Changelog
What is new in ClarAI.
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.
Clearer plan limits and paying accounts out of the beta
- ImprovedAnalysis limits are Pro 80 and Agency 250 a month; automatic checks count as one each.
- ImprovedThe free plan checks automatically every week, or every 14 days when your results are stable.
- NewPaying Pro and Agency accounts leave the beta and open every screen their plan includes.
- ImprovedDeleting your account cancels your subscription straight away; if that cannot be confirmed, nothing is deleted.
- NewTeammates on a paying owner's project get the API and MCP connection for that project.
- ImprovedThe Claude engine answers with Sonnet 5.5 and Google AI falls back to Gemini 3.6 Flash; trends mark the new starting point.
- ImprovedThe free readiness grade no longer scores Markdown negotiation, WordPress refresh handles longer posts, and Weekly Plan tasks run up to 15 an hour.
Turn off usage measurement, and emails without capital-letter trouble
- NewA switch in Settings lets you stop ClarAI recording how long each screen is open. It is on by default; when off, nothing is recorded.
- FixedSigning in, signing up and resetting your password now ignore capital letters in your email address.
Clearer privacy and tighter confirmation links
- ImprovedThe privacy policy now says we record how long each app screen is open and in use, linked to your account, to improve the product: the screen's name only, kept 13 months, and nothing under Do Not Track or Global Privacy Control.
- ImprovedEmail confirmation links sent before 15 October have expired; ask for a new one from the banner in the app if you still need it.
See who each answer named instead of you
- NewOn AI Visibility and AI Mentions, each answer lists the companies it named: "Named instead of you" when your brand is not in it, "Named in this answer" when it is. An engine that did not answer shows no names.
- NewThe free check's sample answer lists the companies it named, with a caption saying it is one question asked once of one AI model through its API: a sample, not a measurement.
- ImprovedThe competitor reading lists every option an answer gives, including names in headings, tiers and tables, and never a name the answer does not contain. It applies to new analyses.
- ImprovedThe email confirmation link signs you in, once, on desktop or phone. Accounts with two-factor authentication, and browsers signed in to a different account, sign in as usual.
- ImprovedReaching the trial's project limit now names the trial's allowance instead of the free plan.
- ImprovedCampaign pages from our ads use the current site design, two new pages are added, every button keeps where you came from, and check pages label their buttons "Run the free check". A 50/50 layout test is built but switched off.
A live first analysis and a check before publishing
New8 changes
- Your first analysis shows your real data while it runs: your site, competitors, personas, the questions asked and each engine's answer as it arrives.
- Your score and answers open on the dashboard as soon as the engines have answered; opportunities and the weekly plan follow when the analysis finishes.
- A pre-publish check in the Content Engine: a person checks the facts, title, meta description, structured data and image alt text before a draft goes to WordPress, Webflow, Notion or Resend, or is scheduled in bulk.
- Refreshing a WordPress post with AI now writes a new draft for your approval and never changes the live post.
- Citations shows whether the engines' search reached your site (not retrieved, retrieved but not cited, or cited), how long your cited pages hold, and sources per market.
- Owned facts: ClarAI checks that your site states your pricing, trial, support and markets plainly, matched on your Knowledge Base.
- Brand accuracy is open on Pro, Agency and the trial, runs after each full analysis, and shows the pages each answer cited.
- Type in your Merchant Center AI performance numbers on the First-party AI data tab and check your product pages for its popular attributes.
Improved6 changes
- Outside events, such as a Google update or a data provider incident, are shown on trends, recaps and verdicts to place them in time, never as a cause.
- Where Google AI Overviews readings are on and a collection looks incomplete, AI Visibility says so and leaves them out of the score.
- The Tech SEO audit explains how to allow AI crawlers by user agent and by their operators' current published IP files.
- ChatGPT with web search and sources is ready but off for now; when we switch it on, the ChatGPT trend starts a new series.
- One 7-day trial per email inbox, and Weekly Plan drafts have their own monthly limit (20 Free, 500 Pro, 1,500 Agency) plus an hourly limit.
- A quicker sign-up from ads, with a note for the Instagram and Facebook in-app browsers, and new single-message campaign pages.
Fixed2 changes
- Search Console data follows project access, scheduled analyses are claimed when they start, and a successful site audit clears an earlier error.
- Weekly Plan task links open the right task, chart subtitles name only engines that answered, Perplexity is never called, and Ahrefs and DataForSEO price notes are corrected.
Bug hunt and up-to-date facts
- FixedAPI keys and notification channels follow project access on every use: removing someone from a project or its workspace, or making them read-only, stops their keys and notifications there.
- FixedContent Engine limits hold when several drafts start at once, two analyses started together cannot both pass the monthly limit, and re-runs share one hourly limit.
- FixedIntegrations, the WordPress refresh and Google or Reddit sign-in check that you can edit the project; beta accounts stay inside the beta, including through the API.
- FixedA failed engine call never counts as not recommended, and recaps, reports, digests and alerts only report a change when the two analyses are comparable.
- FixedTrends, question buckets and tone read the same latest valid answers as AI Visibility, and competitor rates count each answer once.
- FixedBackground workers, site audits and page checks are more reliable, and accepting an opportunity on the dashboard updates the Weekly Plan straight away.
- FixedThe Projects screen lines its cards up in equal columns at every width, without the Beta label.
- ImprovedCompetitor prices, plans and engine lists re-read on the vendors' pages on 13 October 2026, with every change listed on /corrections.
- ImprovedNew AI crawlers from Meta, Amazon, Mistral and Google are recognised, and visits from the published ranges of OpenAI, Google, Microsoft, Apple and Anthropic are verified.
- ImprovedGoogle guidance, Core Web Vitals labels and schema advice now say what the sources say today; Perplexity stays off since its API ended.
Fair limits for drafts and community searches
- ImprovedDeleting a project no longer gives back the Content Engine drafts or the community searches it used this month, the same rule analyses and simulations already follow.
Fair limits and honest unfinished simulations
- ImprovedDeleting a project no longer gives back the simulations it used this month, the same rule analyses already follow.
- ImprovedA simulation that stopped before its results were saved, for example during a service restart, now says it did not finish and offers to run it again, instead of blaming Anthropic.
- FixedThe WordPress audit saves each page's results before fetching the next one, so it no longer holds the database while it reads your site.
Fair simulation limits, steadier simulations
- ImprovedDeleting a simulation no longer gives back its place in the month's allowance; a simulation still running appears in history once it finishes.
- FixedAI Simulation and page checks no longer hold a database connection open while they wait.
Steadier under load
- ImprovedRe-running personas and questions too often now shows a calm note with how long to wait, on Agents and Foundations.
- FixedPicking a draft type, running a Weekly Plan task and refreshing a WordPress post no longer hold a database connection open while the AI works.
