Changelog

Since the move to the single credit wallet, the "Create Pitch Deck" step could open an "You're out of credits" prompt even when your wallet was funded: the wizard was still checking the retired per-deck meters, which are now always zero. The check reads your credits now, the same way the server already did, and the balance shown on the first step is your credit balance.

On a 375px-wide phone most panel pages — billing, settings, team, the deck wizard — could be dragged sideways, with tab strips and tables spilling past the edge instead of scrolling inside their own box. The page shell now stays as wide as the screen, so the strips scroll where they should and the page body never moves sideways.

The home page used to describe one product — the pitch deck — while the menu, the pricing page and the credit balance sold four. It now opens on what Pitchko does for an agency or a freelancer as a whole: turn visitors into leads, build the pitch, sell the service. Three step cards replace the old feature grid, and each one links to the products that make up that step — the lead-gen widget and the pitch deck, the technical SEO audit and AI visibility tracking, weekly tracking with white-label and the partner programme. A new "who it's for" section speaks to the small agency and to the SEO or ads freelancer by name, and a short FAQ answers the questions a first visit raises: how this differs from asking a chat assistant, what a credit is, whether the brand on screen can be yours, and whether your data trains anything. Under the hero, the empty logo wall has become a strip of figures read straight from the code — the number of external data sources, the number of products, the single credit balance behind all of them. Every solutions page now says where it sits in that loop and links to the product it hands off to, and lists the freelancer alongside the agency. The free promise changed too: instead of "your first pitch deck is free", every call to action states the credit gift you receive at sign-up, read from the live catalog, and the closing line says what it buys — the mini pitch deck in the free plan. The founder card and the "coming soon" demo placeholders are gone; nothing on the page stands in for material that does not exist yet.

Scheduled AI-visibility scans run for any account that has set a refresh cadence and holds the credits for it — the subscription you are on has not decided that since August. What was still missing was the other half of that promise: when the wallet could not cover a scan, the notice said how many brands went unmeasured, not which ones. An agency tracking a dozen clients could read "2 brands skipped" and still not know whose numbers had gone stale. The notification centre and the banner on the brands screen now list the skipped brands by name, and the brand list's own copy no longer talks about a weekly refresh "on this plan" — a brand with no cadence is simply refreshed when you ask, and one with a cadence follows it.

Billing has a new tab, Auto top-up. Switch it on, pick a threshold and a credit pack, and set a monthly cap: when your balance falls below the threshold, Pitchko buys that pack with your saved card and adds the credits before the next scheduled scan or deck needs them. The cap is not optional — auto top-up will not enable without one, and it never spends more than the cap in a calendar month. Your card is saved on iyzico's own secure form by ticking "Save my card" during a purchase; card numbers never pass through Pitchko, and you can remove the saved card from the same tab at any time. Every automatic charge, and every attempt that fails, is announced to the workspace's owners and admins by email and in the notification centre, so nothing about your money happens silently.

Deck files stop being hosted by our slide provider after a while; that is expected — the file's lifetime belongs to the provider, not to us. The problem was that nothing recorded it: the deck page asked the provider again on every open and got the same answer, while the deck list and the share page never knew at all.

The first time a file is measured gone, it is now written onto the deck's record:

• In the deck list, those decks carry an "export expired" badge — visible without opening each one.
• On the deck page, the note names the day the file was measured unavailable.
• The share page and the download no longer re-ask the provider about a file it has already refused.

API integrators get the same signal: result on GET /v1/decks/{id} now carries export_expired_at (additive; existing fields unchanged) — when it is set, do not retry the links.

The stamp is only ever written on a measured refusal; a transient connection error never marks a deck as expired. Regenerating the deck clears it with the new file — a regeneration is priced by the image source in your deck settings, per the "regenerate" row on the pricing page; research and section writing are not charged again.

And for new accounts: the last step of the setup flow now shows a sample of the output — the four slide types a deck contains and the sections it covers. Nothing is generated on that step and no credits are spent; your first real deck starts when you pick a brand.

The "How it works" section on the home page used to say Pitchko queries 15 data sources and analyses 20 sections. Both numbers were typed in by hand, and both had fallen behind the product: the pipeline runs 17 external fetches and the deck builder carries 26 analysis sections. Those figures are now read from the same code that does the work — every source tile stands for a named fetch in the research pipeline, every analysis tile is a section group with its real count — and a test keeps the page and the pipeline in step, so the next source we add appears on the page without anyone remembering to update a sentence. The hero's third stat, which read "100% AI-generated" directly above a section explaining that the AI does not generate from scratch, now shows the data-source count instead. The sign-up button in the sticky header is outlined rather than filled, so each screen has one primary action. The hero's "~15 minutes" is now "under 30 minutes": measured across every deck with a real research phase, the median run is about 22 minutes and the longest 29, and the panel's own time estimate now starts from the same measurement. The "download example deck" button no longer appears until the file exists, so nobody downloads an empty file named like a presentation.

The embeddable audit tools that have been sitting in your panel for months — the technical SEO audit and the AI visibility scorecard — were visible to the outside world only as a five-word bullet on the pricing page. They have their own page now: that setup is a single line, how a visitor scans their own website, which of the three points you can ask for the email at, why the domain allow-list refuses sites that copied your snippet, and how the leads that come back are listed in the panel and exported as CSV. It sits in the product menu beside the other three solutions.

This is a marketing page, not a product change: how the tools run, what they spend and which plan carries them are all unchanged. What is new is that there is an address you can send a prospect to.

Both products have been live for months: they run in the panel, they have public API endpoints, they have embeddable widgets, and they spend real credits. Their marketing pages had not noticed — the primary button read "Join the early access", the process heading was written in the future tense as "How it will work", and the page ended on a waiting list offering to email you "the moment it goes live". Three sections higher, the same page quoted what that product costs per scan in credits and told you to start with the credits you get at signup.

So the page told a visitor two things and one of them was false. That is fixed: on both pages the primary action is now signup, the process is described in the present tense, and the waiting list is gone. If you had joined that list you have lost nothing — you could already use both products today.

The same false "coming soon" was also sitting in the Help Centre glossary, and that is corrected too. The site's nav badges had been fixed earlier, which makes this the third and last home of a single drift.

The Overview told you what you hold. It did not tell you how fast you are getting through it, which is the question you actually ask before topping up.

The credit meter now carries a line: credits spent per day, over whichever window the period control at the top of the page is set to. Switch it to 90 days and the line follows — the chart and the counts beside it always describe the same span, so you are never comparing a figure from one window against a shape from another.

A few things the line deliberately does not do:

• It does not draw when nothing was spent. A flat line and a steady one look identical at a glance, and "you are spending nothing" rendered as "you are spending evenly" is worse than no chart at all. No spending, no line.
• It does not count refunds. When a charge is handed back, the credits were never consumed — so the spend disappears from the day it happened, the same way it already does in the burn rate on Billing.
• It does not guess. If the reading fails, or there is more spending in the window than one page will pull, you get no line rather than a partial one presented as the whole picture.

If your account has not spent credits yet, the meter looks exactly as it did before. The line appears on its own with the first spend — there is nothing to switch on.

Customers resident in Turkey can now start a Pro or Scale subscription in lira. The amounts are fixed: Pro monthly ₺2,250, Pro yearly ₺22,540, Scale monthly ₺5,720, Scale yearly ₺57,210 — VAT included.

What matters here is not only the currency but that the amount is FIXED. A lira subscription could not be started at all before; one-off credit packs, by contrast, are converted at the TCMB rate on the day you pay, so the same pack can show two different lira figures on two different days. A subscription carries no such uncertainty: the price is frozen at the payment provider, so your renewals are charged the same amount even as the rate moves. The checkout summary now shows exactly the figure your card will be charged.

Nothing changes for customers outside Turkey — they continue on the dollar prices.

SEO Audits now ends with a Watched targets list. Until now a recurring audit was visible only from the report of the site it watches, so the question "which addresses am I paying to re-scan?" could only be answered by opening every audit in turn. Each row here is a standing charge, so the list shows the target, its cadence, and when the next run is due — and lets you pause any of them without leaving the page.

A paused schedule stays in the list rather than disappearing from it. It is the answer to "why did this site stop refreshing", and hiding it would make the pause invisible from the only screen that lists your schedules.

Turning a schedule on, and choosing how often it runs, still happens on the audit report itself. That is deliberate: picking a cadence is picking a spend rate, and it belongs next to the measurement it repeats.

Two things this list will not do. The "next run" is a lower bound, never a date — the scheduler runs on a fixed cadence and covers a bounded number of targets per tick, so a run can land later than the date shown but never earlier. And if we cannot read your schedules, the list says so instead of showing an empty table: "nothing scheduled" is a claim about your spending, and we will not make it on the strength of a failed read.

