Named, Cited, or Recommended? A 3-Company AI Search Teardown

A site can pass every AEO checklist, score 100 on technical, ship every schema type and a live llms.txt, and still finish seventh in ChatGPT. We ran the audit on three real companies to show exactly why.

This is the annotated board from the session. CorPay does everything the checklists ask and gets cited constantly, but it's the recommendation only 6% of the time. BambooHR is named in 86% of answers in its category and recommended zero times. Linear ships no structured data at all, gets cited once in 46 slots, and takes 38% of every top pick in its field. Named, cited, and recommended are three different things, and they need opposite fixes.

Walk the three teardowns and figure out which position you're in before you go fix the wrong thing.

What's in the download:

• The full annotated board (an Excalidraw file): six panels walking each company's homepage and its content pages, plus a verdict

• The named vs cited vs recommended breakdown for each company, with the real numbers

• The two-position diagnosis (named-not-recommended vs cited-not-recommended) and the opposite fix each one needs

How to open it: it's an .excalidraw file. Go to excalidraw.com and drag the file in, or File then Open. It's free and needs no account.

Built by Will Leatherman and the team at Catalyst, who've built AI-search systems for 150+ B2B companies.