# TrustGrowth > TrustGrowth measures whether AI systems and search engines can reach a website, read it, > cite it, and name it unprompted — then publishes the result as a dated public scorecard. > Built by TechWright at trustgrowth.ai. Free tier available; no account needed for the free tools. TrustGrowth is organised around a five-stage AI visibility funnel. Each stage answers a different question and carries a different epistemic status, and the product deliberately keeps them apart: 1. Reach — an unweighted count of a fixed ten-crawler probe, not a rate. Each crawler gets a robots.txt check and a live per-user-agent homepage fetch. API reachable_count is the number of present crawlers only if none is unknown; otherwise null. Unknown: site or robots unreachable, or live fetch 403, 429, >=500, status 0 or no status. Otherwise robots disallow is absent; allowed and partial verdicts are present. Only robots.txt HTTP 200 and 404 are usable. The UI and API share the same crawler-state rules: any unknown withholds the count; known bot states remain visible. A known HTTP 404 partial response is not unknown. Order completed captures by completion time, then ID. Use the latest completed readiness run with a parseable check date and bots list; older than 35 days is stale. llms.txt is separate: 200 present, 404 absent, otherwise unknown; its body never enters the count. No interval or effective sample size. Crawlers: GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended, CCBot, Bytespider, Amazonbot, Meta-ExternalAgent. 2. Readable — a deterministic 0–100 score, not a page count or the TrustGrowth Score's GEO pillar. One observation is one GEO check on one eligible HTML page. Noindex pages enter no denominator. check_score = 100 × (1 − min(failures, eligible) / eligible). Readable = round(mean of defined group scores); each group is the unweighted mean of its defined check scores. Zero-eligible checks and empty groups are omitted. Groups: structured data (missing_schema_types); authorship/trust (missing_author_markup, missing_publish_date, missing_update_date); answer structure (poor_heading_hierarchy, missing_faq_section, faq_not_marked_up, low_list_usage); citation quality (low_outbound_authority_links). incomplete_schema contributes to no group. GeoCalculatorV3 uses pipeline page evidence or audit-row evidence. Pipeline failures are distinct candidate-check-type/requested-URL pairs in captured geo_readiness manifests, with trailing slashes removed; issue open/fixed status and audit origin do not apply. Audit-row failures are distinct check-type/page-URL pairs originating in the selected audit and open on the reference date; strip whitespace and trailing slashes, excluding older-audit findings. Duplicate pairs count once. Select the newest qualifying complete authoritative pipeline cycle by completion time and ID within the observation bound. Require complete coverage; start <= oldest capture <= newest capture <= completion; positive page and known counts; known == eligible; zero failed/inconclusive counts; current page-checks-v1 contract. Prefer page evidence with completed geo-check-coverage-v1 coverage: nonempty complete HTTP 200–299 measurements for this site's pages within the cycle capture window, excluding GEO not_applicable. The set must match the stored HTML-page count; every manifest needs pass/fail and a candidates array. Denominators use stored eligible_checks, not current-page applicability. If invalid, try that cycle's newest published audit by creation time and ID, with a matching site-owned run, created between cycle start and the bound, updated and published by the bound. Failure does not search an older cycle or fall back to a legacy audit. Only without a qualifying pipeline cycle use the latest audit within the bound, rejecting later publication. Accept no run, no linked cycles, or exactly one linked legacy authoritative cycle belonging to this site and its run, complete by the bound, with valid start before completion and absent/complete coverage. Pipeline cycles cannot re-enter this fallback. Require positive pages_crawled and completed versioned GEO coverage. Missing coverage is not_collected, never 100; missing evidence, no successful pages or no applicable checks is unavailable, never measured zero. Rolling coverage uses the cycle batch. API: latest published audit with a completed run, ordered by audit date, creation time and ID, anchors an end-of-audit-date bound; absent audit gives no_published_audit even with a pipeline cycle. UI: bound is now. Different bounds can select different scores. Only audit-row evidence with explicit as_of refuses historical reconstruction after a later-dated audit; captured page evidence bypasses that refusal. Freshness uses the cycle's oldest capture date, including cycle-backed audit rows, otherwise the audit date. Stale means reference date exceeds that date plus the site's audit freshness window. Do not substitute cycle completion. The API stage date is its anchoring audit date, not the page-evidence freshness date. group_coverage = defined groups / 4; crawl_coverage = pages_crawled / (pages_crawled + fetch_errors_count), 0.5 if errors are missing, 0 if attempted is zero. confidence = group_coverage × crawl_coverage. Partial coverage lowers confidence, not score, and yields maturing with calculator reason partial_geo_check_or_crawl_coverage. The API labels a present score measured and exposes that reason beside it. Confidence is coverage, not an interval or sample size. No n; pages_crawled is not sample size. Uses captured GEO evidence, not Search Console. 