When your credits ran low, a widget visitor left an email, or one of your commissions was approved, we told you — but only by email. If that email was missed, filed somewhere else, or simply arrived on a day you were not looking, there was no way to find out what had happened.

There is now a Notifications screen. The same notifications, newest first, each one linked to the screen where you can act on it: a credit warning to your plan, a lead to the widget inbox, a commission to your affiliate dashboard.

The list is not a copy of your emails — it is the same record, so the two cannot drift apart. Unread items are marked, and one click marks them all read; that position is stored against your account rather than your browser, so a notification you read on your phone does not come back as new on your laptop.

If you buy in Turkey, our catalogue is priced in dollars and your card is charged in lira. The checkout was built to tell you that before you paid — it worked out the exchange rate, took the Central Bank's published figure, and had the sentence ready. You never saw it. The rate reached the page and stopped one component short of the screen, so every Turkish buyer read $15 and met ₺741,60 on the payment form. Nothing failed, nothing was logged, and the amount charged was always correct — the only thing missing was being told.

That line is now where it was always meant to be, with the rate and the bulletin date beside it, so you can check the number rather than take it on faith.

The rest of the screen changed around it:

• The total stays with you. The order summary sits alongside the form on a wide screen instead of scrolling away above it, so the figure is still there when you reach the last field. On a phone it comes first, before you type anything.
• Three steps, named. Paying means leaving our site for your bank and coming back. The screen now says so up front, instead of letting the page swap itself for someone else's payment box unannounced.
• What happens to your card, in writing. Your card details are typed on iyzico's own page and never reach us; every payment goes through 3D Secure; and the cancellation and refund terms are one click away. All three were already true. None of them was written anywhere you would read.
• The discount box is out of the way. It asks every buyer a question almost none of them can answer, so it now waits behind a "I have a promo code" link. If a referral discount applies to you, it is applied automatically and shown on the summary as it always was — that has not moved.

After you pay, a completed purchase looks like one. All six outcomes used to render the same grey strip, so "Payment successful" and "We couldn't confirm your payment" were the same shape in the same place, and you had to read carefully to find out which had happened. A successful payment now gets a confirmation of its own, says the receipt is on its way, and links you to your invoices.

One more thing, for anyone who already has an account: Choose Pro used to send you to the sign-up form, and since you were already signed in, you were bounced to your dashboard instead. The button you pressed and the page you landed on had nothing to do with each other. It now takes you to your plan.

Your affiliate link was a single address: ?ref=YOURCODE. Post it on LinkedIn, email it to forty people, put it in your newsletter, and you would see that eleven agencies signed up — but not which of the three brought them. There was no way to know what was worth repeating.

Now you can name each share. On the affiliate screen, the build a campaign link card takes a name, lets you pick which page the link opens, and gives you the link to copy. Add UTM parameters if you want it to show up in your own analytics too — Pitchko does not need them to attribute.

The comparison table puts the best-earning campaign first, not the one with the most signups. That is deliberate: a share that brings forty visitors who never pay is not better than one that brings three who do.

Two things have not changed: your existing links keep working exactly as before, and a label never affects attribution — a broken or missing campaign name costs you a label, never a commission.

The same screen also gained a reporting period selector (last 30 days, last 90 days, this year, all time) and a line telling you the latest we will start your payout once your balance crosses the threshold.

The checkout screen has been telling you the right thing for a while. The pricing page and the Distance Sales Agreement were not — and both got the same two things wrong.

First: seven currencies. Both texts said you could "pay with TRY, USD, EUR, GBP, RUB, CHF or NOK". That reads as a menu, and there is no menu and never was: the currency is decided by your billing country. Resident in Türkiye, you are charged Turkish lira; anywhere else, US dollars. The other five were currencies our payment provider supports, not currencies we offer — so anyone who came looking to pay in euros was looking for an option that never existed.

Second: "settlement is in TRY." That clause was appended, unconditionally and in English too, straight after "pay with these currencies". It was meant to describe our side — our provider settles to us in lira — but that is not how it read: to a buyer outside Türkiye paying in dollars, it looked like a promise that their card would be charged lira.

Both now state the rule itself: buyers resident in Türkiye are charged in Turkish lira at the Central Bank selling rate; buyers elsewhere are charged in US dollars. The same sentence is in the pricing section of the Distance Sales Agreement, because that document is a promise and a wrong promise costs more there.

Nothing about what you pay has changed. What changed is being able to read it correctly beforehand.

The contract's two languages had also drifted apart

On the Delivery and Refunds page, the sentence describing what lands in your account when payment clears said two different things in the two languages. The English said "the plan allowance (pitch deck generation credits)". The Turkish still used the older vocabulary — a "usage right" — and never mentioned credits. Two languages of one contract should not name the thing you bought differently. The Turkish now says credits too.

If you sell "we watch your client's site", you had to remember to press the button. An SEO audit was a one-off: you ran it, you got a report, and nothing happened again until somebody logged in. The AI-visibility tracker has refreshed itself on a schedule for a while now — the SEO side never could.

Open any audit report and you will find a Recurring audits card under the score trend. Pick daily, weekly or monthly and we re-run that exact target on your cadence. Each run costs one credit, the same as pressing the button yourself, and appears in your history like any other audit — so the trend line fills in without anyone being there.

Three things worth knowing before you turn it on:

• The schedule follows the site, not the report. Switching it on from an old report changes the same setting you would see on the newest one. There is one schedule per address, so you cannot accidentally pay twice for the same measurement.
• "Next run" is a floor, not an appointment. The card says not before a date and means it: runs are batched, and a busy period pushes some targets to the following pass. It will never be earlier than the date shown.
• Pausing keeps your history and your cadence. It is a pause, not a delete — turn it back on later and nothing has been forgotten.

Two smaller things landed alongside it. Deck download buttons now tell you how big the file is before you click, and a deck that has been regenerated now links to the version it came from — and the original links forward to the newer one, which previously existed with no way to find it.

If you wondered whether something was working, the only way to find out was to ask us. There is now a page at /status where you can see for yourself when our scheduled jobs last ran and how they finished: billing reconciliation, the deck pipeline health check, the account-deletion sweep and the rest.

What you will not find there is deliberate: a green box reading "All systems operational". We have no uptime monitor, no synthetic API check and no incident history — so the page has none either. A green box for something we do not measure would make every other box on the page worthless.

For the same reason, jobs we cannot judge show as "not measurable" rather than "running", and the page states how many of them it could judge. And a job whose last run succeeded three weeks ago is not "running" — it is overdue. An old success says nothing about the silence since.

The per-brand AI-visibility report used to be one long page: the score, then four diagnostic grids, then a ten-domain source list and a fifty-row prompt log. Everything was there and nothing was ranked, so reading it meant scrolling past the parts you did not need to reach the parts you did. It now opens on a verdict — the score, how it moved since the last scan, and the three things to do next — with the breakdown, your competitors and the raw evidence one click away each. It is the same frame the SEO audit report already uses, so an agency running both reads them the same way. The embed widget you put on a client's site gets the identical layout, and every panel still ships in the page's markup rather than appearing only after a click, so nothing became unsearchable.

Underneath it, the trend card can now draw your competitors' share of voice beside your own. Those numbers were already being measured on every scan; they simply had nowhere to live where a chart could reach them cheaply, so the card could show your line and nothing to compare it against. All four lines share one vertical scale — a rival at 40% and you at 3% are drawn as what they are, not as two lines of the same height — and a scan that did not measure a rival leaves a gap rather than a zero, because "not measured" and "no share of voice" are different things and only one of them is bad news.

Two smaller corrections travel with it. The report's run-over-run change is now withheld when the two scans measured different prompt sets: the numbers are still real, but the movement between them is not yours, and the header used to claim it anyway. And the evidence section no longer prints a single count that added cited domains to prompt results — two different things under one number.

Your balance, your plan, your invoices and your billing details used to live at three unrelated addresses in two different sections — and the billing details could only be reached from inside a checkout. They are now four tabs of one section at /dashboard/billing (/panel/faturalandirma in Turkish): Balance, Plan, Invoices and Billing details. The sidebar has one entry for it instead of two. Everything you have bookmarked still works — the old addresses redirect, including /dashboard/credits and the settings page billing used to sit on.

The plan card now also names what you actually pay: the price per period and the amount of the next charge alongside its date. If we cannot read the price for a plan you are on, the card shows the renewal date on its own rather than a figure we are not certain about.

A finished pitch deck now has a Share link button next to its downloads. It creates a public address you can paste into an email, and whoever opens it reads the deck without a Pitchko account, on a page that carries your agency's name rather than ours.

The link is yours to withdraw. Revoke kills it for good — the address stops working immediately and never comes back. Pressing the share button again on a deck that is already shared mints a different link and retires the old one, which is the safe default when you are not sure who still has the previous URL.

