A five-time CEO ran a brutally simple test this month, and the result should reset how marketing teams report AI visibility. Tony Uphoff asked ChatGPT, Gemini, Perplexity and Google AI Overviews the questions his prospective clients actually ask, then counted how often the engines recommended the two businesses he launched in February.
The result: Uphoff on Media showed up in zero of 12 category answers. Uphoff Advisory appeared in one of 43. Yet every engine that answered “Who is Tony Uphoff?” knew his career, roles and history.
“The models know who I am. They almost never recommend what I’ve built,” Uphoff said.
Known is not the same as recommended
The gap matters because the B2B buying sequence has flipped. Uphoff argues that the machine drafts the shortlist and the human validates it. G2’s March 2026 survey of 1,076 software buyers puts numbers behind that: 51% now start research with an AI chatbot more often than Google, up from 29% in 2025. Nearly 69% said a chatbot led them to choose a different vendor than planned, and one in three bought from a vendor they had never heard of before.
For years, discoverability was the scoreboard. Teams ranked, earned the click and called it visibility. In an answer engine, that same scoreboard produces a false positive. Branded queries come back accurate, so leadership thinks the company is covered. But the buyer’s actual category question never includes you.
Stale records and borrowed shortlists
Uphoff’s test surfaced a second problem. Three engines still described him as CEO of Pipeline360. Only one mentioned his current firm. The reason: the models default to what the web has said about you most often, and that usually means longest ago.
Category answers also rely heavily on other people’s lists. The models cite roundups and comparison pages more than the recommended companies’ own sites. In a separate Friction AI study across 14,140 activewear queries, Nike appeared in about 71% of athleisure recommendations while New Balance was almost invisible, despite the two sharing the same Google Knowledge Graph description.
Why CFOs and acquirers will care
AI visibility is becoming a valuation input. Uphoff’s Q3 2026 Value Index adds an AI Disruption Protection score. Among roughly 100 active acquirers surveyed, 91% called AI resilience essential or important and 74% said the same of a direct customer relationship. In his worked example, two identical data businesses were valued at 5.6x versus 4.0x EBITDA—$28 million versus $20 million—purely on whether buyers believed AI could route around them.
That reframes a rented relationship from search. Traffic that arrives from Google and leaves without a login, subscription or first-party record is worth less than search that converts into an owned relationship.
What to do about it
The fix starts with a new metric: recommendation rate. Uphoff’s own 1-in-43 is what a baseline looks like. Here’s how to apply the playbook:
- Run a weekly prompt audit. Write 10 to 15 buyer prompts plus two or three branded checks, and run them across ChatGPT, Gemini, Perplexity and AI Overviews. Report recommendation rate first.
- Fix the record before chasing the shortlist. Pull the sources the engines cite, update bios, LinkedIn and press pages, and request corrections from directories.
- Own the narrow phrase. Publish the clearest explanation of the specific problem you solve, then earn placement on the 10 to 20 roundup pages the engines already cite.
- Tie visibility to money. Connect recommendation rate to signups, demos, subscribers and pipeline, not just sessions.
The teams still leading with rankings and organic sessions will likely spend 2027 explaining a pipeline gap their dashboards never showed. The models are already drafting the buyer’s list—the only question is whether you are measuring whether you’re on it.
Source: Search Engine Journal
