Track it right
How to actually measure AI search (without faking it)
What a website scan can really tell you. And what you need to do separately if you want to know how AI talks about your brand.
Updated 2026-04-25 · 6 min read
Website scans measure inputs, not guaranteed recommendations
A scan can confirm whether crawlers can access the page, whether the content is parseable, whether schema is valid, and whether the site exposes enough proof. It cannot guarantee that an AI system will cite or recommend the business.
- Treat scores as readiness indicators.
- Separate direct findings from rollout recommendations.
- Use rescans to confirm that website inputs improved.
Robots, sitemap, and AI crawler rules are direct checks
Robots.txt and sitemap.xml are foundational. AI crawler access is worth checking because some answer engines use their own crawlers. llms.txt can be useful guidance, but it is optional and not required for Google AI features.
- Allow public pages that should be discoverable.
- Reference the canonical sitemap.
- Review AI crawler blocks intentionally instead of inheriting accidental deny rules.
Citation measurement is a separate workstream
To measure actual answer-engine visibility, track a set of real prompts over time, record whether the brand appears, which sources are cited, and whether those visits convert.
- Use long-tail prompts from sales, support, reviews, and customer language.
- Track share of voice across multiple runs, not one answer.
- Pair prompt tracking with analytics for AI referral traffic and conversion quality.
Check your site
See what's hurting your score in 30 seconds.
Run the free check to see if ChatGPT and Google can read and trust your site. No email needed.
Check My Site Free