What's new in AnswerRank

The dashboard now leads with the information that changes decisions, including the number of unique URLs used across all responses in the selected period. A domain count tells you where you are cited; a unique URL count tells you how broadly. One page quoted in every answer and forty pages cited once each look identical at domain level and mean completely different things for your content strategy.

The new metric is not trapped on the dashboard. Filter by it, and it travels into the sources export, so the figure in your client report matches the figure on screen.

Sources queries were reworked for projects tracking thousands of citations. Pages that took several seconds to open now load immediately, and switching date ranges no longer reloads the whole list.

The same page arriving with different utm and campaign parameters was being counted as several separate sources, which inflated both counts and the domain list. Parameters are now stripped before a source is recorded.

Applying a filter no longer sends you back to the first page, and the result count reflects the filter instead of the full list. Exports taken while a filter was active now contain the filtered rows rather than everything.

Microsoft Copilot now runs alongside ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. Copilot sits inside Windows, Edge and Microsoft 365, so for B2B and enterprise brands it reaches buyers your other tracked engines never touch.

Select as many suggested prompts as you want and add them in one action. Setting up a new market used to mean clicking through suggestions one at a time; now it takes a single pass.

Newly added prompts show an explicit starting status, and prompts still collecting market share data show an analysis-in-progress indicator. You can finally tell the difference between a prompt with no visibility and a prompt that has not finished running.

A subset of projects stored Gemini answers with no citations attached. The collector was reading the citation block before Gemini had finished rendering it, so the answer saved and the sources did not.

The Sources page has a new structure, updated filters and scrollable menus for sources and search queries. Projects citing thousands of URLs were technically correct but impossible to read, so the page was rebuilt for that scale.

www and non-www versions of a domain now merge into a single source, so citation counts stop being split across two entries for the same site.

The last inactive prompts moved from API collection to scrapers, so every prompt in your account runs through the same pipeline. Historical trends and current data are now measured the same way, which is what makes long-range comparisons trustworthy.

Sorting and delete actions across accounts and projects, so agencies running dozens of client projects can find and clean up what they need without asking support.

Tag adding was reworked and the data filter updated across pages. Tagging several hundred prompts no longer stalls halfway through.

In some chats the name of a linked publication was being scored as a mention of your brand, which lifted visibility on prompts where you were never actually named.

AI engines rarely search for your prompt as written. They expand it into several sub-queries, search those, and build the answer from what comes back. AnswerRank now captures that expansion, so you can see which searches your visibility actually rests on and which ones you never appear in. This is where most visibility is won or lost, and until now it was invisible.

More dimensions for comparing prompts, periods and competitors, so you can show whether a change in visibility came from your work or from a shift in how the engine answers.

Moving from Prompts to Competitors dropped the tag you had selected, so the second page showed the whole project instead of the segment you were looking at.

Retailer tracking for products, a brand search bar and a ChatGPT-focused review flow. For retail and ecommerce brands the decisive question is often which retailer the AI sends the buyer to, not only whether your name appears.

Run response analysis through Bright Data as an alternative to OpenAI. A second provider lets results be cross-checked, and a single provider outage or model change no longer interrupts your data.

Export visibility per prompt for competitors alongside your own numbers, in one file. Client reports that used to require merging two exports by hand are now a single download.

Re-scan sources you have already collected for brand mentions instead of waiting for the next crawl. Useful right after adding a competitor or correcting how a brand name is detected.

A retailer written with and without its legal suffix appeared as two separate results in product tracking.

Market share now appears on each individual prompt, not only at project level. A healthy project average can hide the fact that your brand is absent from the handful of prompts that actually precede a purchase.

More ways to filter prompts and compare them against each other, so you can isolate where visibility moved instead of reading down the whole list.

Logos failed to resolve for competitors added manually, leaving blank spaces in charts that were otherwise ready to screenshot.

Sentiment was rebuilt across two releases, historical data was recalculated, and every score now carries the context behind it. Dashboard and Competitors previously reported different sentiment for the same brand, which made the metric impossible to cite with confidence. They now agree.

More competitors per project on the agency plan, for categories where ten brands is not a realistic picture of the market.

Search across every account for a project, prompt or product. Support questions get answered without first working out which account the data lives in.

Pull additional attributes for competitor brands inside a project, so competitive analysis is not limited to a visibility percentage.

Configure faster crawl cycles for individual projects. Brands in fast-moving categories, or in the middle of a campaign, can be tracked more often than the standard schedule allows.

Reloading a page dropped the selected range back to the default seven days instead of keeping what you had chosen.

Every active project now collects answers from the AI interfaces instead of through APIs. The two are not the same product: interfaces personalise, retrieve from the live web and cite differently than API endpoints do. Collecting from the interface means your numbers describe the answers your customers actually receive.

Tracked Gemini projects moved to the 3.5 model line, so results reflect the version users are being served rather than the one they have left behind.

Responses beyond a certain length timed out during collection and were skipped, leaving gaps on the days those prompts ran.

