Changelog
Every release, newest first: new features, improvements and fixes.
Release v0.0.26
Slow scans no longer fail
A scan that runs past the usual timeout used to be marked failed while the engines were still answering. It now stays in progress, and your results show up when it finishes.
Usage is counted per window
Day, week and month usage each have their own count. A workspace on a reverse trial keeps the paid data-retention window, a duplicate engine add-on is refused before it is charged, and a payment suspension now applies in full or not at all.
Free-tool and onboarding fixes
The run button on the llms.txt generator, the llms.txt auditor and the AI visibility score disables while the domain is empty. Scan-completion notifications reach the in-app bell again. Onboarding clears the terms error as soon as you tick the box, and the final step no longer hangs.
Support links
The user menu and the error and 404 pages link to support in one click.
Roles and plan gates
Promoting a member to admin is owner-only. Deep Research and Deep Verify appear only on the plans that include them.
Release v0.0.25
Viewers can read but not change anything
Invite a teammate as a viewer and they can read everything the workspace sees (scans, rankings, citations, reports, alerts). They cannot run a scan, share a report, edit a project, or start a Deep Research or Deep Verify run. The same rule covers members, domains, prompts, competitors, recommendations, content gaps and search.
Partial GA4 pulls say so
When a GA4 pull returns less than the window you asked for, the card now says so, the same disclosure Search Console and Cloudflare already give.
Fewer dead ends
A returning user lands further along in onboarding instead of seeing the account form again. Sign-out clears the session on every path, and an onboarding draft resumes where you left it.
Dependencies updated
Key dependencies were updated to clear published security advisories.
Release v0.0.24
Tabbed feature explorer
The homepage is rebuilt around a tabbed explorer. Each capability, from multi-engine monitoring, visibility score and citation intelligence through automated alerts, competitor benchmarking and recommendations, gets its own panel linked to the page that delivers it, and the page loads immediately.
Pricing and plan copy agree
Every surface that describes a plan now says the same thing. The pricing page and the start-here guides agree on the trial length and the competitor limit, and dashboard copy no longer promises a capability a plan does not have.
Accuracy risk and structured webhooks
The Alerts page names heuristic accuracy risk instead of leaving a verdict implied. Multi-engine scans state which models were queried, and Copilot is tracked through search data rather than as a model. Outbound webhooks now include the structured findings behind an alert.
Keys, domains and return links
An API key stops working when it expires. A custom domain used for SSO must prove ownership with a DNS record before it can be claimed, so nobody can take a hostname that belongs to someone else. Signing in returns you to the page you came from.
Forms and metadata
Form fields are labelled for screen readers, iOS no longer zooms the page when a field is focused, and titles, structured data and social cards are corrected across the marketing site.
Release v0.0.23
Integrations hub
The Integrations page is now the home of Search Console, domain ownership, GA4, Cloudflare, the traffic tag, Slack, Teams, webhooks, Looker Studio and the MCP server. Each one has a guide and a live setup, and a dead Google grant now surfaces in the property picker.
On-demand variance sampling
Deep Verify re-asks each active tracked prompt across your project's engines on demand, so the confidence bands come from repeats rather than a single pass. A round bills like a normal prompt check: one credit per delivered sample, free when the round errors.
Signup finishes with a scan
Signup creates your project and runs the first scan. If the scan cannot run, the account still exists and the dashboard says why.
Per-plan email caps
Invites, digests, alerts, scheduled-report delivery and one-off shares count toward a monthly allowance with a daily ceiling. Batched digests and lifecycle mail do not. Workspaces at the ceiling see where the limit came from.
Appeals for blocked domains
A domain blocked by the acceptable-use guard can file an appeal. An operator reviews it, and a resolved appeal is honoured on later checks.
Turnstile on anonymous tools
The anonymous AI-spending tools now verify visitors with Cloudflare Turnstile.
Homepage engine coverage
The marketing homepage now names which of the seven models and three search surfaces you can track.
Pricing copy
The price page, the comparison table and the emails agree with what each plan can actually do.
Release v0.0.22
SSO and SAML
A company identity provider (SAML or OIDC) can now provision Agency seats without a separate password, alongside email and Google sign-in. The pricing comparison lists SSO, so the plan page and the login screen agree.