Fair limits and honest notes for the beta
- ImprovedRe-running personas and questions is limited to five times an hour, and deleting a project no longer gives back this month's analyses.
- ImprovedIf your last analysis did not finish, the dashboard says so, and that it does not count towards your monthly analyses.
- ImprovedAI Simulation examples no longer assume you sell software to startups.
- ImprovedIn the beta, notifications only link to screens you can open; the dashboard bell opens the same places as before.
- FixedThe Content Engine no longer holds a database connection open while it writes a draft.
The beta: focused screens, Give feedback and honest limits
- NewDuring the beta, accounts open the core screens in the Simple view; screens outside the beta say so plainly and take you back.
- NewGive feedback from any project, with the screen you are on. We never post your message or email address anywhere else.
- NewAn honest Beta label that lists what is limited today, and a welcome email with three things to try first.
- NewDelete your account yourself from Settings.
- NewNew guides: how ClarAI measures, and how to show up in ChatGPT.
- ImprovedPricing states the trial exactly: seven days with Pro's monthly limits once you confirm your email, no card; chart PNG export is marked coming soon.
- FixedThe MCP address in Settings is https, the draft badge names your real brand style (or says it was not detected), and "Saved" times are right after a draft is written.
Page drafts that finish, and a weekly summary that counts right
- FixedLanding, pricing and case study drafts no longer get cut off after minutes of waiting: pages keep to a clear size and have more room.
- FixedIf a draft still cannot be finished, a message with Try again stays on screen; nothing is saved and it does not count towards your monthly drafts.
- Fixed"Since your last analysis" counts only the latest analysis's answers (it could add the quick results to the full ones).
- FixedThe Content Engine no longer says "Not in your Weekly Plan yet" while it loads a task that is on your plan.
First analyses no longer stall on some websites
- FixedReading your brand's colours, fonts and spacing could hold a first analysis for about twenty minutes on some stylesheets and then fail. It now takes about a second, with the same results.
Decisions applied: honest engines, paid API access, publishing that follows your plan
- FixedEngine count corrected: ClarAI reads ChatGPT, Gemini and Claude today; Google AI Overviews is built but not switched on yet. Every correction is listed at /corrections.
- ImprovedPages that belong to a Weekly Plan task follow it: Implemented asks for the live address and records it on the task, and an old tab can no longer overwrite newer statuses.
- NewEdit the HTML of a page draft as a new version; Add real data fields removed without a value are flagged before you save.
- ImprovedA warning before writing a case study with no proof in your Knowledge Base, and one partner pitch per site, sent from your own mailbox.
- NewOptional "Who are your customers?" in onboarding and "How did you hear about us?" at sign-up.
- ImprovedA heads-up at 80% of your monthly limit, and faster personas, questions and plans on a newer Claude model.
- ImprovedAPI and MCP access are part of Pro and Agency; chart PNG export is on every plan; unlimited brands on Pro within the monthly analysis limit.
- NewNew /corrections and /press pages; no SSO or signed SLA promised on pricing.
- NewTeam area: skip past pieces, Slack line per re-plan, Waiting on me, @mentions, Portuguese holidays, illustrative Queue cases from one source.
Cleaner demos, a stricter proof check, models ready to switch
- ImprovedSample projects no longer show questions or persona examples about past years.
- ImprovedThe proof check now reads headings and table cells: an unproven result there becomes an Add real data field.
- ImprovedRe-running only the monitoring step asks each question in its own market and language, like a full analysis.
- ImprovedSuggested community replies and reply drafts never carry internal persona names.
- ImprovedPrepared and switched off: newer AI models for each check and generator can be switched on by setting, with nothing changing until then.
- FixedThe MCP citations tool no longer lists Perplexity, which is not measured.
- NewTeam area: fewer background refreshes, names in activity, clear errors for unreadable files, real document dates, and new go-to-market drafts.
A faster, more honest first day
- ImprovedFirst results arrive about a minute sooner: the AI engines get your questions at the same time, and the wait screen counts answers as they come in.
- ImprovedDrafts only state what you can prove: unproven figures and capabilities become Add real data fields, persona names stay out, and the call to action points to your own site.
- FixedA task you add during your first quick results keeps its suggestion when the full scan lands, and long Weekly Plan tasks no longer stop halfway.
- FixedCompetitors' own sites are never suggested for mentions, outreach or partnerships, and integrations such as a CRM are not listed as competitors.
- ImprovedOpportunities, Ask anything and the next step speak plainly: no P4 codes, Enter asks what you typed, and a failed draft is offered as Generate again.
- ImprovedThe tour waits for your first result, the setup bar shows your first analysis as in progress, and engine failures on our side say so.
- ImprovedComparison pages for Profound, Semrush and Peec were re-read on 7 October 2026 and each page shows its own date.
- NewTeam area: a first-week card for new people, sign-up names everywhere and an Unassigned filter on the Board.
Groundwork for web search in ChatGPT and Gemini
- ImprovedSite audits no longer stall for 10 seconds on pages that keep a connection open.
- ImprovedPublic FAQs say ChatGPT and Gemini sources count when they answer with a web search, where enabled.
- ImprovedPrepared and switched off: web-search answers are measured as their own engine and Gemini sources are stored as real sites, so turning it on never reads as an experiment result.
Correct filled proof for free, and clearer sample counts
- NewContent Engine drafts offer Edit filled proof, including values filled before this week; a correction adds a new version and does not use a monthly draft.
- ImprovedAI Visibility, Reports, Analytics and Compare Markets count in questions, engines and answers, and Discovery and Reputation are shown as two separate samples.
- ImprovedAverage position says it covers answers that rank options; an answer about you alone, or one naming other options only in passing, reads Not ranked. Scores are unchanged.
- ImprovedResearch tasks such as customer interviews offer Mark as done instead of Generate draft or Mark as published.
- ImprovedInsights shows whether this month's shared searches are available before you inspect a query.
- FixedConnect a CRM on Revenue Attribution opens your project's integrations; team accounts no longer see a trial countdown; Automations and Community use plain words.
- NewTeam area: notes can be corrected for 15 minutes and removed or restored, and the Home map has Fit to screen.
A plainer Simple mode and pages that match the product
- ImprovedSimple mode opens each main page with one plain sentence built from its own numbers, says questions instead of prompts, and keeps the Dashboard to two indicators.
- FixedPricing and agency pages now show your logo on client reports, the read-only client report link and the monthly client email as included on Agency, as they have been since September.
- NewTeam area: re-plan from a start date with preview and undo, dependencies with a team inbox, a LinkedIn profile review, and Readiness rows that become board cards.
- ImprovedPrepared and switched off: web search for the ChatGPT and Gemini checks, AI-content marking for the EU AI Act, and a check of each integration before launch.
Live Reddit monitoring stops paying twice
- ImprovedLive Reddit monitoring judges buyer intent only for threads it has not saved yet, so your monthly live-monitoring allowance lasts longer.
