I spent six months on a single question. Ask ChatGPT to describe Sam Altman. Then ask Claude. Then Grok, Gemini, Perplexity. Five answers. Same person. Same week. The answers are not the same. This is not surprising in hindsight. It is the most consequential reputation phenomenon nobody is governing. Hundreds of millions of users are forming opinions about the world’s most consequential technology executives — Altman, Amodei, Musk, Sutskever, Pichai, Hassabis — through software those same executives built, fund, or compete with. Buyers, regulators, journalists, capital allocators. Every one of them is being handed a different version of the same person, depending on which engine they happen to ask. I run a PR firm. We rebranded it 5W AI Communications because the work changed. The work is no longer earning a placement and hoping a reporter writes the right thing. The work is making sure the answer the buyer gets — from ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews — is the answer that reflects what the company and its leadership actually are. So we built the research. Six months. Eight founders. Five engines. A structured prompt set. A verified factual baseline. Five scoring dimensions: Accuracy, Sentiment, Completeness, Consistency, Control. The headline finding: in 74% of cases, the engines produced meaningfully different sentiment framings of the same founder on the same prompt in the same week. In 6 of 8 founders, at least one engine produced a factual error. In 5 of 8, Wikipedia content was paraphrased into at least three engine responses — making it the single most recycled source in the entire corpus. **Three sentences on Wikipedia outrank fifty press releases.** That is not an opinion. That is the data. The flagship piece is [on Unite.ai today](https://www.unite.ai/ai-lab-founder-reputation-gap/). The methodology hub is on [Everything-PR](https://everything-pr.com/). The Israeli-specific extension is on [Olam](https://olam.business/). A few notes that didn’t make it into any of those. **One.** The November 2023 OpenAI board crisis is the cleanest stress test of AI-engine reputation dynamics we have. For 72 hours, the answer to “who leads OpenAI?” depended entirely on which engine you happened to ask. Some users were told Altman had been fired. Others were told he was still CEO. Both, simultaneously. That window closed. The pattern hasn’t. **Two.** Most of what PR has done for the past twenty years — issue a statement, brief a reporter, publish a blog post — does very little to move what a retrieval engine says next. The engines index the web on their own schedule. They amplify what’s already there. The inputs that move output get built before the crisis, not during it. **Three.** The infrastructure is buildable. Wikipedia anchors. Primary-source profiles in tier-1 trade publications. Schema-tagged biographical content on owned domains. Dense entity linking. Quarterly audits. A retrieval-crisis playbook. None of this is exotic. All of it is currently absent from the senior communications functions of most companies whose CEOs are being described billions of times a year by AI. **Four.** The category has a name. AI Communications. The audience is the machine. The full research drops continue over the next 90 days. Next up: an audit of the retrieval anchors themselves — which moved the engine answer, which didn’t, and why. Then the Israeli AI Founder cohort. Then the closed-source coding labs. The founders who audit and shape this in 2026 will define the public record of the AI era for a decade. The ones who don’t will spend that decade explaining what the models got wrong about them. Start product research with AI, not Google. That is where the buyer is. That is where the work is. --- _Ronn Torossian is the founder and chairman of [5W AI Communications](https://www.5wpr.com/), the AI Communications Firm. He is the publisher of Everything-PR and the author of two best-selling editions of For Immediate Release_.