Changelog

Fixed: a product named by generic words and a number, such as Bluetti Premium 30, was counted as named by answers about other products that happened to contain those words (Pampers Premium Care … 30 pieces), and their prices were flagged. A model's words must now appear together and in order, quantities such as "30 pieces" aren't read as models, and generic names need their brand nearby. Invalid GTINs now say what's wrong: a model code in the GTIN field, a UPC missing its leading zero, a wrong length or a wrong check digit.

Fixed: a model written with a hyphen, such as Silk-épil 5-251, was matched as its whole line (Silk-épil 5), so answers about other models counted as naming it and their prices were flagged as wrong. Each model is now matched on its own, and the "AI quotes another price" flag only compares prices quoted for that exact model. A title without a model (Pampers Premium Care, or a tyre without its size) now counts as its product line, shown as "its line, … named", without comparing prices. And when a page has the product's GTIN only as its SKU or inside its product ID, "No GTIN" now says so.

Fixed: importing a product feed bigger than what Asheard reads (60 MB, or more than arrives in time from a slow server) failed with "The import stopped before the end". It now keeps every product up to where the feed stopped. And when a feed has more than 5,000 products, the ones that aren't out of stock are imported first, so the Products page checks what you actually sell.

ChatGPT fills its product carousels from Google Shopping: in a 2026 study, 83% of carousel products were in Google Shopping's top 40 organic results, and 60% in the top 10. The Products page has a new Google Shopping card: add up to 25 searches (your tracked keywords, the searches the AI engines ran for your product prompts, or your own), check them, and see where your store's listings rank among the top 40, who the top sellers are, and which of your products each listing is. Check now is charged per search; the searches are checked again every Monday when the project checks anything on a schedule; a check that fails is refunded.

Fixed: depending on the order of a table's columns, Looker Studio could show values under the wrong columns, most visibly the desktop and mobile start and latest positions. Latest positions now always come from the newest check in the date range, also in workspaces with more rows than one data source returns (it now keeps the newest rows). New Keyword rankings fields: country, language, tags, ranked, best position, period start position and period change (from the first check in the date range; also per device), own featured snippet, featured snippet and local pack in the results, sitelinks, CPC, competition, and Search Console clicks, impressions, CTR and position (last 28 days, when Search Console is connected). New AI visibility fields: prompt tags, pages read, own domain read and searches run. And a new dataset, AI citations: one row per source an AI answer cites. Press Refresh fields on your data sources to see them. Also fixed: keywords that share a search with another project's keyword now record AI Overviews too (they read as none).

Each project has a new Products page (under AI visibility). Import your catalog from your Google Merchant Center feed (its address or the file), your Shopify or WooCommerce store's address, or a list of product pages, or take the pages the AI answers you track already use. Asheard then reads each product's page the way AI crawlers do, without running JavaScript, and flags what to fix: pages that don't open, product structured data that's missing, broken or only added by scripts, no price, prices and stock that differ between the feed and the page, missing or wrong GTINs, no SKU or MPN, variants described as one product, descriptions under 50 words, weak titles and photos without alt text. Each product also shows how the AI answers of the last 30 days use it (cited, read, named) and the prices they quote next to its name, flagged when they're wrong. Up to 5,000 products, 250 pages per check, the ones AI uses first; the CSV has it all. Free.

Each project can now connect Google Analytics 4, read-only (Google Analytics in the project menu): sign in with Google, choose the property, and Asheard reads it once a day, free. The new page shows the sessions AI assistants such as ChatGPT, Gemini, Copilot, Perplexity and Claude sent you in the latest 28 days, against the 28 before: their share of all sessions, how engaged they were and the key events they brought, per assistant next to how often the engine's answers cite you, and the landing pages they sent visitors to, next to how often the answers cite each page. Sessions count by their source, so they include the ones Google Analytics files under Referral or Unassigned rather than its AI Assistant channel. All channels are there too; Google's AI Overviews and AI Mode are part of Organic Search, as Google Analytics can't separate them. The AI crawlers page adds each page's visits from AI assistants. Disconnecting deletes everything it read.

Insights has a new Cited when read card: of the answers in which an engine read your site, the share that cited it, per engine, next to the brands tracked in other Asheard workspaces (the middle half of them). A rate below most brands means the engine finds your pages and then quotes other sources, and Gaps and Fixes lists those engines. Only these percentages are shared between workspaces, never a brand, prompt or site. Reddit citations that are Reddit's own machine translations of threads written in another language are now marked on the answer page, Earned media and the citation gaps, with a link to the original thread, where the discussion is. Also fixed: Earned media showed its messages twice.