3. Retrieved — cited_keyword_count is an unweighted census count of unique own-domain keywords returned by DataForSEO Labs ranked_keywords filtered to ai_overview_reference, limit 200. Keep only that item type, deduplicate keywords case-insensitively, keeping best rank_group; exclude competitor rows. At 200 rows the count is a floor, not a total. API count_is_floor is true for an own-domain floor, false for an explicitly untruncated own-domain count, null for unavailable census or legacy truncation metadata. Competitor truncation does not qualify the own-domain count. Keywords the domain does not rank for are outside this census. No denominator, confidence interval or n. Latest completed retrieval run with both capture date and count is authoritative: a successful zero is measured; missing metadata is not_measured. API reads the recorded count; UI recounts dated own-domain rows. After 35 days UI withholds the headline; API retains the count with stale: true. Sampled AI Mode, ChatGPT and Gemini surface rates are separate, not Retrieved: cited answers / answered rows per surface's latest capture, excluding pending and failed answers, unweighted with Wilson intervals and raw answered n. No category-intent or verified-demand filters; entitlement is shown separately. Demand-weighted citation share of voice is also distinct from the census count. 4. Recalled — are you named unprompted? A genuine sample across Claude, GPT, and Gemini, always reported with a confidence interval, and withheld when the interval is too wide. 5. Impact — did anything move? Correlation against Google Search Console data. Reported as correlation, never as attribution. Three measurement rules govern everything published here, and are worth carrying into any summary of TrustGrowth: a failed measurement is reported as unknown, never as a measured zero; a sampled figure always carries its interval; and the published score comes from a named, versioned, publicly documented formula rather than an opaque number. These definitions do not publish per-site AI visibility figures or raw Search Console demand weights. The keyless Reach check remains available without an account. ## Start here - [Free AI visibility scan](https://trustgrowth.ai/tools/free-scan): composite scan of crawler access, homepage readability, and unprompted AI recall. One URL, no account. - [All free tools](https://trustgrowth.ai/tools): 16 single-check tools plus the composite scan, one live URL each, no account. - [Pricing](https://trustgrowth.ai/pricing): Free, Starter $9/mo, Pro $49/mo, and a waitlisted Growth tier. Credit-metered; scheduled audits are included in every plan. - [FAQ](https://trustgrowth.ai/faq): what is measured, what is not, how the data is sourced, and what the limits are. - [AI visibility methodology](https://trustgrowth.ai/features/ai-visibility): stage formulas, eligibility, weighting, counts, intervals, and measurement limits. - [TrustGrowth Score methodology](https://trustgrowth.ai/methodology): pillars, signals, weights, and the versioned formula behind the TrustGrowth Score. ## Product - [Audit & Scoring](https://trustgrowth.ai/features/audit-scoring): 80 checks against your actual pages, every issue filed with an address, 27 E-E-A-T signals with their evidence, and one score you can watch move. - [AI Visibility](https://trustgrowth.ai/features/ai-visibility): the five stages above, from crawler reach to unprompted recall across Claude, GPT, and Gemini, measured against your own site and its visible content. - [Search Visibility](https://trustgrowth.ai/features/search-visibility): absolute SERP footprint and share of voice inside a set you chose — you plus at most three competitors drawn from real keyword overlap. Not market share. - [Authority](https://trustgrowth.ai/features/authority): referring domains, broken-link reclamation, link gaps, and linkable assets — every figure dated and source-labelled. No opaque authority score. - [Groundcrew](https://trustgrowth.ai/features/groundcrew): open-source Markdown skills that let a coding agent audit a site, fix a verified defect in your repository, and confirm the fix after deploy. - [Research & Strategy](https://trustgrowth.ai/features/research-strategy): every pillar names its evidence and every publishing frequency its confidence; what was set aside stays visible. Nothing executes until you approve it. - [Proof & Trajectory](https://trustgrowth.ai/features/proof): a live public scorecard for your site — Google-verified search data, the score, and open issues shown next to the wins. Publishing is off by default. - [The AI team](https://trustgrowth.ai/ai-team): the agent roles that run the pipeline — auditor, keyword research, strategy, calendar, writer, reviewer — and what each one is allowed to do. - [Growth engine roadmap](https://trustgrowth.ai/growth-engine): what is live, what is built but gated, and what is only planned. Written to be checked, not to sell. ## Free tools Each runs one live fetch against a public URL and returns a result without an account. - [AI Crawler Access Check](https://trustgrowth.ai/tools/ai-crawler-access): inspects robots.txt, llms.txt, and homepage access to see which AI crawlers are allowed, blocked, or partly limited today. - [llms.txt Creator](https://trustgrowth.ai/tools/llms-txt-creator): derives a copyable llms.txt from the live