The engagement opened with a full three-source scan. Not a spot-check. Hundreds of query variants across the twelve category queries the client's customers actually ask, run against ChatGPT, Claude, and Google AI Overviews. Search Console pulled alongside for the same query set. A rendered-markup audit of every retrievable page on the site.
The AI mention rate came back at 60%. Named on some of the twelve, invisible on the rest. Position in Google search averaged 12. Page two territory. The site was not broken. It was illegible. Crawlers reached most pages, and the pages loaded fast, and the copy was written by a real person. That was not the problem.
The problem was three sources agreeing that the same entity signals were missing from every place assistants and search engines look for them.
Source A
AI answers
Source-map capture showed the assistants reaching the site but pulling from a category page missing schema, then supplementing with competitor pages that had it. The client's own copy was cited on 3 of 12 queries. Competitors were named in their place.
AI mention rate · 60%Source B
Search performance
Google Search Console showed the client indexed for every category query but averaging position 12. Click-through rate below 2% on impressions that never converted. Traffic pattern consistent with page-two invisibility.
Avg position · 12Source C
Site structure
Rendered-markup audit found empty JSON-LD schemas across every location page, canonical URLs pointing at a staging domain that had not resolved since 2023, and review widgets loaded inside iframes no crawler could read.
Structural drift · high