You can also see whether it has actually been read: the panel shows how many times the live link has been opened, or says plainly that it has not been opened yet. That count belongs to the link on screen, so creating a new one starts it from zero rather than carrying the old link's total forward.

Three limits, said here rather than found later. A share link expires 90 days after it is created. Only a deck that finished rendering can be shared — one still generating has nothing behind the address yet. And opens are counted at most a few dozen times per hour from the same network, so a client whose whole office sits behind one connection may read as fewer opens than it was.

Your lead inbox now has a delete control on every row, so you can act on an erasure request from the person themselves. Deleting a lead removes their address, the site they submitted and your notes on them, permanently — the audit or AI-visibility report they generated is yours and stays. Editors and owners see the control; viewers do not.

Settings has a fifth tab: Activity log. It answers two questions that until now had no screen — who changed the team roster, and what each action cost. Invitations, role changes, suspensions, departures and removals appear alongside every credit-spending action (a deck created, an SEO audit started, an AI-visibility scan run), newest first, with the person who did it and the credits it consumed.

Owners and admins only. The log names who let someone into the workspace and what the workspace has spent, which is the same thing the billing and invite controls are already restricted to — a viewer sees an explanation rather than a blank table.

Two honest limits, said here rather than discovered later. The log starts from the day it was switched on, so anything older is not missing — it was never recorded. And a "—" in the credits column means the action does not spend credits, which is a different thing from spending zero.

Three changes to the reports, all about reaching the thing you opened the page for.

The cited-source list can be searched and filtered. A single 24-prompt brand routinely produces more than a hundred distinct cited domains, and until now the only way through them was Previous / Next. There is now a search box and four chips — All, Yours, Competitors, Third-party.

The third one is the interesting chip. It is every page that answered the question without belonging to you or to a rival — the earned coverage you can actually go and work on. A competitor's subdomain counts as that competitor, the same way it already counted in the Citation share-of-voice number above it; the list and the score cannot disagree about who owns a domain.

Paging stays. The page still shows ten domains at a time, deliberately: the alternative is a list tall enough to bury the finding it is supposed to support.

A brand with no competitors on file now says so. Before, the competitor share-of-voice block simply did not appear — so a report with no rivals looked identical to one where the section had never existed, and there was nothing telling you a measurement was missing or how to get it. That space now explains what the comparison would show, with a button that opens the brand's settings. Public embed widgets are unaffected: a visitor on your client's site cannot add competitors to your brand, so nothing is offered to them.

The SEO competitor comparison moved into the report's tabs. If you had run a comparison, it used to sit below the entire report — past the issue table, which is often forty rows deep. It is now the second tab, next to the category grid it qualifies, one click from the score. It appears only when a comparison has actually been run.

If you have chosen to take your affiliate commission as Pitchko credit, that credit is now added to your balance the moment the payout is settled, in the same step — nobody grants it by hand any more. Your payout history shows how each settlement was made and how many credits arrived, so you can check the figure against your balance. The email you receive stops talking about bank transfers when there is not one: a credit settlement now says so and points at your balance instead. The 40% uplift, the $50 minimum and the 20% rate are all unchanged.

If you are on a paid plan and have never seen the affiliate card on your dashboard, that was our bug, not your plan. Nothing about who may join has changed — the rule always admitted every paid subscription. What was broken is that you could not see the way in.

Two separate things pointed the wrong way:

The reward card did not render. The dashboard decided whether to show it by checking whether your plan name began with "Pro". When plans moved onto credits, the names changed, and that check quietly stopped matching any of the plans we actually sell. The check now asks the same question the rest of the product asks, so it cannot drift apart from it again.

The terms named a tier. /legal/affiliate-terms said the program was for agencies "with an active Pro subscription". If you are on Scale, you read yourself out of something you were entitled to. The eligibility wording — in the terms, in the panel, and in the pricing FAQ — now says any paid subscription, which is what the rule has always meant.

One more thing that follows from the same fix: if you cancelled your subscription but are still inside the period you paid for, you remain eligible to join, and the card now shows that. It always said so in the terms.

Generated decks now carry a one-line note under each section saying which kinds of research produced it — "site analysis · search data", for example. It is there so the client reading the deck can see where the numbers came from.

The line names the kind of research, never the supplier the data was bought from. That is a deliberate confidentiality choice: which subscription reproduces the analysis is not something an agency should hand to its client. The labels come from a fixed list of categories rather than from the source name, so a data provider added later cannot end up on a slide by itself.

The note is written in the deck's own language, independently of the panel's.

Usage & Credits now answers the question the balance alone never could: what would any number of credits get you. Type a figure and the panel tells you how many standard pitch decks, AI visibility scans, instant audits and crawled pages it covers. It opens on your own balance, so it reads as a continuation of the wallet above it rather than a separate calculator.

Underneath it, the thing that has been true since the credit cutover and was not written anywhere: a pitch deck is priced by how many sections you choose. Drop the section count and the price drops with it, images included. The slider shows the new price and how much less it is than a full deck, so the section picker is finally usable as what it is — a way to spend less, not just a way to write less.

One number on that panel deliberately refuses to fall: research runs in full whichever sections you pick, so the price never goes below 17 credits. Halving the sections does not halve the bill, and the panel says so rather than letting you find out at the charge.

The AI visibility drawer beside it does the same for scans. Choose the number of prompts and how often a brand refreshes, and you get the per-scan price and the monthly cost of that cadence. A one-off scan shows no monthly figure at all — it is not a recurrence, and quoting one would invent a commitment you did not make.

Regenerating a deck also shows its price before you press the button. It always cost less than a fresh deck — the research and the writing are not paid for twice — but until now nothing said what it cost.

When revenue from one of your referrals is refunded, the commission resting on it is reversed — the money has not moved yet, so that stays automatic. What changed is that you are told: an approved commission being reversed now sends an email to your affiliate address, down the same path as the other four affiliate notifications. It used to happen in silence, with the figure in your dashboard changing and no way to find out why.

The email names the commission and its amount. On this notification the amount is labelled "Reversed" rather than "Amount", because the number is money you are not getting rather than money you are. Notifications are deduplicated per commission, so an approval and its reversal do not suppress one another — you see both.

The widget lead inbox is a worklist now, not a mailbox. Every captured lead carries the score of the audit or grade it came from, a link to that report, a status you can move through your own funnel — new, contacted, qualified, won, lost — and a note you can search later. Search covers the note as well as the email and the site, because that is where you write the thing you will want to find again, and the whole inbox exports to CSV in one click. A lead whose report has been deleted shows a dash rather than a zero: a zero is a real and terrible score, and it belongs beside nobody's email by accident.

Reports say more about themselves too. The AI-visibility prompt table now quotes the sentence each engine actually used your brand name in, so "Mentioned: Yes" comes with its evidence instead of asking to be believed; the same excerpt reaches the public API as mention_excerpt. The plain-language summary that only the branded PDF used to carry now opens the report on screen, and the recommendation list flags the quick wins — the high-priority items that are genuinely small jobs. Deliberately absent: an estimated point gain per recommendation. Nothing we run measures what a fix does to the score, and a made-up number is at its most convincing sitting next to four real ones. The SEO-audit PDF gained the two sections it had been dropping: the competitor comparison and the Core Web Vitals behind the performance grade.

When you compare an audit against a rival, that rival arrives one of two ways: you type it into the box yourself, or you leave the box empty and we find one from the market's organic search rivals. The report recorded neither.

A month later, looking at your own report, there was no way to tell which — and these are two different sentences to stand behind in front of a client. "We picked your rival for you" is not "you picked it".

The comparison now says which one it was, under its heading.

On reports created before this field existed, the line does not appear at all. The honest answer for those is that we do not know, and labelling them all "chosen automatically" would present something we never recorded as if we had.

The pitch deck wizard's footer now sticks to the screen: on the long steps — picking sections, browsing templates — "Next" no longer scrolls out of view, so finishing the form does not mean scrolling to the bottom of it to find the way forward.

Choosing a tone now shows you how it sounds rather than what it is called: each of the three carries a sample line the deck could actually contain. The market step states what your choices add up to in one sentence — which market the research runs on, and which language the deck is written in — so it can be read before you submit. A category filter appears once there are enough templates to make choosing worthwhile, and stays away when there are not: a filter above four cards costs a row and answers nothing.

The finished-deck screen gained two things. You can now search the content sections — a long deck runs past a hundred thousand characters, and the browser's own find-in-page cannot see a collapsed card, so it was only ever searching part of the deck; this search matches both the section name and the text inside it. The download row also has a shape now: the deck itself leads, the PDF and the preview sit a step behind, and each button says which file it hands you.

