Why AI-Simulated Humans Are on the Rise and Why That’s Not as Unhuman as It Sounds

To the uninitiated, the idea of AI-simulated humans still feels like science fiction.

Until recently, this was a topic that split the research world into opposing camps: those firmly against it, those enthusiastically for it, and a quieter group in the middle trying to work out what actually made sense in practice. Over the past year, that debate has shifted as our understanding of how to use it responsibly has matured and we see adoption rise.

What are AI-simulated humans (sometimes called virtual audiences or synthetic users)? They are AI-built personas that simulate real people in context, including their circumstances, needs, and decision styles. They combine advanced AI with behavioural science and psychology to help us understand how people think, feel, and choose, across different need states. 

If you’re reading this thinking, “Can we trust it?”, “Is it real?” “Does it miss the nuance?” You’re not alone. To make sense of those concerns, it helps to step back and look at how research has always functioned. What makes people nervous is the word “simulated” or “synthetic”, as if human research is somehow pure. It isn’t. What people tell us is shaped by memory, social pressure, context, fatigue, and how they want to come across. Truth in research has always been mediated.

Research has always been a simulation

This part is easy to forget.

Every focus group, every concept test, every conjoint exercise is an act of modelling reality within controlled bounds. None of it is “the market”. It is our best attempt to see clearly enough to make a decision.

AI-simulated humans take that logic to its digital extreme.

They are probabilistic, category-level simulations of human decision-making. They are designed to answer a very specific question:

“Given what we already know about this category, audience, and context - how are people likely to respond, and why?”

Why this matters now: Solving the "Soggy Sandwich" Problem

Anyone who has worked in traditional market research knows the pain of the "soggy sandwich." It's the sad meal you eat at 9 p.m. behind a two-way mirror after listening to focus groups for hours. It’s a vivid metaphor for the practical downsides of conventional research: it's slow, physically exhausting, and logistically heavy. AI simulations are gaining traction now because they solve these very real operational problems.

• Business decisions sped up. By the time many studies land, the moment has already passed. Innovation cycles, comms testing, UX decisions, pricing moves. Marketers and product teams need direction early, not a perfect answer late. 
• Some questions got awkward to ask humans. Early-stage ideas you don’t want leaking. If you show a top-secret prototype to a focus group, you risk it ending up on Reddit within the hour. A simulation doesn't have a Reddit account. Sensitive behaviours people struggle to articulate. Niche audiences that are hard to reach, then hard to re-reach without fatigue. The list goes on.
• And human data got messier. Survey fraud, bots, professional respondents, and low-effort completion. There are safeguards, and they help, but everyone is fighting this battle. More time goes into cleaning and checking, and it still leaves a trust gap. Simulations sidestep a lot of that noise and help teams sense-check thinking without incentives, fatigue, or gaming.

This shift fundamentally changes the nature of research, moving it from “operations-heavy” to “thinking-heavy”. Less time and budget are burned on logistics, freeing up more resources for the critical human tasks of interpretation and strategic thinking.

Where humans still matter most

None of this removes the need for real humans in research. 

AI excels at cognitive simulation, but it can miss the sensory, social, and environmental cues of the real world. It cannot feel the frustration of trying to open a plastic package that requires scissors. It can’t understand the chaotic workaround of using a butter knife as a screwdriver because you can't be bothered to go to the garage.

Simulations are a tool for exploration, stress-testing ideas, and sharpening hypotheses. They are not meant for validation or sizing a market opportunity. You use them to understand the why behind a potential choice, not to predict a final sales number.

That’s the balance. Simulation for qualitative scale and exploration. Humans for depth, meaning, and judgment.

And in a world moving faster by the month, judgment is the part we cannot afford to automate.

By Louise Polders (Head of Innovation at neon octopus)

neon octopus is an insights consultancy working at the intersection of behavioural science and AI, and a first mover in Africa using AI-simulated humans in client research. If you’re curious about where simulation helps, let’s talk.