How this analysis works

Coverage follows the mode selected before the audit. Every mode is bounded; Growthract does not claim to crawl every page on a site.

• Quick scan: Representative scan — homepage plus up to 4 goal-relevant internal pages. Scores homepage links for goal relevance, then selects a diverse representative set.
• Expanded site scan: Up to 25 discovered pages. Uses sitemap discovery first, then a bounded same-site crawler when no usable sitemap pages are available.
• Specific pages: Selected-page scan — up to 25 exact same-site URLs supplied by you. Reads only the valid same-site URLs supplied; it does not discover additional pages.

Only publicly accessible HTML is read — content that requires login, or that's injected client-side after the initial page load, may not be seen. Pages that fail to load are skipped and listed as warnings in your results, not silently dropped.

These are plain rule-based checks against your page HTML — no AI involved, and fully reproducible:

• Title tag presence and length (30–60 characters is a practical heuristic, not a hard Google limit)
• Meta description presence and length (90–165 characters, same caveat)
• H1 presence, canonical tag presence, HTML lang attribute
• Noindex directives (meta robots and X-Robots-Tag header)
• Image alt-text coverage
• Structured data (JSON-LD) validity
• Duplicate titles/descriptions across the pages scanned
• robots.txt and XML sitemap accessibility, and common misconfigurations (a sitewide disallow, invalid or cross-domain sitemap URLs, duplicates)

Everything else is generated by an AI model reading your scanned pages, and should be read as a considered judgment call — not a measured score:

• Business context: industry, audience, personas and primary offer, inferred from your page content. You can correct this directly with "Edit context" before re-running.
• Goal routing: which of the checks above and which strategies below are actually relevant to what you asked for — this tool intentionally skips checks that don't apply to your specific goal instead of running a fixed checklist every time.
• Keyword strategy: recommended search topics, each checked against your scanned pages and marked Found/Missing based on real text matching.
• AEO (answer-engine) opportunities: a qualitative read of how well your content addresses specific answer-style queries. This is not a measured citation rate from ChatGPT, Perplexity, Google AI Overviews or any other live system — Growthract doesn't query those systems.
• Prioritized actions: Impact/Effort/Priority ratings are the model's judgment, grounded in the evidence shown for each action, not a fixed scoring formula.

If you connect a Google account with Search Console and/or GA4 access (Settings → Search Console & GA4), the last 28 days of real data is pulled in during analysis:

• Search Console: total clicks, impressions and average position for the connected property, plus your top ~25 actual search queries with their real clicks/impressions/position.
• GA4: total sessions, conversions and users, plus your top landing pages by sessions.

This is real measured data, not AI judgment. It's used to mark recommended keywords with their real clicks/impressions/position when a close match exists in your actual queries, and to ground the keyword, AEO and prioritized-action sections in real numbers instead of only page-content inference. It does not change the answer-engine disclaimer in Section 03 — Search Console and GA4 measure Google organic search and site traffic, not AI answer-engine citations. Without a Google connection, private Google performance data is unavailable; page-content analysis still works and public Common Crawl evidence is attempted independently.

Bing Webmaster: When connected, Growthract can use verified Bing search and index evidence such as clicks, impressions, queries, page visibility, crawl issues and inbound-link counts. This describes Bing organic search and indexing; it does not measure whether a brand appears in ChatGPT, Gemini, Claude, Perplexity or other answer engines.

Common Crawl: Growthract automatically checks the latest three available Common Crawl indexes for corpus presence on the audited domain. No account connection is required. This is supporting crawl and discoverability evidence: it can show whether qualifying HTML captures appeared consistently across those sampled crawl releases, recently appeared, recently disappeared from the sample, or were not found in the sample.

Common Crawl evidence does not prove that a page was used to train an AI model, and it does not prove that an AI answer engine cites or recommends the site. Likewise, absence from the sampled indexes does not prove that the page is absent from the live web. Growthract uses this evidence only as supporting corpus and crawl-discoverability context.

From an action's fix flow, clicking "Turn this into a deliverable" makes one additional, targeted AI request that writes the specific content implied by that action (a rewritten title tag, a JSON-LD block, ad copy, a content outline, etc.), grounded in that action's own evidence. It does not re-scan your site.

• Without a connected Google account (Section 04), no private Google traffic or ranking data is used. Growthract may still use public Common Crawl corpus evidence alongside the current page scan. Connecting Google adds the last 28 days returned by its APIs, not full history.
• Search Console / GA4 data covers Google organic search and the connected GA4 property's traffic only — not other search engines, paid traffic, or AI answer-engine citations.
• Quick is representative, Expanded is capped at 25 discovered pages, and Specific is limited to submitted same-site URLs. None is an exhaustive whole-site crawl.
• AI-generated sections can be wrong or miss context your page content doesn't make explicit.
• Running, viewing, and downloading the immediate audit as a PDF do not require an account. Saving it for later, report history, and connecting search data require sign-in. See our Privacy Policy for full data handling details.
• Public-web presence is not proof of AI visibility. Common Crawl, Wikidata, and Wayback provide supporting evidence only and cannot prove model training, citation, recommendation, ranking, or a business outcome.
• Change and verification results require a comparable scan of the same site and audit mode. Verification proves only that its stated acceptance check passed; it does not prove a causal business outcome.
• Trend charts and competitor comparisons (see below) track our AI-judgment metrics (and, when connected, real Search Console / GA4 metrics) over time — the AI-judgment portion is not a fixed, precise score, and the same analysis run twice on an unchanged page can vary slightly.

What this tool measures today: deterministic technical checks (Section 02) plus one AI-generated read of your current content (Section 03), optionally grounded in real Search Console / GA4 and Bing data when connected, plus best-effort public Common Crawl corpus evidence (Section 04). These evidence sources support the analysis but do not constitute AI citation measurement or a full historical tracking system.

What's now available: if you're signed in, every analysis is saved, so re-running on the same site shows a trend of both the AI-judgment metrics (technical findings, AEO coverage, keyword coverage) and, once you connect Search Console / GA4, real clicks, impressions, average position and sessions over time — plus a comparison against a competitor site you choose to track (AI-judgment metrics only; we can't see a competitor's private Search Console or GA4 data).

What's not built yet: live querying of ChatGPT, Perplexity, Gemini-with-search, or Google AI Overviews to check whether they actually cite you. That requires paid, metered API access to each of those systems (none of them offer this for free), so it's a deliberate roadmap item, not something this tool silently approximates today.