Two small irritations are gone too: claiming a reward no longer makes the card quietly disappear — it tells you how many credits were added — and a few of the wizard's field hints were too faint to read in light mode.

Our prices are in US dollars, but if you are resident in Türkiye your card is charged in Turkish lira. Until now the checkout screen never mentioned it: you read $15 and first met ₺86,22 after you had already been handed over to iyzico's form.

The order summary now shows the lira amount your card will be charged, just below the total, along with the exchange rate used and the date that rate was published. The Central Bank of Türkiye publishes once a day and not at all at weekends, so if you check the figure against "today's rate" and find a gap, the date tells you where it came from.

The amount is worked out by the same calculation that runs when you submit. A figure on screen that could disagree with the figure on your statement would have been worse than showing none at all.

If you are resident outside Türkiye this line does not appear, and nothing changes for you: you are still charged in dollars.

A small piece of honesty about monthly plans

If you are resident in Türkiye and tried to start a monthly plan, your bank may have refused it. That was not your card: our monthly plans are still defined in dollars only at our payment provider, and a local card cannot be attached to a subscription priced in a foreign currency.

We have not fixed that yet, but we now say so before you start. Previously you ran into the provider's own English error message; now you are told up front that the plan is not on sale. One-time credit packs are unaffected — those are charged in lira and work normally.

Until now the buy card on your Credits page showed you a price, not a value. You knew what the pack cost; working out how many pitch decks those credits would actually produce meant a trip to the pricing page and some arithmetic of your own.

The card now answers that where the question is asked — directly under the button: roughly how many pitch decks, AI visibility scans, or crawled pages the pack you are about to buy is worth.

The figures are quoted against a standard pitch deck, and "roughly" is deliberate: a cheaper image source buys more and AI images buy fewer. Promising an exact count here would be a claim the deck wizard contradicts on the very next screen.

When no pack is available for purchase the line does not appear at all. Not a zero, not a dash: there is no pack to describe at that moment, and a blank line would read as copy that failed to load.

When you share an audit report, you can now see from the audit page whether the link was opened — the count sits next to the link, and it says so plainly when nobody has opened it yet. The counter records nothing about the visitor: no IP, no browser details, no per-visit entry. Just a number.

The number belongs to the link that is live right now. Pressing "New link" kills the old URL and resets the count, so the figure you see always describes the link you actually hold.

Share links now stop working after 90 days. We wanted this because an audit is a photograph of a site at one moment: a year-old report sitting at a live address reads as current to your client. You can mint a fresh link from the audit page whenever one expires. Someone opening an expired link sees the same page as someone opening a revoked one, and is asked to get in touch with you.

If your credits run out before a scheduled AI-visibility scan, that scan does not run — and until now the only place that was said was an email. The brands screen now carries a notice with the date it happened and how many brands went unmeasured, sitting above the scores it explains. Without it, a roster whose numbers had quietly stopped moving looked exactly like a roster that was up to date, and that is the worse of the two: a stale figure presented as current is harder to catch than a missing one. Where the record does not carry a usable brand count, the notice says so rather than printing a zero — zero would claim no brands were affected, which is the one thing a skipped scan cannot mean. The notice clears itself once a scan period has passed, so it never outlives the problem it describes.

Every row of your credit history now states the balance you were left with after that charge or top-up, so you can reconcile the ledger against your wallet without adding the column up yourself. The figure is derived backwards from your live balance, and it is deliberately left blank wherever it cannot be supported — on rows from before credits replaced the old per-product meters, and on the legacy rows that count decks rather than credits. A blank there means "no balance to state", never zero.

The purchase screen showed everyone the same ten fields — contact name, postal code, a separate EU VAT number — and left the buyer to work out which of them applied. The form now knows who it is asking: someone buying for themselves from outside Türkiye fills in three fields, with country and email already filled. Ticking "bill this to a company" opens the tax number, contact and postal code. On a domestic sale the tax number is asked either way — the invoice cannot be issued without it, and hiding it would only make the form impossible to submit.

The tax number is one field now. The same number used to be requested twice — once labelled, once unlabelled below the postcode — and when an EU company did not notice the second box, its invoice was recorded as something other than a reverse charge. The field is still named for your country, so a buyer in Germany reads "VAT ID" and one in the US reads "EIN".

One quiet bug closed with it: on a second purchase the address box was filled with your first invoice's address joined together with its city and country, and that joined line was then appended a second time to the next invoice. The address now comes back only from where we stored it as an address; where we are not sure, we leave the box empty rather than filling it with a guess.

Buying from outside Türkiye no longer requires a Turkish tax number. The tax id is now asked for only where the law actually needs it — on a domestic sale — so an agency in Berlin or New York can complete a purchase with the details it genuinely has. The contact-name field is optional too; if you leave it blank we use the invoice name you already gave us.

Your invoice details are also remembered between purchases now. Fill them in once and the next checkout arrives prefilled, instead of ten blank fields every time. You can still change anything before you pay, and the invoice for each sale is always issued from the details recorded at the moment of that sale — editing your profile later never rewrites an invoice already sent.

Every tracked brand now carries its own refresh cadence — once, monthly, weekly, daily, or only when you ask — and you set it on the brand itself rather than inheriting it from your subscription. The picker sits beside market and sector on the add and edit forms, and shows what a single refresh costs in credits before you commit to a schedule.

This replaces a rule that tied automatic scanning to the Pro and Enterprise plans. Scanning has been priced in credits since the credit economy launched, so which subscription you hold no longer has anything to do with whether you can buy a scheduled scan: any account with credits can schedule one, and an account without them is told its refresh was skipped instead of quietly going stale. Brands you have already added keep their current behaviour until you pick a cadence for them — nothing starts spending on its own.

Your credit history can now be narrowed to top-ups or spending, and to any date range — and whatever you are looking at can be downloaded as a CSV. The file carries one row per movement with the date, what it was for, the amount, the unit, and the job it paid for, so a month can be reconciled in a spreadsheet instead of a page at a time. Spends made through an embedded widget on a client's site are marked as such, and each charge now links straight to the deck or audit it paid for. Paging is faster too: reaching the far end of a long history costs the same as reading the first page.

The wallet card now also shows the widget spending ceiling — how much of your balance anonymous visitors to an embedded report may spend in the current period, and how much they already have. That limit was always enforced; it was simply never shown anywhere.

Affiliate commissions now pass through an approval step before they can be paid. When the monthly accrual writes a commission, the row is no longer treated as ready to pay: it joins an approval queue, and a payout can only settle what has been approved. Each approved commission sends the affiliate a notification, so they learn their earnings are cleared without waiting for the transfer.

The balance on an affiliate's own dashboard is unchanged: commissions awaiting approval and commissions already approved both stay inside the pending figure, because the approval is an internal step of ours. The published payout timing is unchanged too — the clock runs from when a commission is earned, not from when it is approved.

When a credit pack or a subscription was refunded, the credits came back out of your wallet — correctly — but nothing was written to your credit history. The balance simply dropped, and the one screen whose job is to explain where credits went had no line for it. That was also the only kind of credit loss you did not cause yourself, so it was the one most in need of an explanation.

Refunds are now recorded like any other movement, with the amount and the date, and they are excluded from the spending-rate figure — money going back to you is not consumption, and counting it would have inflated the burn rate and the runway projected from it.

Until now there was no good way to show a client the audit you ran for them. The embed widget was the wrong tool — it starts a new scan on whatever URL a visitor types, so it could not show the crawl you had already paid for. The only alternative was adding your client as a member of your agency, which gives them your dashboard, your other clients and your credit balance along with it.

A completed audit can now become a public link. Anyone holding the link sees the report — the score, the category grades, the prioritised issues and the PageSpeed panel — with your logo and your colours on it, and nothing of ours if you are on white label. No Pitchko account, no sign-in.

The same link offers the report as a branded PDF, written in your market's language, for the client who wants something to forward or print.

The link is a secret, and you can take it back. Its address is the only credential, so treat it the way you would treat a password: it is generated from 256 bits of randomness, it is never indexed by search engines, and it is not derived from anything else about the audit. Revoking it is permanent — the old address stops working for good rather than being paused, and sharing again creates a different one. A revoked link is indistinguishable from one that never existed, which is what stops anyone guessing their way into somebody else's report.

Sharing costs nothing. It reads an audit you already ran, so it spends no credits no matter how many times your client opens it.

If your account is set to English, the dashboard could stop rendering — a blank stretch where the meters belong, or the page failing outright. Turkish accounts were unaffected, which is why it took a while to surface.

The cause was the Brands tracked meter. Brand slots used to be a ratio — "2 of 3" — and the ceiling was removed when brand tracking stopped being capped. The counter was updated to show a plain count, but the sentence it fills in was put back to the old two-number form by a later change. So the page kept asking for a limit that no longer exists, and rendering stopped there rather than showing a wrong number.