Trigger a crawl immediately instead of waiting out the 11-hour cycle. Useful when you have just published a change and want to know whether the engines picked it up.

One view of collection health across projects, so a stalled crawl becomes visible to you before it becomes visible to a client.

After deleting prompts, the count in the sidebar kept the old total until the next full page load.

Group near-duplicate prompts into a single row. Large projects accumulate many phrasings of the same question, which pushed the prompts that matter off the first page.

Apply or change tags on many prompts at once, so reorganising a project after a campaign or a market launch takes minutes rather than an afternoon.

Sort prompts by active and inactive state to see what is currently consuming your queries and what is paused.

Grouping similar prompts occasionally rolled up the parent row without its results, which made an active prompt look inactive.

Track the specific products engines recommend, using the same visibility logic applied to brands. In categories where the AI answers with a product shortlist, being mentioned as a brand and being recommended as a product are two different outcomes.

Source exports now carry the country each result came from, which matters when the same brand is cited by different publications in different markets.

Preview Gemini versions were replaced with the stable line, so results stay consistent between runs instead of shifting when a preview is updated.

Shorter deploy times, and the Ctrl+V bulk paste for prompt lists is now visible in the interface instead of being an undocumented shortcut.

A retried collection run could store the same Perplexity answer twice, which double-counted its sources for that day.

AnswerRank now reads answers from the live AI interfaces rather than requesting them through APIs. This is the largest change of the year and the reason to trust everything above it in this timeline: an API returns a model's output, while the interface returns what a person is shown, with its own retrieval, personalisation and citations. Measuring the second is the only way to report on visibility your customers experience.

Groundwork for collecting Microsoft Copilot answers, which went live as a fully tracked engine in August.

Google AI Overviews were sometimes read while the block was still expanding, storing a shortened version of the answer.

Each project now has its own prompt library, rules and lifecycle, delivered in two parts. Agencies running many brands no longer have prompts, tags or settings leaking between clients.

Filter by tag across pages, and carry those tags into sources and prompts exports. The structure you build inside the tool now survives in the reports you send out.

See what each project and each prompt costs to run, with export. Prompt tracking carries a real per-query cost, and this is what lets you price a retainer, justify a project or cut the prompts that are not earning their place.

One page showing the state of every account and project, so expired trials, inactive projects and stalled collection surface in the same place.

Legends on dashboard graphs, corrected dark mode contrast, a repeat-tutorial button, project sorting under accounts, and collaborators seeing shared projects on their first login.

Spend for the current day appeared only after the next collection cycle, so a project mid-campaign looked cheaper than it was.

Push project data straight into a Google Sheet and keep it refreshed. Teams that build reporting in spreadsheets stop exporting and pasting files every week.

The read-only API accepts filters, so an integration can request the prompts, dates or competitors it needs instead of pulling everything and discarding most of it.

Projects with one prompt produced an export containing only headers.

Store the web searches ChatGPT runs while composing an answer. Your visibility depends on what those searches return, so seeing them tells you which queries your content has to win, rather than inferring it from the prompt.

The prompts behind the automated audit were reworked, which raises the quality of the recommendations it produces.

Type your own industry when the predefined list does not fit, and prompt generation adapts to it. Niche and B2B categories were being forced into approximate industries, which produced approximate prompts and therefore approximate data.

Watch project and prompt generation progress instead of waiting without feedback. You can tell whether a slow project is working or stuck.

Gemini 3.0 previews added, Haiku 3 replaced with Haiku 3.5 across existing prompts, and GPT-5 removed from selection. Tracking a model that users are no longer served produces data nobody can act on.

Google sign-up moved to the front and the form cut back to what is actually required.

The days remaining on a trial were calculated in UTC rather than the account time zone, which shifted the number for users outside Europe.

Dashboard, Prompts and Competitors share one date range picker, and exports respect whatever range you selected. Comparing two pages used to mean setting the same dates twice and trusting that they matched.

Downloaded files are named after the project and the page they came from, so a folder of client exports stays identifiable a month later.

The country shown next to a project came from the account rather than the market the project was set up to track.

Sentiment is now calculated against measured visibility and ranking instead of a free-form model judgement. Scores became reproducible and comparable between periods, which is what a metric needs before it belongs in a client report.

When a brand has no visibility, rank is left empty rather than filled with a penalty value. That invented number was being averaged into trends and quietly distorting them.

Endpoints for project data and analytics, prompts and competitors, with API key management in the admin panel. AI visibility data can now sit in your own dashboards next to search, paid and analytics reporting instead of living in a separate tool.

Request verification, OAuth token exchange, store and profile fetch, account linking, and redirect into an existing project or a pre-filled onboarding. Together this is the groundwork for installing AnswerRank directly from a Shopify store.

Per-prompt opportunities plus a unified report for the whole project, which turns an audit into something you can hand straight to a content team.

Newly generated keys did not appear until the admin panel was reloaded, which made it look as though creation had failed.

Both metrics sit directly in the prompts table, so you can scan a hundred prompts for the ones that need attention without opening a single one.

