Measuring AI Search Visibility as Crawl, Citation and Click
Separate AI search visibility into crawl access, answer citations and referral clicks
Definition
An AI-search measurement model that treats a page being crawled, cited in an answer and clicked by a person as separate events rather than one visibility metric.
Perspectives
harsehaj (2026-08-24, X)
Track three layers independently: whether an answer engine crawls the page, whether it cites the page, and whether a user clicks through. Parse citation exports back to post slugs, use referrer domains for ChatGPT and Claude clicks, and read Google AI Overview traffic inside ordinary Google Search Console performance.
Browserbase's strongest Google search posts were cited only twice by AI, so do not substitute organic rank for citation measurement. In the implementation, SEMrush citation data required a manual CSV export and crawl tracking was deferred because Vercel log drains sat behind SSO.
Lee Jae-cheol (2026-08-20, YouTube)
Track each prompt and engine separately across ChatGPT, Gemini, Perplexity and Claude. For every answer, distinguish a brand mention, use of the domain as a source, a linked citation and the brand's recommendation position against competitors; engine-level exposure can differ enough to change which outside publication or owned channel deserves attention.
Monitor organic top-10 coverage alongside AI Overview inclusion, but do not collapse them into one result. Lee states that roughly 60% of AI Overview sources come from Google's top ten and treats top-ten rank as an intermediate target; the interview does not identify the sample or study behind that figure.
Borja (2026-08-27, X)
Pick one answer engine and one backup from customer interviews and countable referrals such as utm_source=chatgpt.com or Perplexity before working on citations. Track every surface and buyer-question shape separately, then rescan monthly rather than reporting one pooled “AI visibility” score.
Borja's snapshot covered 100 buy-intent questions across ChatGPT, Gemini, Perplexity, Google AI Overview and Google AI Mode. It produced 500 answers with 4,231 citations to 1,318 unique sources; 85 of 98 questions answered by all five had no source shared across every surface, and median ChatGPT–Gemini overlap was 0.000. Perplexity cited a median 17 sources per answer while ChatGPT and Gemini cited 4.
How to apply
- Fits a site already investing in Building Product Information Surfaces for AI Search and needing to tell discovery, quotation and traffic apart.
- Map every citation row and referral to a stable post slug before aggregating domain-wide trends.
- Keep crawl coverage marked unknown when server logs are unavailable; a citation absence cannot distinguish no crawl from no selection.
- Read alongside Measuring SEO Content Changes Against a Baseline because normal search and AI visibility may move differently after the same edit.
Limits
- The source's citation feed was export-only, crawl measurement was unfinished and AI Overview clicks were not separately identifiable in Search Console.
- Engine interfaces, referrer behavior and citation exports can change without notice.
- Citation counts do not measure answer prominence, sentiment, factual accuracy or downstream conversion.