Simulations name the engines that did not answer
- ImprovedAI Simulation shows a short note naming any engine that did not answer, and says the others answered normally, so a partial run never reads as complete.
Honest simulation status, no wasted discovery calls
- FixedAI Simulation no longer reports an Anthropic outage when only Gemini or ChatGPT failed; that engine's card shows its own No answer, and saved runs follow the same rule.
- FixedCommunity discovery checks whether Reddit can answer before any paid step, and keeps your existing communities instead of paying for suggestions it cannot check.
Clearer statuses and honest case studies
- ImprovedContent Engine: every status says what it belongs to, the proof list folds away, a Review mode gives the draft room, filled proof values stay editable and the editor follows the task you opened.
- ImprovedCase studies never tell a customer's story as fact without proof: those sentences are marked for you to confirm before publishing.
- ImprovedAI Simulation labels each answer as simulated, with the engine it represents and the model that generated it; failed engines show as unavailable.
- ImprovedRoadmap shows the format you are drafting, Reports counts saved drafts, Billing explains the trial limit and Integrations search covers every card.
- FixedKnowledge Base shows Seeded only after a real import, Insights shows the search cost before you run it, and search marks demo and current projects.
- ImprovedClaude now searches the web from your buyer's country, as a real buyer there would; the first analysis after this starts a new trend line.
Team notices in Slack
- NewTeam area (internal team only): assigning or finishing a card, an @mention in a note, a published piece or an outreach win posts one line to the team's Slack channel. Prospects and note text never leave the team area.
First analysis counts once
- FixedA new project's first analysis now uses one analysis of your monthly allowance, not two.
- FixedCompetitors that share a first word (Mango and Mango Tree Spa) keep their own rows; a brand's product models still fold into the brand.
Clearer team access message
- FixedA teammate whose email is not verified yet is told to verify it, with a button to send a new link, instead of a bare no-access message.
Hide non-competitors and clearer run limits
- NewHide an AI-discovered name that is not a competitor, with Undo and a Hidden names list to restore it. Your own score and mention rate never change.
- NewNew project now offers "Start the analysis now", like sign-up, with a note that it counts towards your monthly analyses.
- ImprovedShare of voice counts your brand and the competitors you track, so the shares add up to 100%.
- ImprovedAt your monthly analyses limit, the Run button says so and opens a panel with the reset date and your options.
- ImprovedMicrosoft Teams alerts work with Teams Workflows webhooks.
- ImprovedA one-time task to set up Bing Webmaster Tools for Microsoft Copilot and Bing.
- FixedTrials unlock their full limits once the email is verified; disposable email addresses cannot sign up. Simulations: Pro 50, Agency 150.
Real competitor logos and a safer team area
- FixedThe Competitors page now shows each competitor's real logo and fills the Top Platform column with the AI engine that names them most, or says not measured.
- FixedTracked competitors with everyday names (for example a brand called Fresh) are no longer dropped from AI answers as generic words; they recover on the next analysis.
- FixedTeam access, unlimited analyses and multi-market now require a verified email address on the team domain.
Improve pages built with JavaScript, safer page reading
- ImprovedImprove an existing page now also reads pages that only show their text with JavaScript, by opening them once in a protected browser.
- FixedThe browser ClarAI uses to read sites now checks every request it makes, including pop-ups, background workers, embedded frames and live connections, and only connects to addresses it has checked.
- ImprovedFair-use monthly limits on live Reddit thread grading and featured images; at the limit, live threads keep arriving with simpler scoring.
Four new kinds of opportunity, and tighter security
- NewFour new kinds of opportunity, built only from answers we already measured: when one AI engine names you much less than another on the same questions, when AI keeps citing other people's statistics where you are absent, when a competitor starts being named on questions where you are not, and when comparison questions have no third-party list to cite. Each states its sample size and claims no cause.
- FixedSigning out now ends your session on our servers too, not only in your browser.
- FixedExtra protection against requests forged from other websites, and libraries with published security fixes are updated (sign-in tokens, uploads, XML reading).
Safer connections and an Improve-a-page draft
- FixedEvery address ClarAI fetches for you (your website, connected WordPress, Ghost, Jira and similar sites, webhooks, sitemaps and the free checkers) is now checked against private and internal networks on every step, including redirects. Sites on a private network or VPN can no longer be connected.
- NewImprove an existing page: paste the address of one of your own pages in the Content Engine and get an improved draft that keeps every fact the page states and marks missing proof for you to fill in. It counts as one draft.
- ImprovedWhen an analysis suggests no opportunities, the page now says whether nothing was found or the step did not finish, and what to do next.
- FixedCompetitor lists keep real names such as Boots, The Whisky Exchange, Essence and Famous Footwear, merge 'Whisky Co' into 'The English Whisky Co', and suggested competitors bring their real website for the logo.
- FixedDrafts no longer keep an invented byline written as a sentence, such as 'By Dr. Jane Smith, MD'.
- ImprovedBuyer personas, Knowledge Base steps and breadcrumbs open in a focused view in the Simple and Standard views, and AI Visibility shows how many measured questions did not name you.
- FixedThe prompts setting now says it controls how many prompts are written per stage, not how many each analysis measures.
Google AI answers read like every other engine
- ImprovedGoogle AI Overviews and Google AI Mode answers are now read the same way as ChatGPT, Gemini and Claude answers: whether they mean your brand, where they rank you, their tone and which competitors they really name, so publications or your own programmes no longer show up as rivals. If the reading fails, the answer still counts with the simpler local check.
A score that measures discovery, not just reputation
- ImprovedAt least half of the questions behind your AI Visibility Score now leave out your brand name, so the score shows whether AI brings you up on its own, not only what it says when asked about you. Your first analysis after this release is marked as a new starting point, never as a fall, on trends, the weekly email, client reports and alerts.
- ImprovedOpportunities now also come from the questions that do not name you: where AI left you out and which competitors it named instead, with the questions linked so the result can be measured.
- ImprovedWhen an analysis finds no opportunity, the page says why with the numbers we measured and suggests what to do next; the Weekly Plan no longer asks you to run an analysis you already ran.
- FixedThe AI-discovered list no longer shows answer headings, audiences, places or generic phrases ("Quick Verdict", "Working Moms", "Como Funciona", "Price Point") as brands, and real names such as H&M are no longer dropped.
- ImprovedWhile we read your website, the fields it fills are locked with a clear "Reading your site" line and a "Fill in myself" button; they unlock on their own when the read ends.
- FixedCompetitor logos try your market's own domain first (continente.pt before .com) and handle accents (Intermarché), and say when no logo exists instead of showing a wrong address.
- ImprovedIn the Simple and Standard views, buyer personas open in a focused view with a link back, without the advanced tabs.
- FixedWhile an analysis runs, missing figures read "measuring…" and no Run button is offered for a run already in flight.