The meter now shows a plain count, which is the honest reading: there is no limit for it to be a fraction of. A new check compares every such sentence against the code that fills it, so the two cannot drift apart again without the build failing.

Also corrected: the English confirmation after a successful purchase said your decks were added. It has said credits since the move to one wallet — the English wording had simply been left behind.

Pitchko no longer stays quiet. Your account managers now get an email when credits run low (at the 25- and 10-credit marks) and when they run out, when a weekly AI visibility scan is skipped for want of credits, when a lead arrives through your audit widget, and when something happens in your partner programme.

Some of this was already visible in the panel — the real change is that you now find out without having to look. The two that matter most: a scan skipped for want of credits produced no error at all, and the panel simply kept showing last week's score; and a lead landing in your widget went cold until somebody opened that page.

Each notification is sent once per period and once per person: the same warning does not arrive on every transaction, but a manager who joins your team later still sees what their colleague saw. Notifications go to the people who manage the account (owners and admins), each in their own language.

Scanning one brand's AI visibility now costs 27 credits instead of 63. Every brand you track monthly costs 36 credits less per scan; on weekly tracking a brand-month is 117 credits rather than 270.

This is a correction, not a promotion. Scans are priced at what they cost us to run, and most of that cost used to be Google's per-search-query fee. When we changed the scan engine so it no longer pays that fee, the price list did not move with it — so for a stretch we billed the scan above what it cost us. That difference is yours now.

The free signup grant also covers a scan, so you can look at one brand's AI visibility while trying Pitchko without paying.

Credits you have already spent are not refunded; the new price applies to scans from here on.

Your own row on the Team screen now offers "Leave agency". Until now every control on that screen was suppressed on your own row: an owner could hand the Owner role to somebody else but could never give up their own attachment, and a member added by mistake had no way out except asking an admin.

Leaving ends your access immediately; the decks, audits and brands you created stay with the agency. If you are the only owner you cannot leave — your row says so, and names the one thing that unblocks it: make another member an owner first, so the agency is never left without one.

Payments made from Turkey are charged to your card in Turkish lira, but the billing history showed the catalogue price in US dollars. For the same payment, the figure on screen and the figure on your bank statement did not agree.

The history now shows the currency and amount the card was actually charged. The TCMB rate used for the conversion and the date of that rate are recorded alongside it, so a refund is always calculated at the rate that applied when the payment was made.

This is a display fix only: no payment was ever charged at the wrong amount, and nobody was over- or under-billed.

Until now your activity history only showed team changes — who was invited, whose role changed. The work that actually spends credits, audits and AI-visibility scans, never reached that list at all. An empty list reads as a quiet week, when what was really happening is that those lines were never being written.

All three are written now:

• Audits — who started one, for which address, and what it cost.
• AI-visibility scans — both the ones you start by hand and the weekly automatic ones. A tracked brand spends credits every week whether or not anyone presses anything, and that was precisely the charge you could not account for.
• Anonymous audits from your embedded widgets — they draw on your wallet, so they belong in your history.

A figure is only written when it was really spent. If a scan stops before it starts, your credits come back and no line appears — so the total in your history stays a number you can hold against your invoice.

The audit screen now tells you before you click

The deep crawl already worked this way: if your budget did not cover it, you saw that before choosing. The instant audit made you find out after typing a URL and submitting. Both looked handled; what differed was the order.

The instant option now behaves the same way — if your credits do not cover it, the notice appears inside the form, with a top-up link.

If your balance cannot be read, nothing is blocked. Saying "you have no credits" would refuse a run to someone who has paid for it, so when we are not sure, we do not stop you.

And when a scheduled scan runs out of credits, we now notice

When a tracked brand's weekly scan was skipped for lack of credits, there was no error to see: the report simply repeated the previous week's numbers. That is now reported to us, so we can reach you about tracking that has quietly stopped.

The Credits screen now shows a spending trend next to your balance: how many credits a day you are using, roughly how long the balance lasts at that rate, and a breakdown of which product is consuming them. The rate is read from the last 30 days and always states the window it came from. When the spending history cannot be read the screen says so rather than reporting zero spending, because those are two different things.

Transaction history now goes all the way back. Previously only the newest 20 movements existed, with no way to reach the 21st — a customer querying an older charge could not find the line. History is paginated on the server now, so you can page back through all of it.

Subscription, purchase and cancellation controls, along with the agency's invoicing identity and payment history, are now shown only to agency owners and admins. Members with the viewer role used to see those buttons and get a refusal from the server on click; the panel now matches the team permission table. Which plan you are on stays visible to everyone.

A new user who is not part of an agency yet also sees a single explanation of that state instead of four separate error messages.

The FAQ on the pricing page still described a refund policy written for packs we no longer sell, and it contradicted the rule in the Distance Selling Agreement. Both pages now state the same single rule: credits you have not spent are refundable within 14 days of purchase, and credits you have spent are not. Partial use does not remove the right, it only reduces the amount.

The same change puts VAT on the pricing page for the first time: prices exclude VAT, and Turkish VAT is added at checkout for buyers resident in Türkiye. Previously that appeared only once you reached the payment step.

Fixed a bug where a credit pack purchase could be approved by the bank without the credits landing in the account. Our payment provider reports the purchased item under a different field name than the one used to create the order, so the verification step could not recognise the item and — by design — refused to grant rather than credit something it could not prove. The correct field is now read, and every completed payment grants its credits immediately.

If you have a completed purchase whose credits never arrived, there is nothing to do: our reconciliation run settles those purchases automatically and moves the credits into your account.

Regenerating a presentation now draws on your credit balance. It is not priced as a full deck: a regenerate skips research and section generation and re-renders content you have already paid for, so you are charged for the image generation alone — between 60 and 155 credits depending on the image source you pick. AI-generated imagery is the most expensive option and stock images the cheapest; the regenerate screen shows what each setting costs.

If your balance cannot cover it, the regenerate stops before it starts and no copy is created, so you are never left with a half-finished deck. The existing one-regenerate-per-deck limit is unchanged.

Pitchko used to meter three things separately — deck tokens, a widget-scan quota and a widget-audit quota — and the public API said so, with a different 402 error code per endpoint. They are one credit balance now, and one balance cannot honestly answer to three words.

• POST /decks returned insufficient_tokens. It now returns insufficient_credits.
• POST /aeo/grade and POST /audit returned insufficient_quota. Both now return insufficient_credits.

This is a breaking change if you branch on the error string. If you branch on the HTTP status instead, nothing changes: 402 has meant "this account cannot pay for the request" since v1 shipped and still does. The full was → is-now table is in §10 of the integration guide.

Nothing else moved. The status codes, the { "error": …, "message": … } envelope and every other code on every endpoint are unchanged. The reason: "quota" value inside engine_failures is a different word — it describes a third-party AI engine's own limit during a scan, not a unit of Pitchko billing, and it is unaffected.

One thing that is not visible on the wire but is worth stating: no v1 endpoint returns an internal failure name any more. Every code now passes through an explicit map, so a state our services invent for their own operational reasons cannot reach your integration until someone gives it a public meaning on purpose.

The SEO analysis screen shows a real figure again. "Pages left" read 0 for every agency — it was summing two counters the credit migration retired, while deep crawls went on working perfectly. It is now derived from your wallet and says so: roughly 1,000 pages per 20 credits.

A deck that fails to generate now returns your credits. The refund was firing correctly but paying into a unit that could no longer be spent. Because we record which part of a charge came from gifted credits and which from purchased ones, the refund restores exactly that split.

Your credit history reads properly. Signup credits, credit-pack purchases and subscription credits had no labels, so those rows would have shown raw system codes instead.

Every card on the reward wall now deposits credits straight into your wallet: the survey pays +4, embedding the widget +10, your first deck +20, and a G2 review +40 — that last one is exactly one mini deck. The feedback (+8) and testimonial (+20) rewards moved to the same unit. Credits you earn show up as their own line in your credit history, so where a balance came from stays readable.

The deck library gained permanent delete. Archive is still archive — a shelf, and reversible — but you can now remove a deck outright, along with its sections, its research data and any feedback it collected. The confirmation names the deck you are deleting and the action cannot be undone. If a deck has a published testimonial, the delete is refused: that testimonial is your public marketing asset and should not disappear as a side effect of tidying a library.

New accounts also start with a working embedded audit widget from day one.

The transaction list on the Credits page only ever showed pitch deck activity. SEO audits, deep crawls, AI-visibility scans and runs of your embedded widgets were spending credits too, with nothing in the list to match them — no line you could point at to explain why the balance had dropped. Every spend now appears under its own name: what it was, when, and how many credits. A refunded charge is not deleted from the list, it is marked as refunded; the spend did happen, and your history keeps saying so. Each row also prints its own unit, because the list still carries your older pitch deck counts and a deck is not a credit. The remaining balance on that page, and the figures on the audit and widget screens, now read from the credit wallet as well — until now they were reading meters that the move to credits had retired, so they all showed zero.