Each project has a new AI crawlers page (under AI visibility). Upload your site's access logs (Apache or nginx, plain or .gz) and see which AI crawlers read it: the search crawlers behind ChatGPT, Claude, Perplexity and Mistral's answers, the ones that open a page while answering someone, and the training crawlers, per crawler and per day. Requests that only pretend to be an AI crawler are checked against the addresses the vendors publish and left out. Each page shows the whole path: crawled, read and cited in the AI answers we track, and clicked on Google when Search Console is connected. What to fix: crawlers robots.txt blocks, crawlers refused by a firewall or CDN, missing pages, server errors, and pages crawled a lot but never cited. Only the crawler requests are kept: visitors' addresses are never stored and the files are deleted once read. Free.

Each project can now connect Google Search Console, read-only (Search Console in the project menu): sign in with Google, allow access, and Asheard reads your search data once a day, free. Clicks and CTR now show next to your tracked positions. The new Search Console page has your clicks, impressions, CTR and position against the 28 days before, by device and country; the CTR of your keywords on days their search showed an AI Overview against days it didn't, from your own rank checks; the searches that bring clicks but aren't tracked, to track in one click in the country they come from; the longer, question-shaped searches, to track as prompts; and your top pages next to how often the AI answers we track cite them. Disconnecting deletes everything it read.

Every AI answer that names you or a competitor is now also read for what it says each brand is known for: up to five attributes per brand (price, quality, customer service, speed and so on), each said as a strength, a weakness or neutrally, and whether the answer raises a concern. Insights gets a Perception by attribute card that compares you with your competitors attribute by attribute, and the answer page and buyer journey turns show what each answer says. Gaps and Fixes adds “Known for it, missing when asked”: what the answers praise you for, where the prompts asking who's best for it leave you out, with the questions buyers would ask, to track in one click. Both go into the action plan, along with answers that name you but raise a concern. It comes from the same pass that already reads each answer, so it costs nothing extra, and it fills in from today's checks on. Also fixed: after Mark done or Reopen, Gaps and Fixes showed the message twice, and share links scrolled sideways on phones.

Gaps and Fixes now starts with an action plan: what it finds, and your latest checks (fact check, broken links, buyer journey, page audits, recall and top-of-mind checks), as one list of actions per project, grouped into your own site, earned media and technical fixes. Each action says what to do and shows the evidence behind it (the answers, engines, prompts and sources), with a priority: impact × ease. Briefs only: the content is yours to write. Mark an action done and the plan keeps how its tracked prompts did in the 28 days before; 14 and 28 days later it shows how often they name you and cite your site since, from your scheduled checks, at no cost. New high-priority actions come in the alert digest. The plan replaces the negative-answers and recommended-actions cards, and the Citation gaps list no longer shows the same page twice for tracking parameters, or your competitors' own pages.

Each project has a new Fact check page (under AI visibility). Save your brand facts there, or let Brand fact audit save them when you pick the project: prices, what you sell, where you deliver, founding year and so on. The check then reads the latest AI answer per prompt and engine that names you (last 30 days, up to 40) and marks every statement about you that touches a fact as supporting it, contradicting it or unclear, with the answer's own words and the source it cites, so wrong prices and outdated details stand out. Answers already checked against the same facts aren't paid for again. It costs a few credits per batch of 8 answers; a batch that returns nothing usable is refunded. Managers run it, and Gaps and Fixes lists the facts the answers contradict.

Each project has a new Earned media page (under AI visibility): every third-party site the AI answers of the last 30 days cited for your prompts, grouped into best-of lists, review sites, forums, directories, news, video, reference, social and other sites. Each site shows how many answers cited it and on which engines, which competitors those answers name, whether you are named, and an opportunity score (answers citing it × competitors named there, less one when you are there too). The few sites cited again and again where more of your competitors are named than you come first, and Gaps and Fixes lists the top ones. Managers can have the best-of lists read, free: the page finds your place and your competitors' in each list (“#4 of 10”), or whether you're on the page at all. Your own site and your competitors' own sites are left out.

