AI search can recognise a company without recommending it to prospective customers. A test described by Search Engine Journal contributor Greg Jarboe suggests that measuring brand visibility by asking about the brand itself may miss the more useful question: does it appear when buyers ask who to choose?
Watch Desk analysis
What happened
Jarboe describes Tony Uphoff testing ChatGPT, Gemini, Perplexity and Google AI Overviews with questions prospective customers might ask. Across 59 answers, Uphoff’s newsletter appeared in none of 12 category answers; his firm appeared in one of 43. The engines could identify Uphoff, but rarely recommended the businesses he had built.
The article also describes a separate problem: AI answers reportedly still identified Uphoff as the CEO of a former company, drawing on older biographies and press releases. Jarboe’s account is a snapshot of one person’s prompts, not a broad benchmark, and he notes that AI answers can vary between runs.
Why it matters
For a business, being accurately described is not the same as being considered. If buyers use AI tools to build a shortlist before speaking to suppliers, a brand may look visible in a branded search while being absent at the moment a customer is choosing.
The article’s practical advice is to test the questions customers actually ask, across several answer engines, and track whether the company is recommended and described accurately. It also suggests checking the sources those systems cite, updating pages a company controls, and correcting stale information where possible. The point is to measure a trend, not to treat one week’s answers as a verdict.
Our read
This is a useful distinction for anyone measuring AI-driven discovery: recognition is not recommendation. But one person’s prompt set is a starting point, not a universal scorecard. Run repeat tests, keep the questions consistent, and connect visibility to outcomes such as enquiries or sign-ups. Otherwise, the dashboard may be reporting a flattering fact that never meets a customer.
What to watch
- Whether repeated tests show stable recommendation rates across engines.
- How often AI answers rely on outdated biographies or company information.
- Whether brands can link AI recommendations to actual customer enquiries or sales.
Discussion spark: Should businesses treat AI recommendation rates as a meaningful marketing metric, or are the answers too variable to trust without repeat testing?
Sources and evidence
- Source update (30 October 2026, 19:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.