AI Engines Know Tony Uphoff But Won't Recommend Him: 1 Hit in 43 Answers

A hands-on test across 59 AI answers found that Tony Uphoff's newsletter appeared in zero of 12 category responses and his advisory firm in just one of 43 — while every engine he queried about him by name knew his career in detail. The experiment, run this month against ChatGPT, Gemini, Perplexity and Google AI Overviews, captures the measurement gap Greg Jarboe argues the search industry is getting wrong heading into 2027 planning.

Uphoff, a five-time CEO who ran Business.com and ThomasNet.com and led the carve-out of Pipeline360, published the test in an October series on Uphoff on Media. "The models know who I am. They almost never recommend what I've built," he wrote. Jarboe, who worked with him at Ziff-Davis from 1988 to 1993, frames it as the industry's core problem: teams measure discoverability when buyers are being served by recommendation.

How Did Buyer Behavior Shift?

Survey data backs the urgency. In G2's March 2026 survey of 1,076 B2B software buyers, 51% said they now start research with an AI chatbot more often than with Google, up from 29% in the 2025 report. 69% said a chatbot led them to a different vendor than planned, and one in three bought from a vendor they had never heard of before.

Uphoff's summary: "The machine drafts the shortlist. The human validates it." For two decades, he wrote, "the scoreboard was discoverability" — and most companies carried that scoreboard straight into AI, testing branded queries and mistaking accurate self-descriptions for coverage. He calls that a false positive.

Why Do Answer Engines Describe Outdated Companies?

A second finding flagged stale training sources. All three engines that described Uphoff still called him CEO of Pipeline360; only one mentioned his current firm. The cited sources were org-chart aggregators, executive databases and old press releases. His rule: answer engines default to what the web has said about you most often, "and most often usually means longest ago." That stale record costs twice — the machine uses it to recommend, the human uses it to verify. A Corporate Ink study reported by eMarketer in May found only 29% of marketers actively track AI visibility.

A related lesson came from asking "What is Demand-as-a-Service?" All engines defined it accurately and cited Pipeline360's pages — but never mentioned Uphoff, who helped create the strategy. "The models credit whoever publishes the clearest explanation, not whoever had the idea first," he wrote.

What Builds AI Shortlists?

Citations, not brand authority. On category questions, the models cited roundups — "best AI newsletters" and "best B2B marketing podcasts" lists from Zapier, Content Allies and others — rather than the firms they named. A May 2026 study by Maryanna Franco and João da Silva on Friction AI's platform found the same mechanism across 14,140 activewear queries: Nike appeared in about 71% of athleisure recommendation answers while New Balance, despite sharing the same Google Knowledge Graph description, almost never did.

Jarboe cautions that 19 prompts over one week is a snapshot; one appearance in 43 could be a floor or noise. Uphoff plans to rerun the same prompts in January and publish the numbers either way. Jarboe recommends weekly tracking, since Citation Share of Voice only means something as a trend line.

What Do Acquirers Now Price In?

Uphoff's October 7 post introduced the Q3 2026 Uphoff Value Index with a new AI Disruption Protection score. Its third question: "Do you own the customer relationship, or rent it from search and platforms?" Collingwood's 2026-27 Market Report, surveying about 100 active acquirers, found 91% call AI resilience essential or important and 74% say the same of direct customer relationships. In his worked example, two identical data businesses value at 5.6x EBITDA ($28 million) versus 4.0x ($20 million) depending on whether buyers believe AI can route around them.

What Scoreboards Matter Now?

An October 19 post used Block as a case: $10.36 billion gross profit with 10,205 people in 2025 is roughly $1 million per person; raised 2026 guidance of $12.51 billion with just under 6,000 people works out to about $2.1 million. Benchmarks: the average marketing agency does about $163,000 revenue per employee, the median public software company about $395,000. Uphoff's spectrum runs from honest signals (contract value, churn) to cheap-to-fake ones (form fills). "A lead score isn't a fact about someone's business. It's a fact about how well your content ranks," he wrote.

His October 14 post, citing Darshan Deshmukh's "death by a thousand pilots" and BCG's 2024 finding that 70% of AI-value challenges are people and process (20% technology, 10% algorithms), proposed a three-gate Application Test: state the job without a vendor name, name the owner and baseline, and pick one signal to check in 90 days.

The five-move playbook that follows: track recommendation rate weekly across four engines; fix stale model sources after any rebrand; own the narrowest category phrase and pursue 10–20 roundup pages; and tie AI visibility to signups and pipeline. Watch the January rerun — the teams still reporting rankings first, Jarboe bets, will spend 2027 explaining pipeline gaps their dashboards never showed.