A useful score should
show its work.
Signal checks the foundations of AI search readiness. Every result includes the evidence behind it, so you can decide what matters for your page.
What happens when you scan
We fetch the URL’s HTML, follow up to two redirects, and request robots.txt, llms.txt, and the first sitemap declared in robots.txt (or /sitemap.xml). We analyze the returned HTML without executing JavaScript. Google PageSpeed Insights supplies mobile Lighthouse lab results when available.
We analyze one page, not an entire website. Sites that block automated access or render everything in JavaScript may yield limited results.
Standard scans accept up to 256 KB of HTML and 4,000 elements. Supporting files are limited to 64 KB each. Pages over these limits are rejected; supporting files we cannot fully retrieve are marked unavailable.
How the score is calculated
Each check has a visible weight: high-priority checks count for 2, medium for 1, and low for 0.5. Crawler checks use 2 for search-oriented agents and 1 for others. A pass earns the full weight, a review earns half, and a failed check earns zero. Unavailable checks are excluded.
The overall score is the rounded, equal-weight average of available category scores. Performance uses Google’s Lighthouse performance score directly. There is no universal industry-standard AEO or GEO score, and this app does not reproduce another tool’s proprietary formula.
Six categories, 53 checks
Headings, short answer passages, lists, tables, Q&A schema, hierarchy, and main-content markup.
JSON-LD syntax, entity and content types, author and date signals, external links, llms.txt, and social summaries.
robots.txt policy for search, assistant, training, and model-use tokens. Allowing training is a publisher choice, not a prerequisite for appearing in search. User-requested fetches may follow different policies.
HTTPS, HTTP status, title, description, canonical, indexing directives, viewport, language, image alternatives, and sitemap discovery.
Visible text depth, paragraph length, attribution, recency, publisher background, contact links, and references.
Largest Contentful Paint, Cumulative Layout Shift, Total Blocking Time, and First Contentful Paint. These are mobile lab measurements; they are not field Core Web Vitals. Total Blocking Time is not INP.
Use the findings in context
These checks detect observable patterns, not expertise, truth, or guaranteed visibility. A short contact page need not have 300 words, a comparison table, or an author. Question and link-label matching is English-oriented. JSON-LD parsing does not validate every schema requirement. An llms.txt file is an experimental convention, not an established requirement.
A robots.txt allowance does not prove an actual bot can fetch your content through a firewall. A blocked training crawler does not establish that your content cannot appear in AI search. Decide access policies intentionally, even if your score remains lower.
Data and retention
Reports store the analyzed URL, findings, small evidence excerpts, and metrics for 30 days. We do not save the full fetched HTML. Anyone with a report link can view it. Local history lives in your browser. Scan quotas retain a hashed network identifier for a two-hour window, with expired records removed in scheduled batches. Do not submit private or signed URLs.