Conversions and revenue attributed to AI referrals
The Traffic page attributes conversions and revenue to sessions that arrived from an AI answer and then converted, through your connected GA4 property. The question it answers is whether the AI channel is selling, not just whether it sends clicks.
Communities as their own source type
Citation tracking groups sources like Reddit, YouTube and Quora as communities: replies, threads and posts are counted as communities instead of being scattered as unnamed domains, so you can see where an engine learned to recommend you.
Aggregate endpoints and a Looker Studio connector
The public API gained aggregate endpoints built for BI tools, and a Looker Studio connector builds dashboards over your visibility data without scraping.
Paid-plan API access
The public API and MCP server became paid-plan features. Free workspaces keep the dashboard but not the programmatic surface. Keys carry read and write scopes and an optional expiry; rotating a key means creating a new one and revoking the old. A revoked key stops working on its next call.
Every measured figure reports its uncertainty
A brand at a 40% mention rate across 20 answers is not the same as one at 40% across 2, and the dashboard now says so: every measured figure carries a confidence band, using the Wilson method described in the published measurement methodology. On-demand sampling that tightens the bands from measured repeats followed in v0.0.23.
In-app notification inbox
An app bell feed pairs the outbound email, Slack, Teams and webhook channels, with badge updates as findings arrive.
New pricing page and a reverse trial
The pricing page was rebuilt, and a 14-day Business trial starts when you begin paid work. Feature pages gained an index describing what each one does.
Free tool rate caps
The anonymous AI tools share one per-IP budget across the seven AI-spending routes.
Consistency pass
The dashboard was rebuilt around one idea: every surface reads the same. Stat cards share one language, charts share one style, filters and downloads sit where you expect them, and the sidebar groups routes by task.
Multi-word queries across the site
Site search handles the queries people actually type, so "citation checker" and "check citations" both land on the citation checker instead of a page of "AI" matches.
Release v0.0.21
Per-competitor and per-engine history
Every completed scan now records each tracked rival's share of voice, and share of voice per engine. The Competitors page plots those series next to your own, so a rival gaining ground shows as a trend rather than a single latest snapshot. Older scans that predate this release show a gap, never a zero.
Place inside an answer
When an engine names a competitor, the mention carries a 1-based place, or unranked when the engine named them without a rank. Rankings and the competitor drill-down show it.
Ground-truth verification
Attach pages on your own domain or a verified host, and a flagged claim is re-checked against them: when the check completes, a matching quote and source URL appear on Alerts. Reputational associations are detected separately from accuracy risk, so a controversial association is not reported as a factual claim.
Up to 20 markets per project
A project has a home market plus up to 20 extra markets, managed on the Domains page. Each market is scanned on your plan's cadence with its own freshness, shown through the dashboard's market filter, and answered in that market's language. A market with no scans yet says so. Each market is one prompt-check: N markets, N checks, N credits, with no per-region discount and no hidden surcharge.
Industry benchmarks
The benchmarks page publishes a bucket only once five or more opt-in workspaces in that industry exist, and household names are kept out of customer percentiles so a familiar brand cannot become "you rank above N% of peers." The named-brand league runs separately, and the aggregate never exposes an individual score.
Business and Agency price update
Business moved to $49.99 and Agency to $299.99, the first change since the plan reshuffle. The comparison tables, checkout copy and entitlements match the pricing page.
Site search polish
The ⌘K dialog and search results page got their first polish pass after shipping in v0.0.14: stable triggers, better result rows and a header that stops shifting when search or the theme toggle opens.
Release v0.0.20
Paid placements are flagged
ChatGPT answers and AI shopping surfaces now carry labelled ads. Every scan flags an answer with a "Sponsored" or "Ad" block, a paid shopping or product card, or another clearly promotional placement, and the Rankings page shows the flag per engine and per answer. A missing flag is not proof of no ads: older scans never recorded the field.
Demand from your own scan history
The prompt library ranks tracked prompts by demand already visible in your scan history: how many distinct phrases engines have been asked, or have claimed, that share a significant keyword with the prompt. High, medium and low are a coarse band over that overlap count, not a market volume. Before the first completed scan the library says "no evidence yet". Search Console import still carries estimated monthly volume where a keyword provider is configured; this band is the AI-side counterpart.