If you opened Widgets since the move to credits, you were told your public embed widgets were not ready to go live, and that your widget credits were used up. Neither was true. The widgets were working the whole time, quietly taking leads.

The meters on that page were still reading the two separate widget balances that existed before credits. Those balances were zeroed when everything moved onto one wallet, and nothing has written to them since — so the page read a real number from a column that had stopped meaning anything, and the readiness banner drew the obvious conclusion.

The page now shows the two figures that actually decide whether a widget can run:

• Credit balance — the wallet everything else spends too.
• Public widget cap — a per-period ceiling on what anonymous visitors may draw from that wallet, so widget traffic can never empty the account your own work depends on.

They are separate on purpose, because the fix is different: one asks you to top up, the other asks you to wait for the period to roll over. A single number could not tell you which.

Nothing was miscounted and nothing was lost — no credits were spent by the bug, and no balance needs correcting. Only the reading was wrong.

The sidebar told you which workspace you were in and never who you were signed in as. If you belong to more than one agency, that was the harder of the two questions to answer. The bottom of the sidebar now shows your name, your email and your avatar, and takes you straight to your personal settings — which is a different place from the agency profile, and now looks like one.

On the developer screen, each API key now reports how many requests it served in the last 30 days, next to the date it was last used. That is the number you actually want before rotating or revoking a key: "last used in June" does not tell you whether something is still calling it. When we cannot measure a key's traffic, the row says so rather than showing a zero — a zero you did not count is the one number you must not act on.

We also corrected the permissions table on the Team screen. It listed API key creation as something every role could do, including Viewer. Creating a key has required an editor seat for a while now, because a key can spend credits, and the table had not caught up — so a Viewer was being told they had a permission the product would refuse them.

The agency profile screen got shorter in three places and gained something in a fourth.

Specialties are tags now. Every specialty you picked sat in its own text field, so choosing ten of them filled half the screen with things you had already chosen — and picking one from the suggestions showed it twice, once as a pressed suggestion and again as a field. Your choices now sit on one line as tags, a suggestion leaves the list once you pick it, and there is a single box for anything the suggestions do not cover instead of one box per entry.

You can reorder client logos. The order goes to the deck exactly as you set it — the first logos are the ones you lead with — but until now the only way to change it was to delete everything after the logo you wanted moved and upload it all again. Each logo now has controls to move it earlier or later.

Every section says where it lands in your decks. Section descriptions either repeated the heading or offered advice like "share your successful projects". They now name the slides the field actually appears on: case studies and testimonials on Cover · Why Us · Closing, certifications and key achievements on Cover · About Us · Closing.

The three fixes at the top of a report have carried a number since last week — "LCP 6.1 s, target 2.5 s". The forty findings underneath them did not. A row said "Structured data missing" and stopped, which left no way to tell a template failing on two pages from one failing on all forty.

Every row states its evidence now. Where the check measures something with a published threshold, you get the reading and the bar: "4.1 s → target ≤ 2.5 s", against Google's own numbers rather than ours. Where it is a yes/no check rolled up across the crawl, you get the share of the sample: "12 of 40 pages fail → target 0". A single-page instant audit says nothing here instead of "1/1 pages", which was true on every row and useful on none.

Findings that cannot move your score — the informational ones — are folded into their own block, so the list above it is the list that costs you something.

Recent audits is a real table. Search by website, filter by status, and page through it, with the score column aligned so two runs can actually be compared. And when you have scanned the same site more than once, the report now draws the score over time underneath it, with every run dated.

The PageSpeed opportunities block is collapsed rather than filling the bottom of the Performance tab, and Google's "Learn more" links are links now — they used to print as raw markup on a report you hand to a client.

All of it applies to the public audit widget your clients see, too.

Once you had produced a few dozen decks, the library stopped being a list and became a pile: scrolling was the only way to find anything, and the only thing telling two decks apart was the brand name. You can now search by name or brand and narrow the list by status (completed · running · failed). Both live in the address bar, so a filtered view is a link you can share and a state the back button returns to. The number beside the title has changed job too — it no longer says how many decks you own, it says which ones you are looking at: 1–20 / 73.

Every deck can carry a name of your own, so five decks for the same brand become "Q3 proposal" and "post-rebrand" instead of five identical rows. The brand name itself does not change, and clearing the field puts the deck back on it. Finished decks can be archived: an archive is a shelf rather than a bin — nothing is lost, and you can bring anything back.

There is also a shortcut. "Use these settings" on any row opens the wizard with that deck's tone, market, template and section choices already filled in. Opening the wizard normally carries your last completed deck's settings forward too — everything except the brand name, which stays blank on purpose so you always type it deliberately.

A deep crawl used to be something you found at the bottom of a finished report — run an instant audit, scroll past the issues, and configure it there. It is a choice about the run, so it now sits on the form where you start one. Pick Instant or Deep crawl before you press the button; choosing Deep starts the crawl straight away, and the report opens already working.

The page cost is stated on that form too, not after. If your balance will not cover a deep crawl, you see that while you are choosing, with a link to top up — rather than filling in a URL and meeting a refusal.

Separately: your embedded widgets now ask for a logo at the point they are about to need one. If neither the widget nor your agency profile has a logo set, the configuration screen says so before you publish, instead of quietly shipping an unbranded widget onto your client's site.

POST /v1/decks returns in milliseconds and the deck takes minutes, so until now the only way to find out it was ready was to keep asking. You can now register an endpoint under Settings → Developers and we will POST to it the moment a deck finishes or fails, an SEO audit completes, or an AEO scan finishes.

Every delivery is signed, so you can verify it came from us: the secret is shown once when you add the endpoint, and rotating it is one click. The endpoint list shows the last delivery and, more usefully, the failures — an endpoint that has been quietly rejecting us for days looks exactly like a quiet week otherwise.

The inbox holding emails your embedded widgets capture used to live under SEO Audit. It has always held leads from both widgets — the audit widget and the AI-visibility grader, and its table has separated the two sources for months — but filing it under the audit meant the AEO side had no link to a screen that was already collecting for it.

It now lives under Embed Widgets, reachable from the header of the screen where you configure the widgets themselves. The old address redirects, so bookmarks and any link you have shared keep working.

An AI-visibility scan queries several answer engines, and until now an engine that failed simply disappeared from the report — while the methodology footer still listed it as queried. A three-engine scan could read as a five-engine one, and nothing on the page or on the API said otherwise. Reports now show a tile for every engine the scan attempted: the ones that answered carry their grade, and the ones that did not carry a plain reason instead. A missing engine is excluded from your score rather than counted as a zero, so a scan we could not complete never looks like a brand that nobody mentions. On the v1 API the report body gained engine_failures alongside engines_queried; if your integration shows coverage, read engines for what was measured.

Live search grounding was 91% of what an AI-visibility scan cost to run — on a 24-prompt scan, a single prompt fanned out into roughly eleven separate search queries — and the provider offers no way to bound that. The Gemini engine now answers from the model itself, which changes what it measures: how present your brand is in what the model already knows, rather than what a grounded answer surfaces today. That is a narrower signal, and a slower-moving one, which suits tracking a trend over months. Two things follow. Gemini no longer contributes cited sources, so citation share of voice now comes from Google AI Mode — the engine built to carry it, with a real URL behind every reference — and from Perplexity. And scans cost roughly a tenth of what they did, which is what makes tracking more brands, more often, worth offering at all.

Affiliate commissions can now be settled two ways, and you pick which one each time: cash in US dollars, exactly as before, or Pitchko credit worth 40% more than the cash amount — a $100 commission becomes $140 of credit. Nothing is applied on your behalf, and choosing credit once does not commit you to it again. Everything else about the programme is unchanged: the 20% rate, the 12-month window, the $50 minimum, and the rule that a balance below the minimum carries over instead of expiring.

The address your embedded widgets are served from was built from your workspace name when the workspace was created, and then there was no way to change it — renaming the workspace left the old name in the URL your clients load. Settings now has a Public address field, so you can put it right. Because the address is already live on other people's sites, it is deliberately separate from the rest of the form: it has its own save, it tells you when it no longer matches your workspace name, and it warns you before a change breaks widgets that are already embedded. Update the embed code on those sites first, then change it here.

Tracked brands have been refreshing automatically for a while, and nothing on the screen said so — a brand measured last Monday looked exactly like one nobody had touched since July. Each row now states when it was last measured and roughly when the next refresh is due, and says plainly when a brand is not in the rotation at all because it is paused or outside your plan's coverage. The dates are estimates rather than appointments, and the list says that too.