homepage and navigation, with notes from those page titles. No generated prose. - [Sitemap llms.txt](https://trustgrowth.ai/tools/sitemap-llms-txt): builds a copyable llms.txt from a sitemap loc list. Same-host links and path-slug notes only; it does not fetch each page. - [AI Crawler robots.txt Policy Composer](https://trustgrowth.ai/tools/ai-robots-policy): fetches a site's live robots.txt and composes allow/block rules for the major AI crawlers, keeping Sitemap lines and non-AI groups. - [Are You Named by AI?](https://trustgrowth.ai/tools/are-you-named-by-ai): asks Claude, GPT, and Gemini one cold category question and reports whether any of them names the site unprompted, with the likely reason when they do not. - [How AI Sees Your Site](https://trustgrowth.ai/tools/how-ai-sees-your-site): one bounded AI read of the homepage, reporting what its visible content currently communicates about the business and its audience. - [E-E-A-T Quick Check](https://trustgrowth.ai/tools/eeat-quick-check): fetches up to three relevant pages and evaluates visible credibility signals for experience, expertise, authoritativeness, and trust. - [Schema Check](https://trustgrowth.ai/tools/schema-check): validates JSON-LD and structured data on a public page, flags missing fields, and shows which rich-result signals are actually present. - [SERP Preview](https://trustgrowth.ai/tools/serp-preview): fetches title, description, canonical, and social tags to preview what search and social crawlers are most likely to see. - [HTML to Markdown](https://trustgrowth.ai/tools/html-to-markdown): fetches one public URL, strips chrome, and returns LLM-readable markdown of the main content plus a deterministic quote pack. No model. - [Canonical Inspector](https://trustgrowth.ai/tools/canonical-inspector): www, HTTPS, and trailing-slash redirect matrix vs the live canonical tag. - [llms.txt Grader](https://trustgrowth.ai/tools/llms-txt-grader): reads a live llms.txt and reports whether it is a briefing, a stub, missing, unusable HTML, or unreadable. Hygiene, not a prediction of whether AI will cite the site. - [FAQ and HowTo JSON-LD Generator](https://trustgrowth.ai/tools/jsonld-faq-howto): fetches a public page and copies FAQPage or HowTo JSON-LD from visible questions and steps. Optional paste composer stays in the browser. - [sameAs Pack](https://trustgrowth.ai/tools/sameas-pack): fetches a public page and copies Person or Organization JSON-LD whose sameAs URLs were observed on that page. No invented Wikipedia or Crunchbase links. - [Page Speed](https://trustgrowth.ai/tools/page-speed): mobile field CrUX for LCP, INP, and CLS, kept separate from the Lighthouse lab performance score and top opportunities. Missing CrUX is null, never a zero. - [Link Graph](https://trustgrowth.ai/tools/link-graph): crawls up to 200 same-host pages and reports sitemap orphans, click depth from home, the most-linked pages, and pages with no outbound links. ## Evidence Published measurements, not marketing figures. Every proof page belongs to a site whose owner opted in, and shows open issues alongside results. - [Proof pages](https://trustgrowth.ai/proofs): the directory of public, dated scorecards for opted-in sites. Each `/proofs/{slug}` page publishes per-site AI-visibility stage values; `/api/v1/proofs/{slug}` carries the same values under `ai_visibility`. The method is at `/features/ai-visibility`. - [Leaderboard](https://trustgrowth.ai/leaderboard): opted-in sites ranked by their published TrustGrowth Score. Rank is earned by measurement and cannot be bought. - [Directory ledger](https://trustgrowth.ai/proofs/directory-ledger): the record of what was distributed to the public directory, and when. ## For developers and agents - [Developers & API](https://trustgrowth.ai/developers): the public REST API — authentication, endpoints, rate limits, and the credit cost of each call. - [OpenAPI specification](https://trustgrowth.ai/developers/openapi): the machine-readable OpenAPI document for that API, served as YAML. - [XML sitemap](https://trustgrowth.ai/sitemap.xml): every public URL on the site, including proof pages and blog posts. ## Writing - [Blog](https://trustgrowth.ai/blog): essays on measuring AI visibility, scoring methodology, and what the data shows. Written by the team that builds the product. - [Blog RSS feed](https://trustgrowth.ai/blog/feed.xml): the same posts as a feed. - [Open Blog Baseline](https://trustgrowth.ai/open-blog-baseline): one list of editorial and technical rules for a company blog, each with its basis. Our own document, not issued by a standards body, not legal advice. Text licence CC BY 4.0. - [open_blog](https://trustgrowth.ai/open-blog): the open-source Rails blog engine (MIT) that an agent operates through Ruby, a JSON API or MCP tools. ## Company - [About](https://trustgrowth.ai/about): who builds TrustGrowth and why it is built in public. - [Contact](https://trustgrowth.ai/contact): how to reach the team. ## Optional Secondary pages. Safe to skip when a shorter context is needed. - [For founders](https://trustgrowth.ai/for/founders): the same product framed for a founder running their own site. - [For agencies](https://trustgrowth.ai/for/agencies): the same product framed for an agency reporting to clients. - [Terms of service](https://trustgrowth.ai/terms): the terms covering use of the product and the public API. - [Privacy policy](https://trustgrowth.ai/privacy): what is collected, why, and how long it is kept.