Case 001 · Recruitment
UK payroll recruitment firm
27 of 32 answers recommended it for an everyday hire. 4 of 32 did for its highest-fee search.
The searches with the biggest fees were the ones AI understood least.
Case 002 · Training
US leadership training firm
Named once in 160 answers. Its closest competitor was named 63 times.
One tool said it couldn't verify the firm's prices, next to competitors that had published theirs.
Case 003 · Construction
US commercial contractor
Recommended in 15 of 50 answers, always under its parent company's name. 0 of 50 used its own brand.
All nine pages we checked on its new site were set to stay out of Google.
What the three have in common
- Each firm was established in its field. The problem was evidence. AI couldn't find the proof for the work the firm most wanted.
- AI judged them job by job. The same firm could lead one buyer situation and vanish from the next.
- The fixes were specific. For one firm it was pricing that AI couldn't verify. For another it was a page setting hiding a whole site.
The three side by side
| Case | Study | Headline |
|---|---|---|
| 001 · UK payroll recruitment firm | 32 answers per hiring situation, four AI tools | 27 of 32 for one hire, 4 of 32 for another |
| 002 · US leadership training firm | 160 answers, four AI tools | Named once; closest competitor 63 times |
| 003 · US commercial contractor | 50 answers, five AI tools, two passes | 15 of 50 under the parent's name, 0 under its own |
Why the names are removed
A study shows a firm where it's weak, in detail. We publish the pattern and keep the name private. We name a client only with their written permission.
What a study covers
An AI Discovery Study now covers 10 buyer situations across ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews. See what the study includes or read the method.
Earlier content work
Looking for our content, video and social case studies? They're under earlier work.