← The AI Hype Audit — all 181 verdicts
PARTLY “Building AI copies of everyone on Earth” — the funding is real; the eight billion twins are a mission statement
The claimSimile creates “agentic twins” — AI stand-ins for real people — so companies can test ads and products on simulated humans first. CVS Health ran 400,000 twins for medication adherence; Deloitte and Wealthfront are clients. Just raised $200M at a $2B valuation led by Greenoaks, five months after its last round. Founder Joon Sung Park (Stanford “Smallville”) wants to eventually simulate all 8 billion people on Earth.
The business facts all check out against real reporting: Simile raised a $200M+ Series B at a $2B post-money valuation co-led by Greenoaks and Index, five months after a $100M Series A (TechCrunch, Jul 30 2026). Joon Sung Park genuinely is the Stanford researcher behind the “Smallville” generative-agents experiment; co-founders include Stanford's Michael Bernstein and Percy Liang. CVS Health did use ~400,000 agentic twins for medication-adherence research, and Deloitte, Wealthfront, Gallup, Suntory, Telstra and Banco Itaú are named customers. For TikTok, this is unusually well-sourced. The oversell is the hook: “building AI copies of everyone on Earth” is the founder's stated eventual mission, not a product. What ships today is synthetic survey respondents — LLM personas calibrated against panels of real people — and the core scientific question of whether AI twins actually predict what real humans will do and buy is unresolved. Market-research professionals actively dispute synthetic-respondent validity, and a $2B valuation prices the promise, not the proof.
What holds up
- $200M Series B at $2B valuation, co-led by Greenoaks and Index, 5 months after a $100M Series A — confirmed by TechCrunch (Jul 30, 2026)
- CVS Health's 400,000-twin medication-adherence study is real; Gallup, Wealthfront, Deloitte are named customers and CVS Health Ventures joined the round
- Park really led Stanford's Smallville generative-agents study and the 1,000-person simulation paper
What doesn't
- “Copies of everyone on Earth” is roadmap language — the shipping product is synthetic survey respondents, not digital twins of you
- Whether AI twins reliably predict real human behavior is scientifically unresolved, and market researchers actively dispute synthetic-respondent validity
- “Test ads on simulated humans before ever putting them in front of a real audience” works only as a pre-screen — no serious brand has replaced real-world testing with it
The catch
The $2B is real money paid for an unresolved question — whether a language model wearing your demographic profile actually buys what you'd buy. The mission statement made the headline; the open question didn't.
How to actually do it
- Read claims like this in two columns: verifiable business facts (raise, customers, founders) vs. product claims (what the AI can actually predict) — this post aces column one only
- If you're pitched synthetic-respondent research, ask for validation data: how did twin predictions compare to real-panel results on YOUR category?
- Use synthetic panels the way the honest customers do — as a cheap pre-screen before real testing, never instead of it
Every dollar figure verifies; the eight billion imaginary friends are the founder's dream, faithfully reported as if it were the product.
- Confidence
- Medium
- Posted by
- a startup-news clips account funneling to a daily newsletter — unusually well-sourced this time
We test hype for free. We build the real thing for a living.
Thirty minutes, no pitch — and you'll leave with something useful either way.
Book a call with Todd or start with the free Business Checkup →The Verdict Weekly
Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.
© Schreier Group · schreiergroup.com · See a wild AI claim? Drop it here and we'll test it.