Question 1

How accurate are your AI personas compared to real humans?

Accepted Answer

Artificial Societies measures accuracy on two dimensions. Opinion distribution accuracy — how closely our synthetic audience responses match real human opinion distributions — reaches 95% of the human self-replication level. Persona internal coherence — how consistently a single persona maintains its belief system across hundreds of questions — reaches 90%. Most competitors measure only distribution accuracy. Achieving higher than 95% distribution accuracy would indicate overfitting on noisy human survey data.

Question 2

Are your simulations backed by research?

Accepted Answer

Yes. We come from a Behavioural Data Science and Computational Social Science background, having worked with leading research labs at Cambridge, Oxford, Stanford, and NYU. We wrote the first ever large-scale societal simulation research paper, presented at the prestigious International Conference on Computational Social Science (IC2S2 2024), and now published at the British Journal of Psychology.

Question 3

What audiences can you simulate?

Accepted Answer

We can simulate any demographic, psychographic, or professional audience. Artificial societies can be constructed from demographic specifications, lists of target organisations and job titles, or your own first-party data including CRM records and existing research. Examples include policymakers, Fortune 500 executives, medical specialists, and representative general populations. Our database includes over 2.5 million real-world persona profiles.

Question 4

How long does it take to get results?

Accepted Answer

Most simulations can be completed within 4 hours but it depends on the size of the audience and the complexity of the questions. What traditionally requires lengthy recruitment, scheduling, and fieldwork can now be accomplished in a day, allowing you to iterate and refine your research in real-time.

Question 5

Can I interview the personas for qualitative insights?

Accepted Answer

Yes. Beyond quantitative survey responses, every persona provides qualitative answers and is available to participate in follow-up interviews. This gives you both statistical confidence and the rich, human texture of open-ended feedback.

Question 6

How do you ensure the personas reflect real human diversity?

Accepted Answer

Each persona is based on a real person and we model not just individual characteristics but also social influence dynamics—how opinions form and spread in real human societies. We compare our results with real world distribution data to ensure representation.

Question 7

What industries do you work with?

Accepted Answer

We work with enterprises across market research, brand strategy, policy analysis, and more. From Big Tech companies to consumer brands, we help organizations that need to understand human behavior at scale.

Subprocessor	Purpose
WorkOS	Authentication & user management
PlanetScale	Database hosting
Cloudflare	API hosting & edge compute delivery
Google Cloud Platform (GCP)	Backend compute processing
PostHog	Application analytics & logging
OpenAI	LLM inference processing (customer-provided input)
Google AI	LLM inference processing (customer-provided input)
Upstash	Caching & message queueing

Subprocessors

Updates to This List

Contact Us