AI Won't Cast Your Influencer — The AI Casting Index 2026

We expected AI to love influencers. It left almost every one of them off the list.

A brand planning a campaign used to start with an agency. A growing number now start with a prompt — they open ChatGPT and ask "who should be the face of our campaign." The answer comes back ranked, in seconds, and it shapes the shortlist before anyone is briefed or a rate is negotiated.

So Everything-PR and Talent Resources made the AI engines do the job for real. 75 casting prompts. Five engines — ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews. Ten passes. We scored every name. The ranking is below.

Here is the part nobody expects: the top of the list is boring. Zendaya. Ryan Reynolds. Serena Williams. Exactly who you'd guess. The story is at the bottom — because you scroll all the way to #25 to find a single influencer, and the biggest creators on Earth never appear at all.

AI doesn't recommend the most famous people. It recommends the most documented ones.

Zendaya tops the list not because she's the biggest name on it — she isn't — but because she's the most legible. A deep encyclopedic record. A decade of consistent press. A long, documented ambassador history — Louis Vuitton, Lancôme, Valentino, Bulgari. Every fact written down somewhere an AI system can read it.

That's the pattern under the whole study. We call it the Structure Premium: AI systems reward talent whose careers sit in durable, machine-readable records. Not charisma — record. It's why athletes overperform their fame, why a handful of versatile names get recommended for categories they've never worked in, and why creators fall off a cliff.

The biggest names on the internet are nearly invisible to it.

Top creators command audiences that dwarf most of the actors in the Top 25. They sell out launches. They own categories. By every metric the creator economy uses, they win.

The casting engines barely see them. The reason is the Structure Premium: a creator's fame is enormous but locked inside the feed — view counts and follower numbers an AI system doesn't read as biography. An actor's fame of the same size is written down. Some of the most-followed creators alive returned zero casting recommendations across all 75 prompts.

One creator beat the penalty: MrBeast, at #20. He's the only creator who built a genuinely documented footprint — years of business-press coverage, a deep encyclopedic record, blue-chip partnerships up to a Super Bowl-scale deal. He did the thing the Structure Premium rewards, and AI casts him where it casts no other creator. The gap isn't about relevance. It's about a record — and a record can be built.

Who AI casts, by category.

The engines don't agree on everything

ChatGPT names the most and is friendliest to creators. Claude is the most conservative. Perplexity is the most current — new deals show up within weeks. Gemini leans hardest on structured data. Google AI Overviews returns the shortest lists. A name strong in one can be missing from another — so casting research run on a single engine isn't casting research.

The shortlist is winnable.

Brands: run the casting prompts yourself, on all five engines, before you write the brief. If your category returns the same five names every time, your "distinctive" face was picked inside a tool your competitor uses too.

Talent: reach is no longer the same asset as castability. A documented record — accurate press, an encyclopedic presence, a current and recorded deal history — is the casting resume now. It can be built on purpose. Most talent aren't building it. That's the opening.

Agencies: a roster needs a structured-presence layer, not just a follower count — because the first round of casting is already run by systems that read records. Talent Resources' celebrity-procurement work sits exactly on that line.

A modeled, directional study of system behavior — not a precise audit, a live leaderboard, or a measure of talent quality. Outputs vary with phrasing and timing. Built to be re-run; movement between editions is itself the finding.