A plain-language record of what we have shipped recently. We are pre-launch and build in the open, so this is honest about what is live, what is opt-in, and what is still rolling out. For the full in-app history, open Help and Docs once you have an account.
The new look is now the standard site
v5.23.0·26 August 2026
ImprovedThe redesigned marketing site (landing page, features walkthrough, AI Readiness Grader, pricing and comparison pages) that was in preview is now the standard experience at the main URLs. The previous design is still reachable for reference.
FixedFixed brand-name detection for sites with a hyphenated tagline (e.g. 'the all-in-one platform'): it now picks the real brand name instead of a fragment, which also improves the suggested competitors.
Experiment tracking and internal cost work
v5.22.0·26 August 2026
Behind the scenesThe marketing A/B test now tracks itself properly: visitors who arrive through the redesigned /clay site keep that identity as they move through the funnel (grader, pricing, sign-up), so we can compare the two versions fairly. The current site and its visitors are unaffected.
Behind the scenesWider internal testing of a lower-cost model: our own QA accounts now run every generation step through the lower-cost model so we can validate it end to end before rolling it out more widely. Your account is unaffected.
A leaner free plan that still shows you something, and a new look in preview
v5.21.0·25 August 2026
ImprovedA free analysis now keeps your headline AI-visibility score, its trend, and your single highest-impact opportunity, and shows a clear 'unlock the full prioritised fix list on Pro' lock where the rest would be, instead of empty space. Behind that lock is a plan we simply did not generate for the free tier, never fabricated data.
NewWe are testing a refreshed look for the marketing site: a redesigned landing page, a features walkthrough with live product animations, an AI Readiness Grader and a pricing page, all under /clay. The current site is unchanged while we compare the two.
NewThe public blog got the new look too, with a hero image on every post.
ImprovedNew accounts outside our own team now open in Simple mode, so the first run is cleaner; you can switch to the full view any time.
Behind the scenesCheaper free-tier analyses: free internal generation now routes through a lower-cost model with an automatic fallback to the usual one, plus a concurrency throttle so those calls do not collide, lowering our cost to run the free tier without changing what you see.
A free trial worth having, and a free plan that still shows you something
v5.20.0·24 August 2026
NewEvery new account now starts with a 7-day free trial of full Pro access, no credit card, with a countdown and a heads-up email before it ends.
NewWhen the trial ends you keep a free Watch plan: your headline AI-visibility score, its trend, and one fresh opportunity each week stay free; the per-engine breakdown, competitor detail and the rest of your opportunities sit behind an upgrade rather than being taken away.
NewTrial reminder emails on day 5 and day 7, so nothing changes silently.
ImprovedTeam seats now match the plan: Free 1, Pro up to 3, Agency up to 10, enforced when you invite teammates.
ImprovedThe organisation rollup, per-client approval routing, and additional client workspaces are now enforced on the Agency plan, matching what the pricing page says.
A grade you can see, and emails worth opening
v5.19.0·24 August 2026
ImprovedThe free grader now shows your A to F grade inside a real score dial, with a bar breaking down how many checks passed, need work or failed. It used to be a bare number, and the dial is coloured by band so a failing grade is never shown in brand green.
ImprovedEvery transactional email was redesigned: a cleaner header, one clear action with the full link shown as a fallback, a security note where it matters, summary cards on the weekly digest, and a proper footer. Billing notices carry no unsubscribe link, because you cannot opt out of a receipt.
ImprovedThe grade page now says plainly that the full grade is free with no signup, invites you to run a competitor next, and explains that a good grade means AI engines can read you, which is not the same as them naming you.
NewAfter you email yourself the report, a couple of optional questions (who you are, how many brands you look after, whether you measure AI visibility today) tailor where you go next. Every question can be skipped, and they only shape your marketing profile if you opted in.
NewA new public benchmark page shows how graded brands compare once enough have run the grader. Below the 200-domain minimum it says so honestly rather than showing a made-up number, and the dataset stays downloadable under an attribution licence.
NewSharing a grade link now shows the real grade in the preview card on Slack, X and similar, instead of a generic ClarAI card.
One real answer, and pages that are actually themselves
v5.18.0·22 August 2026
ImprovedThe free grader now shows you one real AI answer and whether it named your brand. One prompt, one engine, one sample, and it says so: a grade of A never meant an engine actually names you.
FixedThe registration and login pages were serving the homepage to search engines and link previews, along with our ad landing page. A check now fails the build if a public page is left out again.
FixedEvery page told Google it was a duplicate of the homepage. All 42 now identify themselves, including every blog post and comparison page.
FixedAI Simulation ran Gemini twice and labelled one of the two answers "Google AI Overviews", which is a different product we do not query there.
FixedAn agency signing up landed in Simple view, where the client workspaces, rollup and reports they had just paid for were all hidden.
ImprovedThe dashboard's four competing tiles became one chart you can read. Click a reading to change the chart; sample size and coverage stay visible.
FixedError messages stopped telling you to check the console. 213 of them now say what failed in plain words and what to do next.
ImprovedThree alternatives pages for Profound, Peec and Otterly, each opening with where that competitor is genuinely the better choice.
