Opens in a new tabSkip to content
Agent LighthouseAgent Lighthouse

    Searches the text of every published page. The evidence sources themselves are not in this index — search all of them on the trusted sources page.

    GitHub ↗
    Browse checks and page contents
    answer-readiness/meta-description

    Meta description quality

    What it checks

    <meta name="description"> on ctx.pages[0], against the properties Google documents for it — presence, a usable length, prose rather than a keyword run, and relevance to this page.

    State Result
    no description tag fail, priority high
    present, length outside 50–300 warn, priority high
    present, in range, but a keyword string (5+ comma/pipe-separated fragments of ≤ 4 words, no sentence) warn, priority medium
    present, in range, prose, but sharing no content term with the page’s <title>/<h1> warn, priority medium
    present, in range, prose, on-topic pass

    The two quality states are the absorbed half (9.11). Neither can produce a fail: a description that exists is never worse than one that does not.

    Why it matters

    A search engine or answer engine that indexes the page reads <meta name="description"> and may reproduce its text verbatim as the page’s summary/snippet instead of generating one from body text.

    Evidence

    • Google Search Central states the tag is a snippet source: “Google sometimes uses the meta description HTML element if it might give users a more accurate description of the page than content taken directly from the page.” — developers.google.com/…/snippet (verified 2026-08-21)
    • description is one of the meta tags Google Search explicitly supports (the same page notes “Google will ignore meta tags that it doesn’t support”) — developers.google.com/…/special-tags (verified 2026-08-21)
    • WHATWG HTML defines description as a standard metadata name whose value “must be appropriate for use in a directory of pages, e.g. in a search engine” — html.spec.whatwg.org/…/semantics.html (verified 2026-08-21)
    • Google’s AI-features guidance confirms AI Overviews/AI Mode are built on Search and are governed by the same snippet controls (nosnippet, max-snippet, data-nosnippet), which is the only documented link between snippet text and AI answers — developers.google.com/…/ai-features (verified 2026-08-21)

    Limits

    No AI vendor documents reading this tag. OpenAI’s crawler documentation (GPTBot, OAI-SearchBot, ChatGPT-User, OAI-AdsBot) describes only robots.txt and user-agent behavior and mentions no HTML metadata at all — developers.openai.com/…/bots (verified 2026-08-21). Google’s AI-features page likewise never mentions meta description — developers.google.com/…/ai-features (verified 2026-08-21). The audit’s 50-300 character gate is contradicted by the primary source. “There’s no limit on how long a meta description can be, but the snippet is truncated in Google Search results as needed, typically to fit the device width” — developers.google.com/…/snippet (verified 2026-08-21).

    The claim that the tag is the “primary summary” an agent uses when generating answers is unsupported: ChatGPT-User, ClaudeBot and PerplexityBot fetch and read full page bodies, so the tag is at best one candidate among many.

    How it scores

    Google documents that Search reads the tag and may use it as the snippet, and the tag is a WHATWG-standard metadata name with near-universal adoption. But the consumption is explicitly conditional — “sometimes” — no AI-agent vendor documents reading it, and the audit’s 50-300 character window has no primary support.

    Sources