5W study warns AI is fabricating celebrity deals and reputation narratives

AI search is becoming a brand-risk engine of its own. 5W’s new endorsement and reputation research showed that answer systems are not just missing details, they are assembling confident commercial stories that never happened, from fake celebrity partnerships to reputation narratives anchored on a single damaging event.

The 5W Celebrity Endorsement Index, published May 22, 2026 and co-published with Talent Resources, covered 50 celebrities across 4 AI engines and found that 18% of AI-recommended celebrity-brand partnerships were fabricated. Nearly half of those fabrications, 47%, were expired deals presented as if they were still active. In the same study, 64% of surveyed brand marketers said they now begin endorsement scoping inside an AI engine before contacting talent representation, which turns hallucination into a front-end business problem rather than a downstream cleanup issue.

Selena Gomez ranked No. 1 in the index, followed by Ryan Reynolds at No. 2 and MrBeast at No. 3. 5W said the biggest names tended to be owner-operators or long-running brand vehicles, the kind of figures whose business footprints are wide enough for machines to find repeated signals. The company said the endorsement work drew on 600-plus prompts, and that 24% of Gemini’s endorsement picks hallucinated most, with engines asserting partnerships, ambassador roles and product lines that did not exist in the public record.

The companion 5W Reputation Index widened the warning. Its Finance Phase said AI engines were anchoring 53-point composite gaps to single named events, while narrative density, including books, foundations and annual letters, predicted stronger rendering than assets under management or raw capital scale. The broader index was framed as a directional baseline across five AI engines and five dimensions, Accuracy, Sentiment, Completeness, Consistency and Control, spanning more than 220 named individuals across nine sequential studies in the May-July 2026 series.

The June 5 hedge fund principals study scored 20 principals and highlighted how narrative density affects rendering. James Simons benefited from book-based authority signals, while Bill Hwang lagged in AI rendering. The Archegos collapse remained the starkest example of an anchor event that still compresses one principal’s portrait across engines. The U.S. Securities and Exchange Commission said Archegos grew from about $1.5 billion in value with $10 billion of exposure in March 2020 to more than $36 billion in value with $160 billion of exposure at its March 2021 peak, and that the collapse caused billions of dollars in credit losses among counterparties. Reuters reported in March 2026 that the SEC had made substantial progress toward settling civil litigation tied to the case.

That matters because the visibility fight is now about citation ecology as much as rankings. 5W’s AI Platform Citation Source Index 2026 said it synthesized more than 680 million citations, with Reddit taking roughly 40% and the top 15 domains absorbing 68% of the answer pipeline. Stanford HAI’s 2026 AI Index said organizational adoption reached 88%, underscoring how quickly these distorted narratives can spread. For PR, legal and SEO teams, the new watch list is clear: stale endorsements, anchor events, source concentration and the exact phrases AI engines are using to explain a brand before a human ever gets a chance to correct them.