The screen is built around the list now. Adding a brand — one at a time or by CSV — moved into a dialog behind the "Add brand" button, so the roster is the first thing you see instead of a form for a brand you do not have yet. When your plan's brand slots are full, the button explains why and points at your options rather than letting you fill in a form that could only be refused, and the CSV panel tells you how many brands will actually fit before you upload.

Smaller things on the same screens: the row icons now say what they do on hover and on keyboard focus; removing a brand asks first and spells out what is lost (and that pausing keeps everything); and the report, branded PDF and prompt-set downloads are gathered into one Export menu that says what each file is for.

Two fixes that reach beyond the brand list. Removing a teammate now asks in a proper dialog instead of the browser's own confirm box — in your language, and with room to say that removal takes away access, not the decks and audits they created. And the "you don't have permission" message no longer says "add brands" when what you were doing was editing one, running a scan, or importing a prompt set.

The audit and AI-visibility embed addresses you copy from the Widgets page now carry your agency id. Nothing you have already placed on a client's site needs to change — the older addresses still resolve, and will keep resolving.

Here is why they moved. The old address was derived from the name you gave your agency when you created it, and then never updated again. If you renamed the agency later, the address kept the old name; if a second agency shared that name, a number was appended to it. Either way you were left with an address nobody could explain. The new one is keyed on an id, so it is not tied to a name and no future decision about names can invalidate it.

The only practical difference: the address you copy when setting up a new embed will look different from ones you copied before. Both open the same widget.

Your audit and AI-visibility widgets now embed with a single script tag — <script src="https://pitchko.io/embed.js" data-pitchko="your-slug" data-type="audit"></script> — and the widget tells the page how tall it is instead of sitting in a fixed 720px box. That box was wrong in both directions: it left roughly 430px of empty space under the starting form, and it cut the full report off behind a second scrollbar inside the widget. The iframe snippet is still there, under "Alternative", for sites whose platform or security policy will not run a script. Alongside this, the allowed-domains list you configure is now enforced by the server: a site that is not on the list no longer gets the widget at all, rather than merely being unable to frame it. Leaving the list empty still means "anywhere", exactly as before.

Signing up now asks for three things: your name, your email and a password. We generate your agency's name and you confirm or change it from the panel in one click. The confirm-password field is gone too — the eye icon beside the field already shows you what you typed.

The moment you verify your email you land in the product. The four-step setup form that used to stand in the way has not been deleted, it has simply stopped being a door: it opens from the panel when you want to complete your profile, and its first question is your agency's website. One link is enough — we work out your description, your service areas and your positioning from your site and put them in front of you to approve.

The verification screen improved as well: the code verifies itself the moment it is complete, the submit button stays locked until then, and the screen says how long the code lasts. The Terms and Privacy Policy you open while signing up in Turkish now open in Turkish.

Your public audit and AI-visibility widgets now let you pick the trade-off between how many leads you collect and how well-qualified they are. Ask up front requires an email before the scan runs — the most leads, and the least qualified, because the visitor has not seen a result yet. Hold everything back runs the scan but shows nothing until the email arrives. Show the score, hold the detail gives away the score and grades and charges the email for the findings — the fewest leads, the strongest intent.

Nothing changes for widgets already embedded on your clients' sites: they keep running the third mode, which is exactly what they did before today. The setting lives next to the lead-capture checkbox on the Widgets screen and governs both snippets at once. Every lead you collect from now on also records which mode was live when it was captured, so changing your mind later does not rewrite the consent context of leads you already hold.

The dashboard's front page has been rebuilt around your day rather than around the product. Anything waiting on you — a deck that failed, a deck that finished with gaps — now appears in a strip at the top, and that strip is simply absent when there is nothing to act on. Below it you can see what is still running, then a single activity feed covering pitch decks, SEO audits, brand scans and captured leads together, instead of a deck list that made the other three invisible. A period selector at the top says exactly which window every figure is counted over. Product cards keep their usage figures but drop the "here is what this is" introduction once you have used that product, and the rewards section has moved to the bottom, collapsed, with anything you have already claimed hidden.

The sidebar now groups your destinations under Products and Agency, and every page opens under the name the sidebar gave it — the SEO Audit entry leads to a page called SEO Audit, and AEO Tracking to one called AEO Tracking. When the sidebar is collapsed to icons, hovering or focusing one shows its label, so the coin, the handshake and the card are no longer three unlabelled circles. The theme control gained a third position, System, which follows whatever your operating system is set to; Light stays the default until you choose otherwise. Help is now reachable from the sidebar itself, and a mistyped or retired address lands on a proper Pitchko page with a way back rather than a blank error.

Three screens sat at addresses that no longer matched their own names. The menu said "AEO Tracking" while the address said brands; it said "Pitch Decks" while an open deck said presentations; and the usage screen answered to tokens in English and kullanim in Turkish, two different words for one page. They now read /dashboard/aeo, /dashboard/decks/… and /dashboard/credits, with Turkish addresses to match (/panel/aeo, /panel/pitch-deckler/…, /panel/krediler). Everything you have bookmarked still works — the old addresses redirect, including links to a specific deck.

You can now edit the questions Pitchko asks the answer engines about a brand directly on the brand page: rewrite one, add one, drop one, and save. Until now the only route was export the CSV, edit it in a spreadsheet, and import it back — which replaced the whole set to fix a typo.

More importantly, the brand page now tells you when a brand's settings have moved away from its questions. Correcting a sector or a market used to leave the tracked questions describing the brand as it was, with nothing saying so, so the score kept being reported as that brand's score while it measured something else. The page now names which settings changed and offers to rewrite the questions — showing you the exact diff, and the current set stays in place until you approve it.

Prompt sets are versioned from this release, and the trend refuses to compute a run-over-run change across a version boundary. Two scans that asked different questions are not comparable, and saying so is better than showing a movement that is really the questions having changed. The version is on the API too, as prompt_set_version on each history point.

An SEO or AI-visibility report used to be drawn in whatever language the reader had the panel set to, while the written analysis inside it stayed in the language it was generated in — so an English panel could show an English interface wrapping a Turkish competitor commentary. Reports now render entirely in their own language, and switching the panel's language no longer rewrites half of one.

Embedded widgets gained two placement controls. ?lang= sets the widget's language independently of your account, so a Turkish agency can put an English widget in front of an English-speaking client; ?w=narrow|standard|wide|full sets its width, which used to be fixed at a size that overflowed a sidebar and looked stranded in a full-width section. Existing embed snippets are unchanged — both parameters are optional and default to today's behaviour.

Settings used to be sixteen sections stacked on one page — your agency's profile, your own notification preferences, your API keys and your account controls, all in the same scroll. Those are four different things, and mixing them meant you had to already know where a setting lived to find it. They are four tabs now: Agency profile (everything that shows up on a pitch deck), Billing, Developers (API keys, at /dashboard/settings/developers — an address you can link to), and My account (your language and timezone, which emails you receive, and data export or deletion). Each tab is its own address, so a bookmark now lands on the section you meant rather than at the top of everything. /dashboard/settings still works and opens the agency profile.

An SEO report used to arrive as one long page — a score, then a category grid, then an issue table that could open at forty rows, then the performance panel, all at the same visual weight. Everything was there and nothing was ranked, so the first question a client asks in a meeting ("what do we do about it?") was the hardest one to answer from the screen.

The report now opens with the score, and directly under it the three things worth fixing first. Where a finding has a number, it carries one: "LCP 6.1 s, target 2.5 s" instead of "your site is slow" — Core Web Vitals that the audit has been measuring all along but never showed you. Everything else moved into tabs: Overview, Issues, Performance, Deep crawl.

Run the same URL twice and the score now says how it moved since last time, dated. A first scan shows no change indicator at all rather than a zero, because there is nothing yet to compare against — and the comparison is only drawn between runs of the same kind, since a single-page instant audit and a ten-thousand-page crawl are not measuring the same thing.

The public audit widget your clients see uses the same layout, minus the parts that belong to you: no export controls, no deep-crawl tab.

The heading visitors read at the top of your embedded widget was ours, and you could not change it. Widget settings now has a heading and sub-heading you write yourself; leave them blank and the product's wording stays. Two other things on that screen changed with it. The domain list now says what leaving it empty actually means — anyone can embed your widget on any site, and your quota is spent either way — so an unrestricted widget no longer looks the same as an unconfigured one. And the bot check now matches the widget instead of following your visitor's system theme, so it no longer appears as a stray dark box on a light page. The green "your widget is live" banner is gone: it reported the absence of a problem while sitting above a warning, and the screen is easier to read without it.