Structured data is graded, not just recommended
The site-wide content audit grades each page's structured data with the same validator the schema validator ships: a 0–100 score from the JSON-LD in the page <head>, plus the schema.org types it covers. A missing schema is a missed opportunity, not a zero, and a broken one scores lower. Weak grades come with a concrete next step.
Light appearance
The app now follows the system appearance, or a light or dark choice you set. Marketing pages, the dashboard and shared reports use the same tokens, so a report opened in daylight is not forced into a dark theme.
Rankings, alerts and content gaps
Rankings gained per-query deltas, win/loss, sentiment, a Google-gap annotation and JSON export. Alerts gained an unread bell, more trigger types, and per-project threshold overrides on paid plans. Content gaps split citations from mentions and added an engine-visibility matrix.
Traffic, Search and Domains
Traffic gained a realtime view, UTM channels, custom date ranges and CSV export. Search Console flags sustained content decay: clicks that fell past the same floor in two consecutive windows, not a one-week blip. Domains gained a CNAME verification method, a score-history sparkline and an SSL-expiry callout.
Reports
A scheduled shared report can be edited after it is created (recipients, cadence, pause) and shows its delivery history. Saved templates remember a report-builder setup. Agency workspaces can serve the report on their own host (reports.youragency.com): the branding tab attaches and verifies the domain, and a custom mail From is set once the sending domain is verified.
CSV on every read, and on the API
Every dashboard table (recommendations, Search Console, traffic, the monitoring hub) exports CSV. The public API gained the same shape on its read routes. A scan still exports one scan at a time from the browser; there is no bulk history dump.
Release v0.0.19
Answer archive and diff
From the rankings drill-down, any past scan's answer to a question can be reopened, with the wording changes against the previous scan shown as additions and removals and the claims the answer made or dropped. That is answer text diffing, which makes a hallucinated-claims problem visible: what an engine changed its answer to, and when.
Search Console queries to tracked prompts
The Search Console top-queries table imports the queries already earning you clicks as tracked prompts in one click: demand from your own property instead of a guessed list. Imports land inactive, since activation counts against your plan's active-prompt cap. They dedupe case-insensitively against your existing prompts and carry estimated monthly volume wherever a keyword provider is configured. The tracking Google search performance tutorial covers the Search Console side; choosing tracked prompts covers what happens once they are in the library.
AI Mode moved out of the engine picker
Google AI Mode joined the weekly SEO-data pass in v0.0.18. It no longer appears in the AI engine picker in Settings. Like Google AI Overviews, it is not a model you select: on paid plans it runs automatically every week alongside AI Overviews, answered by a search lookup rather than a chat model and billed on the SEO-data meter instead of your credit allowance.
Search Console's generative AI report
Search Console's generative AI performance report (announced June 2026) is impressions-only, reports AI Overviews and AI Mode together, is rolling out to properties gradually, and is not yet in Google's API. The knowledge base says exactly that, so the report is not read as more than it is. AI Mode visibility in LLM Metrix comes from the weekly SEO-data pass instead.
Microsoft Copilot on the weekly pass
Microsoft Copilot is tracked on the weekly SEO-data pass as its own search surface: the answer panel from Bing's copilotsearch page. It is billed on the SEO-data meter, not your credit allowance, and skipped when no search provider serves it.
Research runs
A Deep Research run asks a research model to answer your tracked prompts with a full report, then grades each report for brand mention, rank, stance and claims. On paid plans it bills from your credit allowance on its own meter.
Generated artifacts
A recommendation on the board can be turned into a deliverable (an llms.txt file, schema JSON-LD, or a content brief) and copied or downloaded from the item.
Read and write endpoints, plus MCP
The developer API grew from two read-only endpoints to a read-and-write REST surface: trigger scans, retry jobs, manage prompts and the webhook URL, plus an MCP server for AI agents, all on the same personal access tokens.
Release v0.0.18
Alert history
Every finding a scan derives is now a durable record you can page back through, not only a notification at the moment of the scan. The Alerts page reads that history: each occurrence stays attached to the scan that produced it, read state is per member, and nothing is lost when you leave the page and come back.
Background work is actionable
Activity, domains, prompts, content gaps, recommendations and reports each gained an action surface: a stuck job can be retried from where it failed, a running monitoring loop shows what it is doing, and a gap you have acted on stays visibly acted on.