- FixedA tab left open across an update reloads itself once instead of failing to open a page, and the error screen now matches the rest of ClarAI.
One honest reading on every engine
- ImprovedGoogle AI Overviews, AI Mode, the discovery check and AI Shopping apply the same identity rule as the other engines: an answer that cannot tell which product you mean is not a mention, and positions only when the answer ranks options.
- ImprovedThe tone of each answer that names you shows straight away from the stored reading, with no extra check to click.
- ImprovedYour Knowledge Base positioning, ideal customer and value proposition now shape the questions and personas we generate.
- ImprovedNo invented authors, reviewers or testimonial names in structured data, Content Engine quick actions or interactive widgets.
- ImprovedRecording model costs and sent emails no longer slows your requests, and those records are no longer lost on a slow connection.
- FixedEleven content types and eight scored readiness checks wherever we count them.
Clearer readings, one next step, proof you can fill in
- NewCompany facts in your Knowledge Base (founded year, headquarters, markets, languages, support, free trial, pricing summary), saved exactly as typed; drafts state them as your own records and brand accuracy (early access) compares answers with them.
- NewFill missing proof: every "add real data" field in a draft gets its own input, with progress; filling saves a new version without spending a monthly draft, and publishing stays blocked until all are filled.
- NewOpportunities with no measurement behind them are labelled "Hypothesis to validate", show what backs them, and ask you to check your closest existing page before you create one.
- ImprovedAnswers that cannot tell which product you mean no longer count as a mention ("identity not confirmed"); a position shows only when the answer ranks options. The first analysis after this release is marked as a new starting point on your trends.
- ImprovedWhen every measured question already names your brand, AI Visibility says the headline shows reputation, not discovery, and counts answers that name you, not recommendations; an optional tone check shows answers that speak against you.
- ImprovedNo score before a reading: "Not measured yet" or "First analysis running" instead of 0, and no button offers to start an analysis while one is running.
- ImprovedThe Dashboard and the Weekly Plan suggest the same next step with the task's own time estimate; tasks show what they depend on next to Generate draft.
- ImprovedNew projects keep the name and category you confirm; the suggested description comes from your site's first paragraph, never cut mid-word.
- ImprovedNo invented experts, titles or credentials in any draft type; outreach emails end with a field for your own name. Generated questions no longer assume case studies or figures you never gave us.
- ImprovedA plain message when your browser cannot reach ClarAI, and server errors show as errors instead of "Failed to fetch". One name for the score: AI Visibility Score.
- FixedPricing words such as "Plan" no longer appear as competitors; adding a suggestion to your plan during an analysis always tells you what happened; editing authors or company facts no longer regenerates your questions.
ClarAI 6.0: from a score to proof
Brings together the 127 changes released from 5.42 to 6.5: new features first, then improvements, then fixes.
New44 changes
- Add real authors to your Knowledge Base; article drafts use the one you choose. Without one, drafts carry no byline and the checklist suggests adding it, never blocking publishing.
- Brand accuracy (early access): what AI engines say about your pricing, plans and features next to your own records, with the exact quote, from answers already collected.
- Read-only Google Merchant Center and Business Profile connections feed your commerce and local checklists; Business Profile access needs Google's approval first.
- For the team: each ideal customer profile lists its tagged content and cards, and outreach touches record the template used.
- Blog and documentation drafts open as a preview in your brand's colours, fonts and logo, read from your website; only what we could really read is used, and suggested images are labelled spaces, never stock photos.
- For the team: weekly blocks on the operating calendar, ideal customer profiles on Board cards and content, a personal Today list and message templates to copy (nothing is ever sent).
- Snooze, mark as not useful with a reason, or restore opportunities; they keep their identity across analyses and are ranked by what was measured, with a reason for each.
- Outreach board on Earned Mentions: cited pages, their named author, a short pitch on request sent from your own mailbox, and reminders at day 7 and 14. Plus a review programme kit.
- Technical fixes now come only from your own Site Audit findings on pages linked to the prompts you track, and paid plans get a weekly re-check that marks a fix verified once the page is checked again.
- More crawler and page checks: Bingbot, Applebot, robots.txt rules per page, a crawler user-agent test for firewalls, noarchive, nocache, Apple nosnippet, canonical, dates and hidden text aimed at AI models.
- Readings pooled over 28 days with the number of answers and a likely range, no engine verdict before 7 answers, and whether AI cited your own pages.
- Upload your Bing AI Performance and Google generative AI report exports and see their numbers beside ours.
- List pages, review profiles, videos and community threads that AI cites become opportunities with a result of their own, and earned targets land on Earned Mentions.
- Sub-question coverage, commerce price and stock parity, a local checklist and redirect suggestions for pages AI visitors land on that return 404.
- A “Next step” card on AI Visibility, Trends, Answers, Competitors, Citations and Reports, built from your own data, with “Why am I seeing this?”, “Not now” and “Not useful”, shown at most three times.
- Numbers open their evidence: the mention rate, each engine and each run open the answers behind them, already filtered.
- A “Needs attention” line on every project page for real problems only (failed analysis, disconnected integration, no fresh reading for 8 days, re-check due), and a dot in the menu for what is new.
- Discovery beside your score: up to 8 questions that do not name you, sent with the question only, and how often ChatGPT brings you up on its own (for example 3 of 8 answers), with the likely range and the answers themselves. Starts with your next analysis; your headline score is unchanged.
- Reputation: how often you are mentioned when the question already names you, next to Discovery on the AI Visibility page.
- A data-quality label on every reading (e.g. “Exploratory: 4 prompts, 16 answers”), following the IAB's 2026 guidance.
- Your Monday email arrives at 9am in your time zone, or at the check-in time you choose in the weekly routine.
- Stay signed in while you use ClarAI (up to 30 days), and land back on the page a link pointed to after signing in.
- Agency features during the trial for sign-ups from the agency pages; new accounts start in the Standard view.
- Workspaces that hold your clients: create projects in the selected workspace, move them between workspaces, and list by workspace.
- Client viewers on Agency: read-only access that does not use a team seat.
- Your agency's name, logo and colour on the client PDF, plus a read-only client link you can switch off at any time.
- A monthly client report by email, with month-over-month changes against the normal range and an editable summary built from measured facts.
- One Monday email for all your projects, starting with the ones that need attention.
- Audit a prospect: a quick, clearly labelled first reading to share in your pitch, convertible into a client later (5 a month on Agency).
- Looker Studio via Google Sheets and a weekly CSV export with the number of answers behind each week.
- The same prompts, week after week: analyses re-measure your current set so changes are comparable; your own prompts are kept and removed ones stay removed. New prompts come with “Refresh prompts” or when your brand details change.
- A “first reading is ready” email, a note when your trial ends, and one fresh check before a trial ends if nothing ran in its last 3 days.