The embed widget settings now show a live preview beside the form: change a colour or a heading and the visitor's first screen updates as you type. It renders exactly what your plan actually publishes, so on the free tier you see the standard styling your visitors get rather than the customisation the fields would suggest. Underneath it, a funnel counts the last 30 days — widget loads, audits started, audits completed and leads captured. Where there is not enough traffic to make a percentage mean anything, the card says so instead of printing a confident number: a rate with no one behind it is worse than no rate. Load counting starts from this release, so the first stage fills in over the coming month.

You can buy brand tracking for 1, 3 or 5 brands directly from the pricing page or your billing settings. It was priced in public before this and could not be bought anywhere.

The delay was not the checkout. The packs were being sold as a monthly subscription you could pause and cancel — on the pricing card and in both binding legal documents — while the product had already become a one-off that grants 30 days and then simply stops. Opening the link first would have left a contract describing something we do not sell, so the terms were rewritten in the same change: the distance-selling agreement and the terms of use now state the 30-day window, that it does not renew, and that when it lapses tracking stops while your brand and its history are kept.

Two things follow from that. The checkout tells you the window before you pay, not after. And you can withdraw within 14 days for as long as no brand has been put under tracking and no scan has run — the same rule that already applied to pitch deck packs, rather than the immediate-performance rule that makes SEO page packs non-refundable.

Every playbook on the blog now exists in Turkish. Until today only the launch post did, so a Turkish reader arriving at /tr/blog found one article where an English reader found nine — proposal guides for SEO, web design, social media and multi-channel work, plus the pieces on finding clients for each. Category listings are addressed in Turkish too (/tr/blog/kategori/playbooklar rather than the English identifier they used to carry); the old addresses redirect, so nothing you have bookmarked or linked breaks.

Turning an AI-visibility report into a deck is now free once per brand slot, per period. Three brand slots means three free conversions a period; past the allowance a deck credit is charged exactly as before. Buy a brand add-on and the allowance grows with it, because the counter rides on the unit we already sell.

It works this way so no new currency had to be invented: a fractional charge — "an AEO deck costs 0.3 decks" — would turn your balance into a number you cannot reason about. This is a discount on the first N, not a third meter.

Brands can now be paused. A paused brand drops out of the weekly sweep and stops occupying a slot, while its row, its prompt set and its entire scan history stay exactly where they were. Removing a client who was between engagements used to take the trend series you would want back with it — pausing is the non-destructive version. If your slots are full when you try to resume one, the panel says so specifically, because the fix is specific: pause a different brand, or buy an add-on.

One smaller annoyance is gone too: with no deck credits left, the "convert to deck" button still looked clickable and failed on every attempt. Your balance is now checked before the click.

The pricing page listed three sizes for each add-on pack but only ever linked to the cheapest one, so the 5- and 10-deck packs and the 100,000- and 250,000-page packs were priced in public and could not actually be bought. Every tier is now its own row with its own checkout link. Your billing settings gained an add-on section too — previously the only pack link there was a single deck, and it was shown to free accounts only, so a Pro subscriber could not buy more decks from the dashboard at all.

Refund messages got more specific at the same time. Every refusal used to read "the 14-day window has passed, or the credits were used" — a guess offered to someone who already knows which it is, and for several cases both halves were simply wrong. Each reason now has its own sentence, including a new one for a purchase whose credits have not reached your balance yet, which previously said you had already used them.

And a promo code entered while the discount service was unreachable used to be reported as an invalid code — a claim about your code rather than about our lookup. The checkout now tells you the check itself failed.

Pitchko sells three things in three different units — pitch decks by the deck, SEO analysis by the page, AI visibility by the brand — but the overview only ever showed the first, so the other two balances were invisible unless you went looking for them under Usage & Credits. All three now sit together on the front page, with the same wording that page uses. The deck figure is called "deck credits" rather than "tokens", which is what it has always been; a balance that cannot be read shows a dash rather than a zero, because "we could not check" and "you have none left" are not the same thing. Alongside that: a reward you have earned is now marked as ready and its button lines up with the others instead of jumping, a locked reward tells you where to go to unlock it, and decks generated from an SEO audit or an AI-visibility report are labelled as such in your deck list. On a deck's page, "Rate this deck" has moved up into the header and the jump-to-section bar starts collapsed, so it stops sitting between you and the deck.

Two older faults went with it. The language switcher in the sidebar was painted for a dark background the panel stopped having, so in light mode its label was white on white — readable in dark, invisible in light. And if we briefly could not count your decks, the list quietly sent you back to page one and hid the pagination, which looked exactly like having only one page; it now keeps you where you were and simply does not claim a total it could not measure.

You can now buy extra AI-visibility brand slots. The 1, 3 and 5 slot packs have been priced on the pricing page for a while and were not actually purchasable; they are now, each for 30 days, and overlapping purchases stack — buy 3 today and 1 next week and you hold 4 slots that expire on their own dates rather than on a single shared one.

Tracked brands also gained an on/off switch. Pausing a brand keeps the row, its prompt set and every scan it has ever produced — it simply stops being refreshed weekly and stops occupying a slot, so you can rotate a client out for a quarter and back in without throwing away their trend line. Reactivating is refused, rather than silently allowed, when it would put you over the slots you hold.

Converting an AI-visibility report into a deck is now free once per brand slot per period, instead of always costing a deck credit. Past that allowance it costs a deck credit exactly as before.

One correction that goes with this: the slot limit is now enforced by the database on every path that adds a brand. CSV import never checked it, so a roster larger than the slots you hold could be uploaded — those brands were never going to be tracked. An import that would exceed your slots is now refused as a whole, and says so, rather than landing brands that sit idle.

The report reads as a report rather than a stack of tables. The score and its composition lead, the recommended actions come next, the per-engine and per-category breakdowns support them, and the raw cited-source and prompt-level logs sit behind a disclosure instead of dominating the page. Each of the four metrics now carries an explainer — what it measures, exactly how it is calculated, how to improve it, and how much of the composite it carries — and the metric bars were replaced by a view of the points each metric contributed against the points it could have. That makes an unmeasured metric visibly different from one measured at zero, which the old bars drew identically. The branded PDF was rebuilt around the same figures: a summary page with the score dial, the composition and deterministic key findings, then a detail page adding competitor share of voice, sentiment and the most-cited domains. A white-label export no longer falls back to Pitchko's brand colour when an agency has not set its own.

Your two public embed widgets — the SEO audit form and the AI-visibility grader — now have their own entry in the main sidebar, alongside Pitch Decks, SEO Audit and AEO Tracking. They were previously three levels down under Settings, with no menu entry at all, which meant most agencies never found the one feature that puts a lead-capture form on their own site. The SEO Audit screen now points at them too. Your existing bookmarks and links still work — the old address redirects to the new one, in both Turkish and English.

The affiliate programme is now a two-sided offer. You earn 20% of the qualifying payments made by every agency that signs up through your referral link, for their first 12 months — and that agency gets 10% off its first pitch-deck or SEO page pack, applied automatically with no code to type. The discount covers one-time packs only, not a Pro subscription.

The pages describing the programme had been quoting 30% with no mention of the discount, which was never the offer we agreed. The discount half has been live since 2 August; the commission rate is now what the terms say it is. Commissions already accrued are untouched — each one records the rate it was calculated with, so a rate change only ever applies from the change forward.

Pitchko now measures how AI answer engines talk about the brands you manage. Track visibility, sentiment, and competitive position across engines from a single roster.

• Multi-brand roster — add every client brand once under AEO Tracking; scheduled scans keep scores fresh without manual runs.
• Honest scoring — 0 means measured-and-invisible, null means not measured. Empty runs never fabricate a score, and methodology ships with every report.
• White-label PDF reports — export client-ready visibility reports under your agency's own branding.
• Public grader widget — embed an AI-visibility grader on your site and convert visitors into leads, mirroring the SEO audit widget.

Agencies can now embed a white-label SEO audit widget on their own website. Visitors enter a URL, get an instant technical audit, and become qualified leads in your dashboard.

• One-line embed — an iframe snippet configured from Settings → Widgets, locked to the origins you allow.
• Lead capture built in — unlocking the full report asks for a work email; leads land in your Audits → Leads queue with the report attached.
• Separate quota — widget traffic draws from its own audit pool, so public embeds can never drain your team's internal audit quota.

Anti-abuse rate limits and per-origin controls ship enabled by default.

Pitchko now has a public REST API. Generate pitch decks from your own tools — CRMs, internal dashboards, partner integrations — with a single authenticated request.

• API keys are issued per agency from Settings → API Keys, with scoped permissions and instant revocation.
• Sandbox mode lets you integrate end-to-end without spending deck credits — sandbox decks are watermarked and free.
• Deck status polling via GET /v1/decks/:id returns snake_case JSON with a stable, versioned wire shape.

All v1 responses go through additive DTOs: fields are published deliberately, never leaked from internal types.