Acceptable-use domain screening
New projects are screened against known adult, gambling and drug lists before they can be tracked, with a hard block on adult content and a soft flag on the other categories. The lists refresh nightly. A screened-out domain can appeal the decision, and a resolved appeal is honoured.
Per-engine position history
The Rankings page gained a per-engine position timeline, stat-card deltas against the previous scan, and query search with drill-down, so a movement in your aggregate position can be traced to the engine and the query that moved it.
Editable facts, site-level @graph and auto-detection
The schema generator can edit facts in place, emit a site-level @graph, auto-detect the entity type, score completeness, and cover LocalBusiness and Event. It checks structure rather than only well-formedness.
Next steps on the Search-to-AI gap
The Search-to-AI gap analysis on the Search page now offers a next step when Google ranks a page or query that engines do not cite, instead of stopping at the finding.
Empty states, errors and confirmations
Empty and error states say what happened. Plan limits surface where they stop you. Destructive actions ask for confirmation before they run.
Release v0.0.17
Rival comparisons across the dashboard
Rankings, Citations, Search, Content Gaps and Competitors each gained a competitive view: where rivals rank higher than you, which of their prompts you do not track, per-engine breakdowns, and the queries where Google and the AI engines disagree about you. Competitor projects can re-scan automatically on your plan's cadence. The competitor benchmarking tutorial covers the workflow.
Content gaps with demand attached
The content-gaps analysis enriches each gap with search demand, the competitors already ranking for it, its Google status and an opportunity estimate, so the table tells you what to write next, not only what is missing.
Scheduled reports
A shared report can be scheduled: pick a frequency, add recipient emails, and the report link is delivered on schedule. A report can also be pinned to a specific scan, so its link always shows that dated snapshot rather than the latest result. The printable and public report views share one layout.
Recommendations board
Recommendations moved from a list you could not track to a durable board. Each item has a status (suggested, to-do, done, snoozed or dismissed), so a fix you have made stays done instead of resurfacing, and the board keeps the evidence behind each recommendation so you can measure its outcome against the next scan. See the GEO recommendations feature for what the board produces.
Domains as a monitoring hub
The Domains page is now the hub for everything about a project's domain: verification status, tracking state and entry points to each project's health.
Richer views
The Traffic page gained a GA4 channel breakdown, a Cloudflare card and GA4 backfill on connect, with a wider bot taxonomy. The Search Console view gained device and country breakdowns with period-over-period comparison on every row.
More triggers
Alerts gained lost-citation and score-floor triggers, editable position thresholds, a per-engine delivery filter and per-workspace snooze, with a daily digest option that batches lower-priority signals.
Clearer consent
Signup asks for explicit consent to the terms and privacy policy, a refund policy page sets out the money-back terms, and checkout requires accepting the DPA.
Release v0.0.16
Three plans, one credit
> A note on the plan ladder. This entry describes prices as they shipped > then (Business $29.99/mo, Agency $199.99/mo). Today's figures are on the > pricing page. The sentence below reflects what shipped then.
Brand safety, claim by claim
The brand-safety view now works claim by claim: each claim an engine makes about your brand is scored for accuracy risk and sentiment, and the positive, neutral and negative split is shown as a distribution over time. Every finding links to the fix it points at, so a claim that hurts you names the remediation, not just the problem.
Alerts you can tune
Alerts are now per-trigger: each trigger type (position drops, share-of-voice shifts, citation losses) has its own editable threshold and frequency mode, so a low-stakes signal can go into a daily digest while a position collapse pings immediately. See the setting up alerts tutorial. A baseline-deviation mode alerts when your metrics drift from their own recent history, and a position-drift mode watches ranks shift across scans. Delivery is by email or Microsoft Teams webhook.
Bulk import for tracked prompts
Tracked prompts can be imported in bulk from CSV or JSON, with validation reporting which rows were rejected and why.
Grounded llms.txt generator
The llms.txt generator reads your sitemap, grounds the generated file in your pages, validates the result, and can emit llms-full.txt.
Citations per market
The citations view gained a market selector, so citation sources can be examined for one region instead of mixed together.