FixedThe grader said you had been added to the mailing list even when no mailing list was configured. It now reports what actually happened.
ImprovedA campaign click that becomes a signup three days later is now credited to that campaign instead of to "direct".
Saying what we actually measured
v5.17.0·21 August 2026
FixedPrompts we never sent to an engine are no longer counted as "not present". Those are different claims, and merging them overstated how much of your brand had been checked.
ImprovedStep 4 of an analysis now shows progress inside itself, and the run records which markets it covered, with a guaranteed floor of prompts per market.
FixedAn analysis killed partway through (by a deploy, say) used to sit at "running" forever. It now flips to failed within ten minutes so you can re-run it.
FixedScrolling is fixed on five screens that were always slightly too tall, which left a permanent scrollbar and dead space below the content.
ImprovedNine industry figures on our homepage carried no link and no date, five of them credited to a competitor. Six are gone; the three that stay name their institution, sample size, date and a link that resolves.
ImprovedThe homepage now says on page one that we are early access with no customers yet. That line already shipped on seven comparison pages.
ImprovedPricing is in the top navigation, and every nav item is visible on a phone.
FixedThe Agency plan stopped underselling itself: white-label PDF export ships today and carries your agency name. Your logo on it, and client portal links, are still to come.
FixedPrivacy, terms and legal now serve their own page. They were registered but never pre-rendered, so a crawler or a direct link received the homepage instead of the policy.
ImprovedIllustrated empty states, per-post blog artwork, one chart style across the product, and the monospaced typeface our figures were meant to use finally loads.
ImprovedWe now measure our own funnels first-party: no third-party tracker, no raw IP stored, Do Not Track and Global Privacy Control honoured, disclosed in the privacy policy.
Teams, your own domain, and email that sends
v5.16.0·21 August 2026
ImprovedInvite teammates by email, even before they have a ClarAI account. They get an invitation, land on an accept page, and join by signing in or registering with that address.
ImprovedClarAI now lives at myclarai.com. The app, its links and its emails all point at the real domain.
FixedPassword reset, email verification and invitation emails now actually send. Before this, a reset link only went to the server logs.
ImprovedEmails were redesigned: centred, one clear action, expiry and security notes in their own panel, and they render properly in dark mode.
ImprovedOperators get a Users tab in Admin to change a plan, promote, suspend or delete an account, with guards against locking yourself out.
FixedEnergy-drink brands like Red Bull now detect correctly and get real competitors instead of a generic beverage list.
FixedEmails no longer link to a domain we do not own, so the account-settings link and the buttons in digest and alert emails now go somewhere.
FixedThe chip shown while we work out your industry is readable instead of a blank pulsing pill.
A simpler, more consistent first run
v5.15.3·18 August 2026
ImprovedSigning up no longer auto-starts your first analysis. You start it yourself from the dashboard, the same as a new project.
Improved'Run your first analysis' now lives inside the Get Started checklist, so there is one clear place to begin instead of two cards.
ImprovedRemoved the duplicate Get Started button from the sidebar (the journey page is still there).
ImprovedThe AI Visibility result now shows a plain-language one-liner in Simple mode, so you get the takeaway without reading an info point.
ImprovedTrimmed the extra hover text on the signup goal cards.
QA fixes from Maria's Test 3
v5.15.2·18 August 2026
Fixed'Review' in Get Started now opens your results instead of starting a new analysis.
FixedThe new-project card shows your brand logo instead of a rocket when a favicon is available.
FixedYour detected industry now shows as a readable label, not a raw code.
FixedThe 'Prompt distribution' info point now shows its explanation on hover.
FixedThe AI-discovered competitor list keeps real brand names only, no more phrases like 'Sizing Advice' or 'Dev Mode'.
ImprovedClearer benchmarks label, score tooltip, and prompt-count wording so the numbers read as one consistent story.
Public-site deploy fix and a crawler-accuracy safeguard
v5.15.1·17 August 2026
FixedFixed the public site's deploy configuration so it builds and publishes correctly.
ImprovedAdded a safeguard so ClarAI can never silently overstate which AI crawlers it has verified.
A Weekly Plan you choose, Get Started in the Dashboard, and authority-building
v5.15.0·17 August 2026
ImprovedWeekly Plan you choose: generated tasks now arrive as a 'Suggested this week' list. Your active plan starts empty and you add the ones you want (one at a time or all at once), so nothing reads as overdue before you decide.
ImprovedGet Started now lives at the top of your Dashboard until you finish setup, then collapses to a small 'Setup complete' mark.
ImprovedWhen you draft an article for a brand, the Content Engine now suggests internal links to your own pages, real sources to cite, and where to place images with alt text. It never invents a link, a source, or an image.
ImprovedOur own blog posts gained real internal cross-links, verified sources, image placements and honest author bylines.
ImprovedCrawler verification now also confirms Googlebot, Bingbot and Applebot by forward-confirmed reverse-DNS, not just published IP ranges.
FixedThe first-run 'setting up your workspace' card now says roughly how long the scan takes, so it never looks stuck.
FixedA tracked competitor that was not named in the prompts tested this run now says so, instead of showing a bare dash you cannot interpret.
ImprovedRealtime Reddit finds now also appear in the notification bell.