AI engines sometimes link to pages of your site that don't exist. Each project has a new Broken links page (under AI visibility): it checks every link to your site that the AI answers of the last 30 days gave, written in the answer, cited as a source or among the pages the engine read, and lists the ones that are not found, send visitors to the homepage, keep redirecting or point to a subdomain that doesn't exist. For each broken link it suggests the closest real page from your sitemap, and you download the 301 redirects for nginx, Apache (.htaccess) or the WordPress Redirection plugin. Free; managers run the check, and Gaps and Fixes shows what it found.

A new free AI Lab tool (Tools → Page audit). Give it a page and the prompt it should answer, and it checks in one go whether AI engines can reach the page, quote it and use it as the answer: snippet controls (noindex, nosnippet, max-snippet and data-nosnippet, plus Bing's noarchive and nocache), whether the first 160 characters answer the prompt, the page's length against how much of a page Google's AI uses, whether the format fits the prompt, robots.txt for this page grouped into search, user-fetch and training crawlers, readability without JavaScript, structured data and citability, with a score out of 100. llms.txt is shown for information. The Snippet Lens runs on the audited page for its usual credits and is refunded when it can't read the page. Gaps and Fixes and the Google & Bing page link to it. The Snippet Lens and Selection Duel also read pages better: words at the edge of a heading and the next paragraph no longer run together.

Every tool that reads a page you give it (AI readiness, AI bot access, "Can AI read this page?", Selection Duel, Snippet Lens and the website step when you create a project) and the alert webhook now go through one stricter check: the address must be a public website on a normal web port, every address the name points to is checked (IPv4 and IPv6), and each redirect is checked before it is followed. Private and internal addresses are refused. A webhook address that isn't a public HTTPS endpoint is refused when you save it. Normal websites behave exactly as before.

A new AI Lab tool (Tools → Buyer journey). Describe a buyer, and each engine answers five questions in one conversation: the problem, the options, a comparison, the reviews and the final choice. You see whether your brand is named at the start, survives the comparison and is the one pick at the end, where it drops out, and which competitors the engines name instead, next to how often single checks of your tracked prompts name you. ChatGPT, Gemini, Claude, Perplexity and Grok search the web as they answer; DeepSeek and Mistral answer from memory. Each turn costs the same as a normal check on that engine, and turns that get no answer are refunded. Question templates in English, Slovenian, German and Croatian, all editable.

Each project has a new Google & Bing page (under AI visibility) for the reports Google and Bing now publish on AI search: Search Console's generative AI performance report (impressions of your pages in AI Overviews and AI Mode) and Bing Webmaster Tools' AI Performance (citations of your site in Bing and Copilot answers, and the searches behind them). Upload the exports as they are to see the daily trend, the pages Google shows and how much of the total they cover, countries and devices, and Bing's searches with your share of their citations. Pages are matched with what the engines we track cite, Bing's searches can be tracked as prompts in one click, and Gaps and Fixes lists the Bing searches where other sites get most citations. Importing the same days again replaces them.

Each project has a new Gaps and Fixes page (Gaps in the sidebar) with everything that says what to fix, most urgent first: negative answers, recommended actions, where your brand drops out of the answers, coin-flip prompts, citation gaps and the searches where your site never came up. These cards moved off Insights, which keeps the measurements. Share links get a Gaps and Fixes tab.

Every prompt now has an intent: awareness, consideration, decision, comparison or branded. It is set automatically when a prompt is added (existing prompts included), and you can change it on the prompt row. Headline numbers (visibility, its trend and per-engine split, win rate and share of voice) now leave out branded prompts, which name you, and comparison prompts (X vs Y, alternatives to X), which are about the brands they name: both inflate your numbers. If you track such prompts, your headline numbers change from today. The new By intent card shows visibility for each intent, and one click counts branded and comparison prompts again. Insights, reports and share links use the same numbers, and Looker Studio gets a Prompt intent field.

A new Search queries page for each project (AI visibility → "What the engines searched for") lists the searches ChatGPT, Gemini, Claude, Perplexity and Grok ran while answering each prompt. It shows the words engines add to your prompts, flags searches that never brought up your site and English searches on non-English prompts, suggests the page type a search calls for (best-of lists, comparisons, dated pages), and tracks any search as a prompt in one click. Prompt fan-out now shows the real searches next to its simulation for prompts you track.