Navbar, onboarding and dashboard shell
The navbar menu was rebuilt as a viewport-aligned surface with featured rails and live filtering, onboarding shows the engines you can choose from in a preview carousel, and the dashboard shell gained refined domain context controls, a responsive layout, and no white flash on scroll.
Release v0.0.15
AI assistant
A chat panel in the dashboard answers questions about your workspace in plain language: how your visibility is trending, what the engines said in the latest scan, which prompts you track, how your jobs are running. It reads workspace data through a bounded set of tools (listing projects, summarising scans, pulling trends, checking account status) rather than answering from memory, and replies stream in as they are written. Each plan includes a monthly allowance of assistant turns.
Every headline number has history
Mention rate, share of voice, average rank, citation counts and brand-safety signals now have trends. The series are computed from the frozen per-scan rollup (the numbers recorded at scan time), so a chart cannot disagree with the headline card and neither changes after the fact.
Share of voice per engine
The competitor view breaks share of voice down per engine, so you can see where a rival leads on ChatGPT but trails on Perplexity.
Logo mark animation
The logo mark draws itself in on the navbar and landing hero, and carries through sign-in and onboarding. It respects reduced-motion settings.
Release v0.0.14
Site-wide search, two ways
A new search covers the whole public site: knowledge base, glossary, tutorials, blog, changelog, features, solutions and free tools. It opens as a ⌘K dialog from the navbar on any page, and lives at its own /search page for shareable result links. Multiple words match together, so "citation checker" narrows to what it names instead of returning half the library.
Thirty new articles
The knowledge hub grew by thirty articles: product and educational posts plus ten audience-specific pieces closing the solutions coverage gap, with editorial cross-links threading the existing posts into the new ones. Every article carries its own brand cover image.
Tutorials that show the dashboard
Tutorials embed the same dashboard UI they describe (stat cards, charts and tables fed by sample data) instead of screenshots. A new tutorial walks through the Activity page.
One credit, one prompt
> A note on the trial. This entry was written when the product had no > trial at all. Today every new workspace starts on a 14-day free Business > trial (no card) and moves to the free plan when it ends. See the > pricing page. The sentence below reflects what shipped then.
Free allowance that survives a deleted project
The free weekly prompt allowance is counted on the workspace, not on scan history, so deleting a project does not reset it.
Release v0.0.13
Status page runs independently
status.llmmetrix.com is hosted by an external monitoring provider, independently of the application it reports on, so it answers whether or not the product itself is reachable. The in-app /status route redirects there.
Live health in the footer and dashboard
Current system state shows as a live badge in the site footer and, when something is degraded, as a banner on the dashboard, so you find out from the product instead of wondering why a scan is slow.
Release v0.0.12
Google Search Console, on every plan
A new Search page shows clicks, impressions, click-through rate and average position for your verified property, with your top queries. The tracking Google search performance tutorial walks through connecting and reading it. Connecting backfills up to sixteen months of daily history immediately, plus query-level detail for the last quarter, so the page has something to show on day one. It is available on every plan, including Free. Average position is weighted by impressions, and because Google withholds low-volume queries entirely, the share of your traffic the query table can account for is reported rather than hidden.
Ranked but not cited, cited but not ranking
Your Search Console queries are matched to your scanned prompts, and the disagreements are surfaced two ways: pages that sit on Google's first page while no AI engine mentions you, and prompts where engines mention you but Google does not rank you. Each row shows the query and prompt it matched on, so you can judge the match. Queries with no matching prompt are counted as coverage, not asserted as findings: nobody asked the engines about them.
Page-level citation comparison
A sharper version compares the pages Google ranks against the pages engines actually linked to. It works only on search-grounded scans: an ungrounded answer yields a domain extracted from prose rather than a link, so when no citation carries a URL the comparison reports that it needs grounding rather than marking every ranking page as uncited.
Pick your Search Console property
Settings lists the properties your connected Google account can access, filtered to the ones that will return data, so binding one is a choice from a list.
Annual billing (staged rollout)
> A note on the prices. This entry describes annual billing as it shipped > then, at the list prices of the time. Today's figures are on the > pricing page. The 15% annual discount is unchanged.
Release v0.0.11
Transfer ownership
An owner can promote another member to owner from the members panel, which is what makes it possible to leave a shared workspace, close an account, or hand a client account over without stranding it.
