Rule reference

robots.txt disallows paths that contain primary content, or the site unknowingly blocks AI crawler user agents. Blocking is a decision — the problem is doing it without knowing.

Scope

Parses robots.txt directives against a list of key paths and known crawler user agents. It cannot detect server-level blocks (WAF, rate limits).

Basis

The Robots Exclusion Protocol is a documented standard; major AI providers publish their crawler user agents and how to control them.

The main content is only built by client-side JavaScript, so a plain HTTP fetch (no browser) returns an empty shell. Some crawlers render pages, others do not — content that is not in the HTML is at risk.

Scope

Compares text volume and key phrases between the raw HTML response and the rendered page. It flags risk; it cannot guarantee how any specific crawler processes your site.

Basis

Crawler documentation across search engines describes rendering pipelines and their limits; client-only content is a well-known failure mode.

The page is missing a title element, repeats another page’s title, or has a title that does not describe the page content. Search engines and AI systems use titles as a primary label for the page.

Scope

Applies to every indexable HTML page. Does not evaluate whether the wording is persuasive — only presence, uniqueness and rough relevance.

Basis

Title elements are a long-standing, documented indexing signal for search engines and are commonly quoted by AI answers when naming a source.

The page has no meta description, or it is a template string that does not summarize the page. A missing description leaves the summary shown in search results to chance.

Scope

Applies to indexable pages. Presence and non-generic content are checked; actual snippet rendering in search engines is not guaranteed and is out of scope.

Basis

Meta descriptions are a documented way to influence result snippets; they are frequently reused verbatim or in part by answer engines.

The page declares a canonical URL that conflicts with itself (e.g. different protocol/host), or different variants of the same page declare different canonicals. This splits signals and confuses crawlers about which URL to keep.

Scope

Applies to pages that should be indexed under one URL. Parameters, pagination and cross-domain syndication need explicit decisions per case.

Basis

rel=canonical is a documented link element supported by all major search engines.

The page has zero or multiple h1 elements, or heading levels jump around (h2 → h4). Structure signals help both readability and machine parsing of sections.

Scope

Mechanically checkable: counts and level ordering. Whether heading texts are meaningful requires human review.

Basis

Heading structure is documented in HTML semantics and is used by parsers to build document outlines.

The page makes claims (numbers, comparisons, "best", "proven") without sources, data or examples. Answers built from such pages inherit weak grounding, and readers have no reason to trust them.

Scope

Requires human judgment: what counts as sufficient evidence depends on the claim. The check can only surface claims that appear unsupported.

Basis

Answer engines weight corroborated content when selecting sources to cite; supported claims are also more likely to be quoted accurately.

Sections depend on context scattered across the page, or headings do not correspond to questions people actually ask. Answers and snippets are built from fragments — fragments that stand alone get quoted.

Scope

Human judgment over heading wording and section completeness. No automated verdict is possible; a check can only organize the review.

Basis

Answer systems extract passage-level content; writing for standalone passages is a documented content practice.

The page shows no date, or a date that changes on every deploy without content changes. Fake freshness erodes trust with both readers and systems that compare versions.

Scope

Detectable: presence of a date and whether it matches the actual content revision history.

Basis

Visible dates are a common trust signal; search documentation warns against misleading date manipulation.

Across a fixed question set, answers do not mention the brand, or mention competitors instead. This is an observation about samples — it has a denominator and conditions, not a verdict.

Scope

Always reported with: question set, sample size, valid response count, language/region, assistant identity and date. A low rate in one sample does not prove a brand cannot be seen by AI.

Basis

Mention counts over repeated samples are the only honest way to talk about "AI visibility"; single anecdotes are not measurement.

Answers that mention the brand get facts wrong: wrong positioning, outdated pricing, confused product names. Inaccuracy usually traces back to stale or contradictory public sources.

Scope

Human review of sampled answers against current official sources. Reports which fact was wrong and where the correct value is published.

Basis

Answer systems synthesize from public sources; the actionable fix is on the source pages, not on the assistant.

When the brand is mentioned, the answer cites third parties only — or nothing at all. Citations are the measurable link between answers and your site.

Scope

Counted over the same samples as mentions. Whether a specific assistant cites links at all is a platform behavior, not something a site can control.

Basis

Citation presence varies by assistant and answer type; treating it as a site-side failure would misread the measurement.