- Ranges beside your score and mention rate: “likely between X and Y, from N answers”.
- Real milestones: your first analysis, the first time an engine names you, your first finished task, the first re-check that differs from its baseline, and your first week of AI visits.
- An optional weekly routine (off until you turn it on): read what changed, finish one fix, re-check one experiment. Items tick themselves; one missed week keeps your streak.
- With Google Analytics connected: key events GA4 counted in visits from AI assistants (last click, a correlation), plus a weekly trend of AI visits.
- Your HubSpot pipeline on the recap card, next to your visibility and never presented as a result of ClarAI; plus a Monday recap in the notification bell.
- “Since your last analysis” at the top of the dashboard: your score move in points, “named in X of Y AI answers”, prompts gained or lost, competitors that overtook you, citations of your own site and the work you finished. It says “steady” when a change is within normal variation, and leaves out anything not measured.
- “This week”: up to four things worth doing, in order (due experiment re-checks, drafts waiting, your oldest plan task, a new high-impact opportunity), with “Do this next” on the card, its reason and roughly how long it takes.
- With Google Analytics connected, the card shows the last 7 days of visits from AI assistants, labelled as a floor.
- An email when an experiment is ready to re-check: baseline and latest reading side by side, no winner called, sent once, and it follows the weekly email's on/off switch.
- A weekly automatic check, on by default for paid plans and trials: your projects are re-measured early on Monday so the Monday email and the dashboard are fresh. Free accounts get it while you sign in at least every 14 days; turn it off or change the day in Project Settings. It counts toward your monthly allowance.
- “Proof you can use” in the Knowledge Base: add your real results, quotes (with permission) and statistics once, and the Content Engine uses only these, or public sources it cites.
- Drafts that need proof you have not added show a highlighted “Add real data” field instead of an invented number or quote, and cannot be published until it is filled.
Improved65 changes
- Your brand's style is read from your live homepage when there is no crawl yet; defaults are never presented as yours.
- Built and ready to switch on later: sub-question coverage matched by meaning, and weekly checks sent as a batch at half the model cost.
- The Content Engine editor stays inside the app: the menu and top bar stay in place, with a clear page header and one main action.
- A sharper read of your website: pages some sites sent compressed in a way we could not read are now always received in a readable form.
- “Tasks this week” counts only this week, impact shows as High, Medium or Low, and snoozed opportunities are no longer counted as open in Ask anything or the agency PDF.
- Cost tracking now covers every model call, including engine calls that time out.
- The whole public site shares one design: blog, changelog, manifesto, comparisons, landing pages, legal pages, the public brand page and the 404, with Log in on phones and Terms, Privacy and Legal in every footer.
- Personas, tracked questions and the Weekly Plan use a newer model: about a third cheaper per analysis and three times faster for the plan.
- Weekly Plans arrive in the language of your main market, like personas and questions.
- Content Engine drafts know which sources AI cites for the opportunity's questions and the latest result of its check.
- Restaurants have their own category, counted as a local business, separate from fast food.
- Accepting a measurable opportunity records its starting point straight away, and a “Do this first” row points to the next step.
- Alert emails only when you have not seen the notification, notifications grouped by category, and past-week tasks shown as carried over.
- Re-running only the opportunity step now builds every kind of evidence-based suggestion the full analysis does, with the same limits.
- Drafts from Weekly Plan tasks know which sources AI engines cite for that task's prompts, and video creator pitches read properly.
- Site Audit keeps the crawler access test after a reload, shows the engines each issue applies to and flags language versions that do not link back.
- Sub-question coverage also lists sub-questions an AI engine is likely to search for, refreshed weekly on paid plans and always labelled as generated, beside the real ones from Bing and Search Console.
- Content Engine rules and audit findings say which engines they apply to; the FAQ is optional and built only from real buyer questions.
- Structured data is no longer presented as a way to get cited by AI; FAQ markup is optional and for Bing and Copilot only (Google stopped showing FAQ rich results on 7 May 2026).
- The free AI Readiness grade no longer scores llms.txt (Google does not use it); grades can move by up to about 2 points.
- The Content Engine picks the right template, checks each draft again, and its quick actions produce a real meta description and valid JSON-LD.
- WordPress posts arrive as clean HTML; links in drafts come only from stored sources or your own pages.
- robots.txt is read per page path, the redirect check works, and JavaScript-only content is checked as AI crawlers see it.
- Weekly Plan drafts use only facts from your Knowledge Base, your site or a cited source; anything missing becomes an “Add real data” marker, and cut-off or empty drafts show an error.
- Every opportunity names the questions it should move; choose them in the Proof Loop. No readings are promised until it has some.
- Page tasks open the Content Engine with the evidence behind them; each opportunity has one clear next move; tasks have an owner, a due date and a “Done when”.
- Mark a task published with its live address; ClarAI checks the page is live before counting readings.
- Your actions appear on the trend line, placed in time, never as the cause.
- Done, Snooze and Mute on every notification; teammates get what concerns them; experiment emails only if unread; a switch to turn celebrations off.
- Agency: approve the monthly client report before it is sent (on by default); a change in how answers are measured shows as a new starting point, never a fall.
- The AI engines now receive the buyer's question only, not your brand name, and we find you in the answer ourselves. In our test, naming the brand lifted ChatGPT mentions from 12 to 16 of 20 answers.
- A small reader resolves ambiguous names (Patagonia the region versus the brand) and reads position, sentiment and competitors from each answer, instead of trusting the engine's own report.
- Trends mark the change of method, so a lower score after this release is not read as lost visibility.
- Experiments say what really happened: Measuring, Early reading, Changed beyond normal variation, No measurable change, Not enough data or Ended by you, with the answers behind it, the real re-check date and one email when decided.
- Content Engine drafts are saved with versions, never lost when you leave, linked to their task, and offer to start measuring after you publish.
- Finishing a task confirms it and offers the next step; “Act on this” tasks link to their opportunity and are not duplicated.
- Trends mark a break when two analyses did not measure the same thing (different questions, an engine that did not answer, a smaller sample), and “What changed” says why.
- “What changed” says when a move could be chance, and labels question moves based on a single answer.
- The normal range is measured on the analyses before your latest one and never narrower than its sampling margin.
- AI content drafts have a monthly fair-use limit (5 Free, 100 Pro, 300 Agency); rewriting a section does not count, and reaching it shows your plan options.
- A failed AI engine no longer counts as “not mentioned”: it is left out of your score and the dashboard says which engine did not answer.
- Your headline number now matches your latest analysis and its trend line.
- Sources list only the pages an answer actually cites; ChatGPT and Gemini no longer show sources they cannot have.
- Opportunities no longer promise invented results and are built from what the analysis measured.
- Community replies always disclose your affiliation, and consistent labels replace random ones.
- A clear view-only message for read-only members; prospect audits run on background workers; deleting your account deletes its product events.
