While traditional market research remains the foundational standard for deep consumer understanding, the typical six-to-eight-week timeline and required financial investment can sometimes limit a brand’s day-to-day agility. This is where synthetic audiences and data offer a practical alternative. By using AI to mimic real-world populations or groups, this approach can generate useful insights within minutes. Rather than relying on a single AI request, this methodology utilizes “swarm technology” – where thousands of distinct, localized synthetic personas respond individually – allowing brands to efficiently pre-screen campaign strategies, validate creative concepts, and complement their existing first-party data in a privacy-compliant manner.
In this interview, Quintijn van Kessel, the Global Head of Innovation at Incubeta explores the mechanics, practical applications, and necessary guardrails of this technology, and how, when deployed in tandem with other tools, it can serve as an efficient addition to modern marketing strategies.
What is synthetic audiences and data?
Synthetic audiences uses AI to mimic audiences and give you insights on how people would react to certain topics, like price increases or a new set of campaign creatives.
This means you ask AI to answer a specific question as a group of users or people. It is important to NOT do this as one single AI request per audience. Instead you ask this question to the AI for every persona in the audience, allowing the AI to really take on the role of each specific persona. This can easily scale up to hundreds or even thousand requests and that is what we refer to as “swarm-technology”.
From there, all the various responses are used to form a single answer, but the insights of all the different persona-responses are still available for deeper insights and research.
What are “Synthetic Personas,” and how do they change campaign planning?
Synthetic personas are the mimicked people within synthetic audiences. Not all synthetic audience providers take the route to actually create all different sorts of synthetic personas that, when done right, mimic the population of a country or different countries. But that is where the strength of synthetic audiences actually lie. It should not do one AI interaction with the request to act like a synthetic audience. No, it should make a big amount of AI interaction to different personas that have been matched to synthetic audiences. Only then you get the same level of insights as you would get via traditional market research.
Seamless Research (Incubeta’s synthetic audiences tool) uses a cluster of 10 000 synthetic personas per country, mimicking the population distribution (age, gender, marital state, income, interests) of those countries.
How is this relevant to marketing, and why should marketing leaders care?
Within marketing, it’s very common to do extensive market research, but we also know that this tends to be very expensive and time consuming. That is exactly where synthetic audiences come into play – not as a complete replacement for marketing research, but rather as a way to do pre-research in a faster and cheaper way. Imagine you can have market research insights within a minute against a very low cost. What would stop you to validate your campaign strategy or your campaign creatives before actually putting them live?
I think that every marketer, from trader to campaign strategist to CMO should be thrilled to have this to their disposal.
There is often a struggle with small sample sizes in niche markets. Can synthetic data help with this?
Here synthetic audiences could indeed provide a solution in a way that you can amplify this group to get some real usable insights. But you should be careful not to amplify such groups too much, because that will take their voice out of the actual context. They are a small group after all in the real life environment you operate within.
If you would use Seamless Research you will automatically be limited in amplifying specific groups, because the base of personas is a proper representative group towards the actual population. Of course this can still help amplifying groups that are under-represented in your own dataset, but never to the extent that they will be relatively bigger than they actually are in the real world.
With privacy regulations tightening globally, how does this impact compliance?
The nice part of synthetic personas and audiences is that you don’t use actual user data. Instead you ask AI to act as different persons, without sharing the actual user data with the AI. All the persona enrichment is done in a generic and aggregated way via, for example, data from each country’s national statistical institute. Sometimes client data is also used to enrich the synthetic audiences, but always in a non-personal and non-identifiable way. The actual users are not of interest, but rather how they – as a group – behaves.
How does synthetic data speed up go-to-market testing?
With synthetic audiences you are able to get market insights for a fraction of the cost and at a much faster rate. Your research questions will be answered within a couple of minutes instead of the standard six- to eight weeks. That is a massive difference, and allows you to go-to-market much faster, especially if you want to do a secondary research to get deeper insights.
But there is a big caveat here as well. Although synthetic audiences can be very useful to quickly gain market insights, they do not completely replace the traditional real market research. I strongly believe you still need to do that type of traditional research, but in another way. Do this research in such a way that you learn not only about a specific topic, but also about your users or clients – because those insights can later fuel the synthetic audiences.
