Simile's public record supports a $100 million Series A, not the reported $200 million Series B, and that distinction matters when you're judging whether AI-simulated customers are becoming real infrastructure or just receiving AI-era funding treatment.
Simile has a strong story. It just doesn't have a verifiable $200 million Series B in the public record I could find. TechCrunch search results, Index Ventures job posts published in July 2026, venture funding databases, and Simile-linked materials still point to a $100 million raise in February 2026 led by Index Ventures, with Bain Capital Ventures, Hanabi, A*, Andrej Karpathy, Fei-Fei Li, Adam D'Angelo, Guillermo Rauch, and Scott Belsky listed among backers. That is already a large round for a company most people outside AI research had barely heard of. You don't need to inflate it.
The company itself is real. Simile was founded by Joon Sung Park, Michael Bernstein, Percy Liang, and Lainie Yallen, with roots in the Stanford research behind Smallville, the 2023 generative agents project that put 25 AI characters into a simulated town and watched them form memories, make plans, and interact. That work gave Simile a cleaner origin story than most AI startups get. It wasn't a chatbot wrapper looking for a market. It was a research idea looking for one.
The pitch is strong.
Simile builds what it calls AI simulations of society, populated by agents based on real humans. Companies can use those agents to test product changes, messages, store layouts, earnings-call questions, or other decisions before they spend real money putting them in front of real people. According to Index Ventures job postings from July 24, 2026, the company says it's developing a foundation model to predict human behavior at scale and is hiring in San Francisco, Palo Alto, and New York. That tells you the company is still building, not just announcing.
The customers are the real signal #
CVS Health is the name that makes this more than an academic curiosity. Fierce Healthcare reported in March that CVS Health partnered with Simile to use agentic twins in consumer-facing product work, built on 2.9 million consented responses from more than 400,000 people across more than 200 behavioral scenarios. IT Brew separately reported that CVS was using more than 100,000 agentic twins and that Sri Narasimhan, CVS Health's vice president of enterprise customer experience and insights, said compact studies could run in 15 to 30 minutes instead of the four to six weeks a traditional research process might take.
That is a serious operational claim. If you run product, customer experience, or market research, you know why it lands. Recruiting the right people is slow. Follow-up studies cost money. Hard-to-reach groups are hard to reach for a reason. A model that gives you a fast read before you commission the human study is useful, even if it never replaces the human study.
Gallup, Wealthfront, Telstra, and Suntory also appear in public Simile materials and funding databases as customers or partners. Deloitte didn't check out. It doesn't appear in the searches I ran, so it shouldn't be presented as a signed customer - that is how fabricated authority creeps into a piece: one familiar enterprise name too many, no source under it.
The gap matters.
The trust problem has not gone away #
Synthetic research is moving faster than researcher trust. Development Corporate's June analysis of User Interviews data said 97% of researchers surveyed use AI somewhere in their workflow, but only 8% regularly use tools that generate synthetic participants. The same analysis said zero out of 150 respondents reported having no significant concerns. You should take that seriously, because these aren't people allergic to AI. They already use it.
NielsenIQ has also warned, as Research World summarized, that rushed synthetic feedback tools can produce answers that pass a gut check without enough evidence behind them. That is the danger. Bad research you can spot is irritating. Bad research that sounds plausible is expensive.
Simile's advantage is that it's not simply asking a general-purpose language model to pretend to be a customer. Its public materials describe agents grounded in real interviews, consented responses, and behavioral data. CVS's reported dataset gives that claim some weight. Still, grounding isn't the same as proof across every use case. A model that helps test pharmacy refill messaging may not be reliable for pricing, political polling, litigation strategy, or demand forecasting.
That is the test.
The $100 million Series A is enough to show serious investor belief. The customer list is enough to show serious enterprise curiosity. What Simile still has to prove is narrower and harder: that simulated people can be trusted when the answer changes a launch plan, a health intervention, or a large consulting budget. For now, the honest version is simple. Simile is one of the more credible companies in synthetic human simulation, but the public evidence supports a $100 million company milestone, not an unverified $200 million follow-on at a $2 billion valuation.
Also read: Travis Kalanick raises $1.7 billion for Atoms as a16z bets on specialized industrial robots over humanoids; Together AI raises $800 million at an $8.3 billion valuation as enterprises abandon closed AI models; A federal judge says the government's case for banning Anthropic has gotten worse