- The privacy policy covers in-app usage events (kept 13 months, off with GPC/DNT); email addresses are masked in sending logs.
- “Viewer” project members are now read-only, as the Members screen always said.
- One “Do this next” on the dashboard, a quiet setup row, at most five setup steps, and a first-analysis screen that stands on its own.
- The dashboard works on phones and tablets.
- While online checkout is being set up, upgrade buttons say “Talk to us” and email our team instead of leading to an error.
- Your first day counts one analysis, not two: the quick first look and the full analysis are one run.
- Before your first analysis, ClarAI lists the AI engines your account is actually monitored on; every screen gives the same waiting time (about 6 minutes).
- Proof Loop experiments compare against a real baseline recorded from your stored answers.
- The Agency dashboard only calls a change a real move when it is outside that client's normal range, and flags clients without a fresh reading.
- The workspace switcher is on every project page for accounts with several workspaces.
- AI-discovered competitors must be named in at least two answers, so phrases the AI used (like “Important Nuances”) no longer show up as brands. Tracked competitors are always listed.
- Pro, Agency and trial accounts can leave Simple mode to reach Integrations, Reports and the Content Engine; nothing switches by itself.
- The “analysis complete” pop-up only appears for your first analysis, a real milestone, or a move up outside normal variation.
- The Monday email is built from the same summary as the dashboard, and “How we measure” names the engines your last analysis actually checked.
- ClarAI now reads Claude's full answer: many answers were cut off before the end, so part of what Claude said (and whether it named you) could be missed.
- The weekly email tells the truth: only tasks you actually finished under “Completed this week”, open opportunities under their own heading, and a score change that says “steady” when it is within normal variation.
- Notifications open the exact item: the opportunity, the Visibility trends, or the experiment to re-check.
- The dashboard no longer overstates small moves, and never shows a placeholder 0 while numbers load.
- Community reply drafts are written by you, from the brand, with a short disclosure, and never invent experiences.
Fixed18 changes
- Article drafts no longer invent an author, team or reviewer.
- Suggested image descriptions no longer contain internal numbers, and the featured-image option states the real price and that AI images are unavailable for now.
- The customers page says plainly that no stories are published yet, and the agencies page no longer says every plan can switch views.
- Editors can save content status in the Content Engine again.
- A list page not re-read in a run keeps its last result instead of being reported as not cited, and homepage schema advice no longer suggests the retired sitelinks search box.
- “Latest AI answers” loads again on the AI Visibility page.
- No more suggestions to earn mentions on Wikipedia, YouTube or a competitor's site; the Citations filter works for Google's AI surfaces.
- “Write with AI” works again on documentation pages and other long drafts: they now have room to finish (one to three minutes), instead of being refused as cut off.
- The brief and the “Write with AI” area sit together on one scrollable background, so the brief no longer looks cut off.
- Outreach is sent from your own mailbox; no more suggestions to edit your own Wikipedia page; generic words no longer appear as competitors.
- Honest copy on AI Visibility and Answers, and a new analysis no longer removes an opportunity you drafted content for.
- Brand mentions are matched as whole words (“Hims” no longer matches inside “himself”), and the What changed card fits on phones.
- Run monitoring, re-runs and the API respect your plan limits, the same analysis can no longer run twice, and sample projects are for exploring only.
- “Refresh prompts” works again.
- Fixed links: the pop-up's “Open opportunity”, approval notifications, and Weekly Plan ↔ Opportunities links open the right item.
- The “Turn on weekly checks” step completes when the check is on; no more “step null”; clean site descriptions; project invites respect seat limits; trial emails open Billing; sample projects send no Monday email.
- Fewer false competitors (compliance terms like “Type II” or “Trust Center”), “Engines live” counts engines that actually answered, and HubSpot figures say “deals” since they include closed ones.
- Content Engine emails send again; an analysis no longer erases saved competitor links or custom Query Fusion axes; a failed Claude check is no longer counted as an answer; community scans and Sentinel simulations respect your monthly allowance; the platform tour starts again.
Tune your prompts, and a calmer running-analysis bar
- ImprovedCurate your prompts: remove one you do not want (it stops being measured but its history is kept) and add your own, right on the AI Visibility page.
- ImprovedSet how many prompts are generated per bucket (4 to 12) on the prompts page itself, with a live total (it sizes the prompt library; each analysis still measures a sample), instead of digging into System Settings; and the Refresh prompts button now sits directly above the list.
- ImprovedWhile an analysis is running, cells that are not measured yet read “measuring…” rather than “not measured”, so an in-progress result is never misread.
- ImprovedThe running-analysis bar can now collapse to a slim strip so the page content shows straight away; it remembers your choice per project and still shows live progress.
- ImprovedA “New project” shortcut in the project switcher opens the create form in one click.
- ImprovedThe “AI-discovered” competitors section now explains itself in plain words (brands the AI engines named that you are not tracking) with an info tip, keeping the note that only real names that appeared are listed, never invented.
Teammates can name their shared workspace
- NewThe invited teammate can rename the shared workspace on first entry: it is created with the project's name, and the first time your teammate opens it they are quietly offered a different name. You never see the prompt (you named it), and it shows at most once.
- FixedThe "Try this first" label is no longer clipped on the Projects cards.
A brighter Projects screen
- ImprovedThe Projects screen has more life: a subtle brand mark, an honest at-a-glance line (projects tracked, demos, and any needing attention), and cards that respond as you move - a brand accent edge, a gentle lift, a clearer status dot (pulsing while an analysis runs), and a rounded arrow that fills in. A project with no analysis yet stays plainly marked - nothing invented.
- ImprovedConsistent wording finished: breadcrumbs and the command palette match the renamed navigation (Answers, Sources AI cites, Missed answers, Answer changes), and the score reads AI Visibility Score in the Reports widgets too.
Shared workspaces on invite, encrypted webhooks, plainer words
- NewInviting a teammate to a project can now create a shared workspace for it: invite the first collaborator to a personal project and you are offered a shared workspace named after the project (editable), which moves the project into it and adds you both. A quiet "just add them to this project" link keeps the simpler invite.
- ImprovedYour connected Slack and Microsoft Teams webhook addresses are now encrypted at rest and no longer returned by the integrations status view.
- ImprovedClarity pass: one name for the score everywhere (AI Visibility Score), plainer setup wording, clearer navigation labels (Sources AI cites, Missed answers, Answer changes), the AI replies called "answers" consistently, and jargon like ICP rewritten in plain words.
- FixedA deep scan that times out no longer looks like your whole analysis failed when your quick results are intact; PNG export shows a friendly message instead of a technical command; and each Weekly Plan task records the kind of move it represents.
Live-QA fix batch: honest tables and eight interface bugs
- FixedFailed engine calls no longer appear as an engine: a "gemini_error" row could show in the Competitors table and trends. Error rows are excluded from the numbers (a failed call is no signal, not a zero) and historical data is cleaned.