Win rate is new on your AI visibility dashboard: of the answers that name you, how often you're the first pick, overall and per engine. Insights has a visibility funnel per engine (the engine reads your site, cites it, names you, puts you first) with the fix for each drop-off. Every prompt now shows how stable its result is per engine: hover a cell in the prompt × engine matrix or click a prompt. After 5 checks it is labelled stable in, coin-flip or stable out, and the coin-flip prompts, where content work can tip the result, are listed on Insights. Answers reused from the 24-hour cache don't count as extra checks.

Open any AI answer to see the searches the engine ran and every page it read, with the pages it cited marked. ChatGPT, Gemini, Claude, Perplexity and Grok show both; Google AI Overviews, AI Mode and Bing Copilot show the pages behind their answer; DeepSeek and Mistral answer from memory. Gemini citations now credit the actual website instead of a Google redirect link, so your site can count as cited by Gemini. Grok now counts only the sources its answer cites, not everything it looked at, so Grok's cited-source numbers are lower from today. ChatGPT checks for prompts without a location no longer default to United States results. Answers checked before today don't include searches or pages read.

Perplexity retired the API behind our Perplexity checks on 27 September, so they now use its replacement: the same Perplexity model with live web search, and answers still list the sources they cite. DeepSeek checks moved to DeepSeek's current model and Claude checks to the latest Claude Sonnet. Nothing to do on your side, and your history is unchanged.

Team members limited to specific projects now see only those projects on every page — AI visibility, Insights, Alerts, Reports, share links and the AI Lab project pickers — not just in the sidebar.

AccuRanker joins the migration importer: upload a keyword export and your keywords arrive with positions for both desktop and mobile, search volumes, CPC and competition — plus AI Overview presence where the export recorded it. Same rules as every migration import: free, and safe to re-run.

Moving from SerpRobot? Import → Migrate from another rank tracker brings your keywords AND every historical position check, so progress charts, starting positions and change tracking carry over from day one. Search volumes come along too, and re-importing the same file never duplicates anything. More providers coming.

Google AI Mode has started showing ads inside its generated answers. Asheard now separates those from the real answer: they never count towards your visibility or citations, and Insights gets a new card showing who is paying to appear on your prompts — including whether a tracked competitor is. Individual answers list their sponsored placements too.

Every workspace now has a personal referral link — find yours under Refer & earn in the sidebar. When someone signs up through it and becomes a customer, you get bonus credits, automatically. We only ever show you when your referrals joined, never who they are.

AI Visibility Trend charts monthly mentions and AI search volume for any brand or domain back to August 2025. New & Lost Mentions shows the churn behind flat totals — which AI answers you newly appeared in and which you dropped out of. Brand Radar reveals which brands AI actually recommends for any topic, so you can spot competitors before you track them. All three are credit-priced with automatic refunds when nothing comes back.

The Insights "know" cards — Recall, Top-of-mind, and Fact accuracy — now chart every run you've made: a sparkline, the latest score, and how much it moved. Run them monthly and watch whether your work is actually sticking in the models' memory.

The sidebar now separates Rank tracking (your domains) from AI visibility with clear group labels — no more hunting for a site in the middle of AI links. Keyword history also got cleaner: failed checks now show a simple "credits refunded" note instead of technical internals.

Track whether Bing's Copilot AI answers recommend your brand — the ninth engine, included in Pro and Agency plans at the low SERP-surface rate. Like Google AI Overviews, an empty result is a signal too: Bing chose not to show an AI answer for that prompt.

Gemini checks now run on Google's latest generally-available model — Google retired the previous one on their side, which briefly made Gemini runs fail (always auto-refunded).

You can now move a domain between groups from its settings — keywords and history ride along. Fixed doubled headers on two tool pages. URL tools now explain nicely when an address can't be checked instead of showing an error page, page reading got more reliable, and refunded AI Lab runs are rarer thanks to sturdier answer parsing.

The Insights page now tells the story in three acts: what engines KNOW about you (your latest Recall, Top-of-mind, and Fact-accuracy results now persist and show here, with age and one-click re-runs), what they FETCH about you, and how STABLE it all is. Three new free cards too: Mention prominence (lead recommendation or footnote?), Co-citation neighbours (who answers lean on next to you — and who owns the room when you're absent), and your own answer-volatility chart per engine.

Prompts in Slovenian, German, and 30+ other languages now carry an explicit answer-language instruction to every engine. Short prompts were ambiguous between related languages — a Slovenian prompt could come back in Czech. Cached answers refresh within a day.