Project-scoped keys
An API key can be scoped to specific projects rather than everything you can reach, validated at creation against the workspaces you belong to. The key screen shows the secret once, in a dedicated banner, then lists how many projects each key is scoped to and when it was last used, so a key nobody is calling is easy to spot and revoke.
Accessible print reports
The charts on the printable report carry the same captions and hidden data tables as the rest of the dashboard, so a printed or shared report is readable without the graphics.
Release v0.0.10
Structured data validator
The ninth free tool checks a page's JSON-LD: what types it declares, which required and recommended properties are missing for the common ones (Organization, Product, FAQPage, Article and BreadcrumbList), and which errors would stop a search or answer engine using it. The structural checks run locally, and an external validation service is consulted only when one is configured. Try it on any page at the schema validator.
Release v0.0.9
A status page backed by measurements
The status page reports each component from live checks (the database, scan and notification job outcomes, overdue work, and scheduled-job heartbeats) rather than a static list. Anything that cannot be measured is not listed at all.
Charts everyone can read
Every chart carries a caption and a visually-hidden data table, so screen-reader users get the figures rather than an unlabelled graphic. Muted text and outline buttons were re-tuned at the same time to clear the WCAG AA contrast ratio on every surface.
Polite crawling
Every page we fetch on your behalf (the site audit, the content grader, the domain-verification check) reads the target's robots.txt first, honours the most specific rule that applies to us, re-checks after every redirect, and waits out a declared crawl delay between fetches, up to a two-second cap. If robots.txt cannot be reached, we treat that as "do not crawl" rather than as permission. The AI crawlers guide covers how crawlers reach your site and how to welcome them.
Release v0.0.8
Delete and restore a workspace
An owner can delete a workspace. Deleting marks it Deleting in your workspace list and blocks all work in it, revokes every public share link immediately, cancels its queued work, and schedules the content for removal after a 30-day window. It stays in the list while that window runs: hiding it would make a mistaken deletion look instantaneous and irreversible. An owner can restore it at any point, and members keep their access throughout, so a deletion started by mistake costs nothing. Billing and invoice records are deliberately kept beyond the window, because "delete my workspace" should never mean "delete the record that I paid you".
Verify domain ownership
Prove you control a domain with a meta tag on its home page, or with a DNS TXT record. The verifying your domain tutorial walks through both. Verification is not a gate on scanning: tracking a competitor's domain is a normal thing to do, and observing public data about a domain is not a claim on it. It gates only the places where an unproven claim would cost someone else: uploading a domain's server access logs, and taking a white-label custom domain, which is unique across all customers.
Your first scan, during signup
The last step of signup creates your project and runs a scan against it, reporting each stage of the job as it happens, so you arrive at the dashboard with a result rather than an empty state. If the scan cannot complete, signup still succeeds and tells you why: your project is saved either way, and the scan can be run again from the dashboard.
Release v0.0.7
Track your own prompts
Rather than relying only on automatically discovered questions, you can maintain a list of prompts per project, and scans draw from that list. The setting up tracked prompts tutorial walks through it. Each scan runs up to your plan's per-scan prompt limit, so order the list by what matters most. Suggested prompts come with topic clustering and, when a keyword-data provider is configured, estimated search volume.
Site-wide content audit
Point the audit at your domain and it discovers and crawls your pages, scores each one for how well it answers the questions AI engines are asked, and aggregates the result into the Content Gaps view.
Search-grounded answers
Engines can be queried with their provider's live web search enabled, so an answer comes back with the URLs it actually cited rather than domains inferred from the prose. There is no switch for this in the dashboard: it is off by default, and where it is enabled a grounded prompt carries a credit multiplier, because a grounded query costs more to run. Perplexity and Google AI Overviews are the exceptions in both directions: they search on every call, so their result is grounded already and carries no surcharge.
Google AI Overviews
Scans can include Google AI Overviews, the answer block above Google's results. It works differently from the other six: rather than querying a chat model it reads a search-data provider, so it is included when one is configured and skipped from a mixed scan when none is. That makes seven surfaces in total.
Two more free tools
The free tool set gains an llms.txt audit and a schema markup generator.