- FixedIdentity jargon and answer fragments stopped appearing as competitors (SAML SSO, Single Sign, Honest Answer and friends), near-identical names collapse into one column, and demo projects were scrubbed of the same noise.
- FixedWeekly Plan tasks no longer invent a website address from your project's name - tasks always use your real site, so executing a crawl task works.
- FixedInterface polish: the first-analysis card stays gone once you have run one, misleading competitor evidence links removed, pages open at the top on navigation, the tab strip pins correctly, the score chart renders cleanly with hover values, and two dead controls were removed.
A Monday digest by default, and evidence-grounded opportunities
- NewA weekly email digest, on by default: every Monday, a summary of what changed on each project. Turn it off any time with the new toggle in Project Settings or the one-click unsubscribe in the email itself. Projects with no data yet are skipped rather than sent an empty email.
- ImprovedOpportunity suggestions are grounded in your measured run, not general strategy: the generator now sees which prompts were measured, how each engine answered, which sites won the answers you lost, and your positioning notes, within a strict cost cap - and routes recommendations (earn a mention, drive reviews, create content) from the actual citation pattern.
- New"Your page lost a citation it used to hold": when a consistently cited page of yours stops being cited, a refresh opportunity appears naming the page, prompt and engine with dates. Solid history only - never a first-run guess.
- New"Get onto the sources AI already trusts": when third-party sites repeatedly win AI answers where your brand is absent, an opportunity names them and the prompts they win. Every count states its sample honestly.
- NewA review-platform presence nudge when a review site relevant to your sector keeps appearing in AI answers for your prompts, claiming only what the answers show.
A friendlier journey from first grade to weekly habit
- ImprovedThe dashboard finally explains itself: every headline number carries a plain-words hover explanation, the score shows its reliability band, and a full "What changed since your last analysis" panel appears once you have two runs - a change within the noise band honestly says "no real change".
- ImprovedSimple mode got genuinely simpler: AI Visibility and Competitors tuck their analyst-depth panels behind one quiet "Show the full analyst view" control, and jargon was rewritten in plain words on the surfaces beginners see. Standard and Advanced are unchanged.
- ImprovedInsights now lead somewhere: competitor gaps link to the matching opportunities, citation gaps link to the work that closes them, and notification bell entries open the exact panel they refer to.
- FixedA failed bonus deep scan no longer looks like your analysis failed: when the quick results succeeded, you see a calm note that they are unaffected instead of a red failure.
- FixedGeneric phrases like "GDPR Compliance" stopped appearing as competitors, with real brands that have generic-sounding names explicitly protected.
- ImprovedThe domain you graded follows you in: sign-up from the grader is attributed correctly and onboarding pre-fills the site you just graded (stored privately in your browser, never in a link). Onboarding fields are usable immediately while your site is read in the background.
- ImprovedThe trial is transparent about its end: the banner and reminder email state the exact date, what the free plan keeps, and that nothing is charged. The free plan's weekly refresh limit reads as a friendly note with your actual refresh date.
Your work survives, and the Proof Loop measures for real
- FixedRunning a new analysis no longer erases your work. Previously a full analysis deleted every opportunity and task, including accepted ones, completed tasks with their drafts, and active experiments. An analysis now refreshes the suggestions only; everything you have touched survives and dismissed items stay dismissed.
- NewThe Proof Loop now measures for real: every analysis records a per-engine reading for your active experiments (with date and honest sample size), the panel shows it next to your starting point with change markers, and you get a reminder when it is time to re-check. Failed engine calls are excluded rather than counted as absent.
- NewOne status per piece of work: the stage on Opportunities reflects the task's real state (In your plan, In progress, Done), starting an experiment marks the task in progress, and an experiment chip shows where the Proof Loop stands.
- ImprovedPlainer words across the flow: Open instead of Review, To do instead of pending, Proof Loop buttons that name their destination (Mark as Published, Start measuring), stages renamed to Work in progress and Result ready, and Find communities to share in.
- FixedGenerated drafts render as formatted content with headings, tables and lists instead of raw text markup, Regenerate asks before replacing a draft, and the Weekly Plan reads comfortably at laptop widths.
- ImprovedDemo projects now teach the real flow: a few tasks arrive as suggestions awaiting your approval, so the accept step is visible from the first minute.
Defence in depth: quieter locks on every account door
- ImprovedTwo-factor setup endpoints are now rate limited, so a stolen session token cannot hammer the enrol or disable flows.
- ImprovedPassword-reset tokens are now stored hashed. The link in your email works exactly as before; what sits in the database can no longer be used directly if it ever leaked.
- ImprovedRequests with unexpected extra fields are now rejected on the core account endpoints (sign-up, password flows, two-factor), closing off a class of field-tampering attempts.
- ImprovedProfile picture URLs are validated (web addresses only) and transcript uploads are checked for their real file type, not just the name, so a disguised binary is refused with a clear message.
- FixedThe "Verify you are human" widget now appears as soon as the free grader loads. A timing issue meant it only showed after a first failed attempt, forcing a double submit.
- FixedThe MCP settings tab now shows the correct server address instead of a placeholder domain from before the move to myclarai.com.
A much faster first load
- ImprovedThe first download is about 80% smaller: the code fetched before anything renders went from 554 kB to 113 kB compressed. Every page now loads only when you open it, and the underlying libraries sit in separate files your browser caches across releases.
- ImprovedThe release history no longer ships with the app: Help & Docs and the public changelog show recent releases instantly and fetch the older history only when you press "Show older releases".
- ImprovedFonts start loading earlier, so text settles into its final look sooner on a first visit.
- ImprovedThe AI Visibility page computes its numbers in the database instead of loading every stored answer, verified identical to before. Pages get faster as your history grows rather than slower. The analysis progress check also now reads just the two fields it needs.
- NewThe forgot-password page now carries the optional bot-protection widget (Cloudflare Turnstile), completing the set alongside sign-up and the free grader. It stays invisible until the operator switches Turnstile on.
- Behind the scenesNightly maintenance jobs staggered across 04:00, 04:15 and 04:30 UTC, and a stale build-configuration artefact that had silently overridden the real build settings since May was removed.
A responsive Weekly Plan flow: act and see it happen
- FixedStarting a Proof Loop experiment now updates the panel instantly. It used to keep showing a stale snapshot, so the click appeared to do nothing until you reopened it. The panel now advances the stage strip, records the per-engine starting point and switches to "Advance Stage" the moment you click.
- FixedDemo projects no longer show a wall of overdue tasks. Sample data is re-dated relative to today when it is seeded, so a new account sees the current week with due dates inside it.
- ImprovedA clear thread from opportunity to result: a dismissable "How it works" strip (accept, generate the draft, review and publish, measure) now sits on both Opportunities and the Weekly Plan, highlighting where you are.