A rebuilt onboarding flow and an audit screen reorganised around what to do next rather than what was measured.

Tooltips rounded small values down to zero, so prompts with genuine but low visibility looked identical to prompts with none.

Select many prompts and act on them together. The page had been built for dozens of prompts while users were running hundreds.

A prompt search bar, sorting by last prompt added, and 7, 14 and 30 day presets with a calendar dropdown.

Row height stays fixed no matter how long a prompt is, so one long question no longer stretches the table out of shape.

Prompt suggestions are generated in the market's language instead of translated from English. Translated prompts do not match how people actually ask in that market, so the visibility they measure is not the visibility you have.

Gemini upgraded from 2.0 to 2.5 platform-wide, Perplexity configuration updated across plans, prompts migrated from GPT-5 to GPT-5.1, and Claude availability aligned to plans.

Top metrics respond to the selected date range, with a clearer tooltip and visibility and position reordered so the number you read first is the one that matters.

Updated prices, prompt counters and monthly query limits, reflected on the landing page pricing cards and the free trial rule.

Sentry was added, which shortens the distance between something breaking in production and the team knowing about it.

Export domains, linked prompts and complete URLs to Excel and CSV. Sources are the actionable half of AI visibility, because they name the pages you need a mention on, and that list now leaves the tool with you.

Reworked pie chart, colours that stay stable when a competitor is added, and pagination for manually added lists. A chart that reshuffles its colours between screenshots cannot be used in a monthly report.

Competitor names beyond a certain length were truncated mid-word in the chart legend, making two similar brands hard to tell apart.

The Opportunities section audits a project and returns prioritised recommendations. Up to this point the platform told you where you stood; this is the first release that tells you what to do about it.

See the vocabulary AI engines consistently use when answering questions in your category, and where your brand is missing from it. It shows the language your content has to speak in order to be retrieved at all.

An audit started immediately after adding prompts could use the previous collection run, producing recommendations for data you had just replaced.

Backfill and aggregate historical visibility, with aggregation limited to the selected date range. New projects open with a trend behind them instead of an empty chart and a month of waiting.

One click exports the full source list for a project.

A two-link referral system for sharing access.

Login frequency, next payment date, linked project and user data, and filters for users and projects.

Missing data points display averages instead of breaking the line, so one failed collection no longer looks like a collapse in visibility.

A pass over every step of the first run, guided by where new users were actually dropping out.

Notifications for account and project events, sent from a verified answerrank.ai domain so they arrive in the inbox rather than in spam.

Projects stop when a plan is cancelled, and failed payments and cancellations are handled explicitly instead of leaving an account in an undefined state.

Reset links stopped working sooner than the stated window, forcing users to request a second one.

A dedicated page for reviewing, editing and accepting the prompts the platform proposes. Which prompts you track is the decision that determines whether any of your visibility data means anything, so it needed a proper workspace rather than a step buried in setup.

Prompt suggestions are drawn from curated industry pools instead of being generated from scratch every time, which makes them more consistent and comparable between brands in the same category.

Account, project and plan management without developer access.

The redesigned sign-in page went live.

The performance graph puts the brands you added ahead of auto-suggested ones, so the chart answers the question you arrived with.

Batched analytics requests and proper queue handling. Large projects had been timing out; after this release they complete.

The first full export across dashboards, prompts and competitors, which is what turns tracked data into something you can put in front of a client.

The Sources filter lists only the models a project actually uses, instead of every model the platform supports.

Selecting a range that contained one day of data rendered an empty chart instead of a single point.

See how the answer space splits between your brand and everyone else in it. A visibility percentage alone never says whether you lead the category or trail it, and market share does.

Competitor position is shown as a rank number rather than a percentage, because rank is what people compare, remember and act on.

Switch the bar chart to the complete list of up to 70 brands, on a tighter layout, so categories with a long tail of smaller competitors can be seen in full.

Chart colours were assigned per session, so the same competitor appeared in a different colour each time the page opened.

Set what a project is called and what the tracker searches for as two separate fields. Display names had been leaking into brand detection and quietly distorting visibility.

A plan for agencies evaluating the platform across several client brands before committing.

Models outside your plan appear greyed out with an upgrade path instead of being hidden, so you can see what a higher plan would add before paying for it.

The models available for prompt generation were audited and brought up to date.

A queue system triggers collection when a project is created and keeps runs ordered under load, which is what stops prompts failing silently when many projects run at once.

See exactly when the next batch of prompts starts instead of guessing at the schedule.

The redundant graph and upper metrics block were removed, prompt review opens in a new tab, and filters survive navigation and refresh.

Editing a project reset its prompt language back to English, so the next batch was generated in the wrong language.

Group prompts by campaign, market, funnel stage or product line, then filter the dashboard by tag. Without that structure, a few hundred prompts is raw data rather than reporting.

Switch auto-suggestion off and track only the brands and prompts you chose. Auto-suggestion is useful while you are exploring a category, but the moment you report to a client the competitive set has to be the one you both agreed on.