Are there risks of bias or inaccuracies when using AI-generated data?
Yes, that is 100% a risk, as it always is when using AI. There are times it can definitely be biased, or even flat-out wrong. Of course, there are ways to minimize this risk by, for instance, using well structured and complete prompts and making the sample size big enough to reduce the effects of wrong answers. But you will always need to look at the output of AI with a cautious mind – and with synthetic audiences that is no different.
Should marketers replace their first-party data strategies with synthetic data?
No, absolutely not. First of all synthetic data is the mimicking of users, not actual users. This means that the users of the synthetic audiences are not targetable, unless you have a CMS integrated into this process that allows you to map the synthetic audiences to real users or user groups. But in that case you are still using a first-party data strategy, which you have extended with synthetic audiences to make the personas interactive. Secondly, the synthetic personas and audiences are not a fail-proof know it all solution. I think nothing can beat first party data when processed, analyzed, segmented.
Essentially, I see synthetic audiences as an enrichment of first-party strategies. You can take these strategies to a higher level by making sure you have representative synthetic audiences, which will allow you to pre-screen your campaign-strategy and creatives for instance.
Where do marketing leaders start if they want to deploy, and unlock the power of synthetic data?
Synthetic audiences is quite a complex topic, since it is not just a case of asking AI to act like one. So my suggestion is to look into a tool that will help you get the most out of synthetic audiences, like Incubeta’s Seamless Research.
Once that is in place it is crucial to take all the possible steps to enrich the personas and audiences with your company data, so that the mimicked personas and audiences fit the users/clients of your company as much as possible.
Finally it is a question of experimentation with the tool, the personas and the audiences. Make sure you do not only ask questions, but also dive in the data the tool returns. For instance, with Seamless Research it is possible to do much deeper insights on the answers to see what part of an audience (personas) answered in line with the overall audience level answer, also split out on demographic level, and more. It is even possible to see the actual answer of each persona individually, allowing you to really learn the “why” behind the answers. This will allow you to really fine-tune your campaign strategy or creative for instance.
Could a Synthetic Audience Help Build or Influence an AEO Strategy?
A synthetic audiences can play a big role in a company’s AEO strategy, specifically with the following four areas:
It can help identify how your clients would interact with an LLM: Traditional keyword tools tell you what people type into Google (e.g., “best CRM for small business”). But people talk to AI engines differently – they use longer, highly specific prompts (e.g., “I have a 10-person agency and need a CRM that integrates with Slack and costs under $50/month. What are my options?”).
You can use synthetic audiences to generate hundreds or even thousands of these conversational, long-tail prompts. By seeing exactly how your ideal buyers would interrogate an AI, you know exactly which specific questions your website content needs to answer to get pulled into those results.
It can help with pre-testing LLM outputs: With the previously mentioned insights on persona queries you can check with an LLM how well your new content will score in the well known LLMs. This will allow you to further optimize your content to be the authority on specific topics for those LLMs.
It can help to simulate topic authority gaps: Synthetic audiences can help you map out the entire universe of a persona’s anxieties, objections, and niche questions. AEO relies heavily on being the most comprehensive, authoritative source on a specific micro-topic. By running simulations with synthetic personas, you can discover hidden information gaps that your competitors have missed, allowing you to publish the exact missing data the AI is looking for.
It can help to keep content interesting for different humans: In the end all your content will not only be read by the LLM, but also by the humans visiting your website. It would be a shame to have won your visits only to lose them afterwards. So let your new content also be checked by the synthetic audiences to verify how the different groups would respond to it. It is very hard to create content that will resonate with everybody, but synthetic audiences can help you identify how each group would respond to the new content, allowing you to optimize it to your prime target audience.
Synthetic Audiences and Data in Action: See How It Works
To demonstrate how synthetic audiences and data can be used in practical ways, Quintijn developed a dedicated synthetic audience configured specifically for the 2026 FIFA World Cup. It ingests fresh World Cup data every hour, and importantly, during live matches, the data refreshes every five minutes.
Once you’ve asked a question, wait a few moments for the tool to gather information from the synthetic audience. It will produce a general sentiment, how the audience splits, as well as the voices behind it. In this demonstration, the practical application of synthetic audiences and data can be seen in important global moments for markets, like the 2026 FIFA World Cup.
