Brands Are Now Testing Campaigns on Synthetic Consumers
Updated: 1 day ago
A 12 October 2026 Digiday report finds agencies planning against AI-generated synthetic audiences, with a Kenvue/Aveeno strategy brief falling from about 30 days to roughly 3.
What happened
A Digiday report on 12 October 2026 detailed a fast-growing practice: agencies and CMOs planning media and creative against AI-generated synthetic consumers rather than only real respondents. Goodway Group and Arima are running simulated incrementality tests, while Vurvey Labs works with brands including Vans and Kenvue.
WPP Media, Dentsu and Stagwell's Code & Theory are all experimenting with the approach. The eye-catching figure: a Kenvue and Aveeno strategy brief that once took around 30 days now takes roughly 3 using Vurvey's synthetic method.
The report is careful to include the skeptics, who warn that weak underlying data produces unreliable personas. That caveat is as important as the headline speed gain, because it defines where the technique is safe to use and where it is not.
Why it matters
Compressing a 30-day brief to 3 days is not a marginal efficiency, it is close to a tenfold change in cycle time. If it holds up, strategy and testing stop being gatekeeping stages that slow everything down and become something teams can run continuously alongside the work.
But the skeptics in the report have the stronger long-term point: weak input data produces unreliable personas. A synthetic audience is only as good as the real behaviour it was trained on, and a confident, fast, wrong answer is far more dangerous to a budget than a slow one.
The deeper risk is cultural. If synthetic testing feels authoritative, teams may stop pressure-testing it against reality, and the industry could quietly optimise campaigns toward a model of consumers rather than consumers themselves.
The speed-versus-validity framework
The right mental model is a trade between speed and validity. Synthetic consumers are superb for divergent exploration, generating and killing ideas cheaply at volume, but risky as the sole basis for a high-stakes, convergent decision like committing a national media buy.
The emerging best practice is a hybrid loop: use synthetic audiences to narrow the field fast, then validate the finalists against real people before any money moves. Treat the AI as a hypothesis engine that proposes, not a judge that decides on its own authority.
This also reframes the value of real research. Rather than replacing it, synthetic testing makes human validation more precious, because it is the gate that catches the confident errors the model will inevitably produce from imperfect data. Think of it as a funnel: synthetic audiences widen the top cheaply, and real people guard the bottom where the money is, which is exactly the division of labour each is best suited to.
What businesses should do
Pilot synthetic testing on low-risk, high-volume decisions first, such as early creative screening or message sorting, where speed matters most and a wrong call is cheap and quick to reverse before it reaches the market.
Insist on provenance for the training data and always keep a real-world validation gate before spend. The firms that win will not be those that replace human research with AI, but those that sequence the two so each covers the other's blind spot.
And govern it explicitly. Write down which decisions may rely on synthetic audiences and which may not, so speed does not quietly creep into choices that should always be checked against actual human behaviour and real money on the line. The discipline is not anti-AI; it is the opposite, because a clear boundary is what lets you lean on synthetic testing confidently in the places it genuinely helps.
Quick FAQ
What is a synthetic consumer?
It is an AI-generated audience or persona, trained on real behavioural data, that brands use to simulate how real consumers might respond to media and creative before testing with live respondents.
How much faster is it?
The Digiday report cites a Kenvue and Aveeno strategy brief that previously took around 30 days and now takes roughly 3 using Vurvey Labs' synthetic approach.
Cutting a brief from 30 days to 3 is exciting, but speed is not the same as truth. A synthetic audience is only as good as the data underneath it, and a fast, confident, wrong answer is the expensive kind. Use these models to explore, then validate the finalists against real humans before you spend. — Daniel Nikolla, Founder of Merx Marketing
Treat synthetic consumers as a hypothesis engine, never as the final verdict before you commit budget.
Which decisions in your process are safe to speed up with AI, and which still need real people?
If you would like to know more about this topic, please contact us on danieln@merxmarketing.co.uk
Sources: Digiday
Related reading
Written by Daniel Nikolla, Founder of Merx Marketing Ltd and Marketing Minute