- ImprovedAccepting an opportunity tells you what happened and what is next: the review panel stays open with a persistent "Added to your Weekly Plan" note and a direct link, and the opportunity's stage reads "In your plan" with a link.
- ImprovedGenerating a draft is easier to follow: the Weekly Plan scrolls to and briefly highlights the task that just updated, and the pending state says what is happening.
- Behind the scenesThe API now lives at api.myclarai.com (same site as the app), and signing in also sets a secure HttpOnly session cookie alongside the existing token. First phase of moving sessions out of localStorage; behaviour is unchanged for existing sessions.
Security hardening: stronger protection for your data and credentials
- ImprovedYour connected credentials are now encrypted at rest: WordPress application passwords and Google Search Console sign-in tokens were previously stored in a reversible form and are now protected with strong encryption.
- ImprovedThe generated-HTML preview and the interactive-widget preview no longer run generated code with access to your logged-in session. Previews are sandboxed to an isolated space and live execution is switched off.
- ImprovedThe tools that fetch a web page for structure and site-audit checks now refuse to reach private or internal addresses, so a supplied URL cannot be used to probe places it should not.
- ImprovedAdded more protective layers, switched on carefully: new browser security headers (HSTS and a Permissions-Policy) plus a Content-Security-Policy running in report-only mode for now (it observes and reports without blocking anything yet), and optional bot protection (Cloudflare Turnstile) on sign-up, password reset and the grader that stays off until the operator adds keys.
- ImprovedAdded database indexes on frequently used columns so busy pages and lists load faster.
- FixedHardened how your saved prompt answers are protected so they can only be reached from your own account, with an automatic detector that keeps that protection from slipping again.
- FixedFixed an analysis that could appear stuck forever showing 'Full deep scan continues'. The quick-to-full handover is now covered by stuck-run recovery, and a periodic check heals any run that gets caught this way.
- FixedCompetitor lists no longer show generic phrases like 'Challenge Validation' as if they were brands, and the Haiku model cost is now accounted at the correct rate.
- Behind the scenesMoved our sign-in token handling off an older library that carried a known vulnerability onto a maintained one, with no change to how sign-in works, and added internal security, performance and cost audit notes.
Design polish, a working plan flow, and cleaner generated pages
- ImprovedEvery redesigned screen was checked against the design mockups and brought closer to them: text-and-underline tabs (no icons), page actions back in the header, clean divided lists, consistent colour families, and full brand panels on the sign-in and create-account pages.
- NewAccepting an opportunity now puts a real, runnable task in your Weekly Plan instead of only changing a label.
- ImprovedThe Weekly Plan's real action is now obvious: the prominent 'Generate draft' has ClarAI produce the deliverable, while the old 'Start' becomes a quiet 'Mark in progress'.
- FixedGenerated brand pages are readable and complete: no more magenta body text, no more broken stock images (on-brand placeholders instead), and a whole page with nav, hero, sections and footer rather than a thin draft.
The redesign reaches every screen
- ImprovedAll 37 surfaces now share the new design language, from login to admin: Content Engine, Distribution, Community Inbox, Realtime, Approvals, Foundations, Personas, Agents, Citations and Sources, Reports, Reach Gap, Revenue Attribution, Engine Volatility, AI Simulation, Compare Markets, Channels, Help, Workspaces and the auth pages. Same data, same buttons, same honest empty states.
- NewTrends plots real mention-rate and average-position history from your stored runs; a single measurement shows 'A baseline, not a trend.' instead of a fabricated line.
- NewResponses becomes a reader: pick a response on the left, read it highlighted on the right, with a plain note that missing citations are not the same as a missing mention.
- NewReports renders your widgets inside a branded report document frame with Export PDF as the page action.
- FixedA fabricated community coverage bar on Distribution was removed, and every redesigned screen was probed so long content can never stretch a page sideways.
One sidebar everywhere, and a guided setup home
- ImprovedOne sidebar for the whole app: the dashboard now renders inside the same shell as every other page, so the left column no longer visibly changes when you move between sections. Its Discover, Plan, Execute, Measure and System groups are collapsible and remember your choice per project.
- NewThe dashboard's Guided setup tab opens a proper guided page: your seven setup steps with honest progress, real measurement numbers beside them, and links straight into each step.
- NewThe projects list gains search and honest filters: All, Demo projects, and Needs attention for projects whose last analysis failed.
- ImprovedGetting started shows all seven steps in one continuous card with stable numbers and clear Complete and Next markers; onboarding steps now explain themselves beside the form, including that we read public pages only.
The new design reaches onboarding and projects
- NewOnboarding redesigned: five clear steps in one card, and the final step starts your first analysis for you by default (uncheck to start later). Creating a project lands on the dashboard with the analysis already running.
- NewThe projects list redesigned: cards with your brand logos, an honest analysis status on each, and the Try this first ribbon kept on the samples.
- FixedDashboard shell polish from live feedback: no stray band at the top, the sidebar stays fixed while you scroll, its labels are clearly readable again, and its green is lighter and softer.
- FixedOnboarding no longer discards your edits when it loads details from your website, and an address problem takes you back to that step instead of failing at the end.
A new home: the redesigned dashboard
- NewOpening a project lands on the redesigned dashboard: deep green sidebar with your setup progress, four headline cards with plain sentences under every number, a grouped brand comparison with logos and real trend lines, a next-step panel driven by your actual setup, and shortcuts into answers, opportunities and alerts. Illustrated with the brand's clay artwork.
- NewThe dashboard search field opens the command palette, and the bell lists your latest notifications.
- ImprovedNothing was deleted: the previous dashboard lives on at /classic with every panel intact while the new home grows into them.
- FixedAccepting a suggested opportunity from the dashboard confirms inline with a 10 second Undo.
A new logomark, real competitor trends, a tidier Get Started
- NewThe you-vs-competitors table now draws each competitor's trend from your stored run history (mention rate per run). Fewer than two measured runs shows 'no trend yet' rather than a flat line.
- FixedPer-competitor Sentiment and Average position columns are gone: they were never computed and always showed a dash. The table now shows only what is measured, with sample sizes.
- ImprovedGet Started shows the next step first and folds finished steps into a single 'N steps done' row you can expand.
- FixedYour first analysis no longer emails an opportunities digest; in-app notifications still appear and later runs send one digest each.
- ImprovedOne market per project while multi-market analysis rolls out account by account; the picker says so plainly. Projects already tracking several markets keep everything, including the market selector and Compare Markets in every viewing mode. On multi-market projects, each prompt is now asked in its own market's context and Google results use the matching country.
- ImprovedThe Search Console block points at Google's own generative AI report and says plainly that Google offers it in its UI only. Pricing copy reads 'up to 5 AI engines' while engine availability is confirmed on production.
- NewThe new C-aperture logomark replaces the old waveform in the browser tab, the sidebar and email headers.