Release v0.0.6
AI traffic analytics
A new Traffic page shows which AI crawlers fetched your pages and which AI assistants referred real visitors, rolled up per day. The monitoring AI traffic and crawlers tutorial has the full walkthrough. There are two ways to feed it: a one-line beacon script for referrals, and an upload of your server access logs for crawler hits. Every crawler hit is checked against the operator's published verification method (reverse DNS or a published address range) and marked verified or unverified, so a user agent that merely *claims* to be an AI crawler is shown as unproven rather than taken at its word. Several operators publish no verification method, so their hits stay unverified permanently; verification never blocks ingestion, so nothing is silently dropped.
Tag the links you place
A UTM link builder inside the dashboard tags the links you place in AI-visible content, so the visits they generate are attributable when they arrive on the Traffic page.
Analytics connectors
Connect Google Analytics 4 and Google Search Console from Settings → Integrations, and Cloudflare through the API. Credentials are encrypted before they are stored, and the connectors sync on a schedule rather than at page load, so a dashboard never waits on a third party.
Release v0.0.5
An Activity page for background work
A new Activity page lists your recent background work (scans, notifications, exports, syncs) with its live status, so you can see what is queued, what is running and what has finished without inferring it from the dashboard. The diagnosing background jobs tutorial covers reading the queue and acting on a stuck job.
Restart a job yourself
A job that has failed, or been given up on after its automatic retries, carries a Retry control. Restarting re-drives the *same* job rather than creating a new one, which is what makes it safe for scans: a restarted scan adopts the result the original already committed rather than charging you a second time.
Expiring export links
Data-export download links carry an expiry, and the exported data is purged from the job record once the link has lapsed, so a link that is forwarded or left sitting in an inbox does not stay live indefinitely.
Release v0.0.4
Your brand on the dashboard
On the Business and Agency plans, a workspace can set its own display name, logo, favicon and accent colour from Settings → Branding. They apply to the logged-in dashboard, so a client you have given a seat to sees your brand from the moment they sign in.
Your brand on client reports and email
The same logo and accent colour carry onto the printable report and shared read-only report links. Report, invite and digest email carries your workspace name and accent colour as a text wordmark rather than an image, so it renders the same in every mail client.
Release v0.0.3
Figures frozen at write time
Each scan's headline figures (mention rate, average rank, share of voice, citation counts) are computed once when the scan is written and stored on the scan itself, so they stay fixed to what that scan found.
Faster score history
The score trend reads those stored figures directly, without pulling the individual answers, citations and competitors behind each scan, so charting months of history stays quick as your history grows. The understanding your visibility score tutorial walks through reading the trend.
Release v0.0.2
Access follows membership
Whether you can see a project, its scans, its reports and its background jobs is decided by one thing: whether you are a member of the workspace that owns it, and what your role is there. Roles run viewer → member → admin → owner, each adding to the one below it. Because access is evaluated against current membership on every request, removing someone from a workspace revokes all of their access immediately. For what that means when a content or SEO team runs a shared account, see working as a team on AEO.
Release v0.0.1
Multi-engine visibility monitoring
Track how ChatGPT, Perplexity, Gemini, Claude, Grok, and Meta AI answer buyer-intent questions about your category, and whether your brand shows up, from a single dashboard. Discovery prompts run automatically, or you can track your own list of queries.
Visibility Score
A single 0–100 score you can watch over time, the visibility score derived from three signals in each answer: how often your brand is mentioned, how prominently it ranks among the named options, and the sentiment of the mention. The score is calculated per engine and as an aggregate.
Answer engine ranking and citation intelligence
See where your brand lands in each answer (first, prominent, mid-list, or fine-print) and which third-party sources engines cite for your queries, so you know where to earn presence.
Competitor benchmarking and GEO recommendations
Compare share-of-voice against the competitors you track, and get optimization actions generated from your own scan results, grounded in the queries and engines where you are weak.
Brand safety monitoring
Surface answers that make questionable or potentially inaccurate claims about your brand, ranked by risk, so you can review and correct the record.
Scheduled scans, alerts and exports
Automated re-scans on your plan's cadence: weekly on every paid plan, with Free scanning on demand. Notifications by email or webhook, per scan or batched into a weekly or monthly digest. Plus CSV / JSON exports and shareable read-only report links.
26 releases so far. Follow along as we ship. The roadmap shows what's next.
Want to shape what we build next?
See what we're building next, and tell us what you need.