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/descriptive-urls

    Descriptive URL slugs

    What it checks

    AI engines use URL text as a content signal. Descriptive slugs help agents understand page topics before fetching the content.

    Why it matters

    URLs containing readable, hyphen-separated words describing the page’s topic increase the likelihood of the page being retrieved and cited by AI answer engines, relative to opaque ID-based URLs.

    Evidence

    Descriptive URLs

    Google’s URL structure doc gives clear, quotable guidance. First: ‘When possible, use readable words rather than long ID numbers in your URLs.’ Second: ‘Use words in your audience’s language in the URL (and, if applicable, transliterated words).’ Third: ‘We recommend using hyphens (-) instead of underscores (_) to separate words in your URLs.’ Semrush analysed 5 million cited URLs across ChatGPT Search and Google AI Mode. Citation counts peak for 21–25 character slugs (~87K citations), with a secondary peak at 6–10 characters.

    Moderate slug lengths of 17–40 characters consistently outperform very short and very long URLs. Mechanistically a descriptive slug does carry topical tokens that a retriever can match and that a synthesiser can render as meaningful anchor text, so the convention is coherent and costless to follow.

    Limits

    Google’s own doc frames descriptive URLs purely as crawlability and human/machine comprehension, and makes no ranking claim whatsoever — it says only that they help ‘Google Search (and your users) better understand your site’. The Semrush slug data is a distribution over already-cited URLs, with no uncited control group. It therefore cannot separate a URL effect from the confound that well-edited sites both write good slugs and produce citable content.

    Semrush labels the whole study correlational. No AI vendor documents URL wording as an input to source selection — OpenAI’s and Anthropic’s publisher-facing docs are silent, and Anthropic’s crawler doc contains no content-selection guidance at all. The GEO paper did not test URLs among its nine methods, and the 2026 critical survey does not list URL structure among replicated levers. Google’s AI-features doc reiterates there are no special optimizations for AI surfaces. Plausible, conventional, cheap — and entirely unproven as a citation driver.

    Sources