Prompt Fan-out maps the sub-queries an AI actually runs behind your question and flags which facets your tracked prompts don't cover yet — with one-click adding. Top-of-mind Check asks every engine, from bare memory, which brands it names first for your category. Plus two free tools: a public AI Answer Volatility index (how much each engine's answers changed day over day, from our aggregate tracking data) and an AI Facts File generator — llms.txt's sibling for canonical brand facts, with ready-to-paste JSON-LD.

Two new Insights cards: every AI answer classified into the four mention×citation states — including the Evidence Gap, where your content fed the answer but someone else got the credit — and a per-engine breakdown of WHAT each engine cites (forums, reference sites, news, reviews…), so you invest where that engine actually looks.

Brand Fact Audit (what each engine believes about you from memory — and where it's wrong), Recall Check (memory vs. search: is your visibility durable or rented?), Selection Duel (which of two pages an AI would pick as its source, judged in both orders to cancel position bias), and Snippet Lens (exactly which sentences an AI would lift from a page). Credit-priced, auto-refunded when nothing usable comes back.

Type your brand and one question on our homepage and see the actual AI answer — with your brand highlighted, or conspicuously missing while a competitor gets recommended. Two free checks a day.

Every alert digest is now permanently recorded per project — open a project's Alerts page to see everything that ever fired, long after the bell was cleared.

Track whether DeepSeek and Mistral's Le Chat recommend your brand — both included in Pro and Agency plans at the low engine rate, no add-on fees. Prompt generator also went worldwide: one clean English template set plus an optional market field.

The prompt generator now speaks four languages (EN/SL/DE/HR) and can localize prompts to your market; the CTR estimator got a more conservative curve; the Tools sidebar section collapses like the project tree.

Research ANY brand's AI visibility — even ones you don't track. Search a large corpus of real AI prompts and answers for mentions, sample answers, and the sources powering them. Credits per lookup, auto-refunded when nothing comes back.

"Can AI read this page?" (JS-shell detector), an AI prompt library generator, a CTR estimator with an AI-Overview toggle, and a maintained AI crawler list with copy-paste robots.txt rules.

Page titles now condense out of a subtle violet mist. Blink and you'll miss it — that's the point.

PDF reports got a redesign: executive summary, a visibility trend chart, movement arrows on every keyword, and citation-share bars — the same data, dressed for a client meeting.

The project tree no longer folds shut every time you click into AI visibility, Insights, or a domain.

Tag prompts by persona, product line, or funnel stage — then filter the whole AI visibility dashboard, every chart and score included, to just that segment.

Eleven guides covering everything from your first rank check to reading the perception map — public, searchable, and linked from the app sidebar.

Add your company name, address, and VAT ID on the Billing page — every invoice carries them, valid EU VAT IDs bill with reverse charge, and PDFs download straight from the new Invoices section.

The credit ledger now speaks human: "Monthly plan credits", "AI visibility check · run #5" — instead of internal codes.

Scheduled Google AI Overview checks are processed in the background queue — same data, lower cost, which keeps credit prices where they are.

Insights now shows exactly which of your URLs the AI engines cite, plus Citation Gaps: source pages where competitors are mentioned and you are not.

AI Bot Access Checker (does your robots.txt block the crawlers behind ChatGPT, Claude, Perplexity?) and the AI Citability Analyzer (paste content, get a citability score with fixes).

Every project can generate a read-only rankings page for clients — no login needed, revocable anytime.

Purchase confirmations after checkout, and the trial banner now names exactly which module is still on trial.

Every domain now gets a monthly snapshot of ALL keywords it ranks for — across every country — included free with every plan. One-click tracking for anything worth watching.

Change a keyword's text, country, or language without deleting and re-adding. History stays.

The new bell shows run alerts even if you haven't configured email or Slack — including the new "Rising rival" alert when an untracked brand keeps appearing in AI answers.

The Insights page: AI Visibility Index, brand perception map, rank-vs-AI quadrant, and concrete recommended actions.

AI-readiness analyzer, llms.txt generator, and the GA4 AI-traffic setup guide — free for everyone.

Agency plans can brand reports with their own logo, colors, and client name.

Rank tracking and AI visibility in one product: keywords across Google, brand presence across ChatGPT, Gemini, Claude, Perplexity, Grok, and Google AI Overviews.