While much of the industry is asking how AI will change research, we are asking how AI is changing the consumer we are researching.
What Esomar Congress 2026 in Valencia tells us about the next phase of innovation in insights
Coming back from this year’s Esomar Global Congress in Valencia, one thing felt clear: AI in market research has moved beyond the “Can it do this?” phase.
The conversation is increasingly about where AI genuinely adds value, where it can already be deployed at scale, and where human expertise and traditional research remain essential.
Across presentations and exhibitors, the innovation agenda ranged from Digital Twins and Synthetic Data to AI-moderated interviewing, automated analysis and increasingly integrated research workflows.
Two areas in particular stood out.
Digital Twins: the real question is “accurate for what?”
Digital Twins and Synthetic Data were certainly among the hottest topics in Valencia. But the discussion felt more mature than the hype surrounding them might suggest.
Research agencies, technology providers and client organisations are no longer simply asking whether synthetic respondents can reproduce human responses. They are increasingly trying to establish which research applications they already work well for and which remain unreliable or require considerable caution.
This distinction matters.
We should probably stop asking whether a Digital Twin has one overall “confidence score”. Reliability is application-specific.
A model may perform well in identifying a rank order of needs, exploring hypotheses, supporting early-stage screening or simulating reactions to familiar situations. The same model may be much less dependable when asked to reproduce precise market incidence, predict genuinely new behaviour or capture decisions strongly influenced by context, emotion and lived experience.
The right question therefore moves from: “Does synthetic data look like human data?” to: “Can I trust it for this specific research question and business decision?”
For Eumetra, this is also the most constructive approach: embrace the potential, continue experimenting, but validate by use case rather than assuming that one level of reliability applies across every research application.
Qual at Scale: moving rapidly into the mainstream
The second change felt even more tangible.
The number of exhibitors, platforms and presentations dedicated to AI-moderated interviewing and Qual at Scale suggests that this is rapidly becoming a mainstream research option rather than an experimental niche.
Commercially, one of the more counter-intuitive aspects of Qual at Scale is that it is not necessarily cheaper than traditional qualitative research: the cost per AI-moderated interview can be comparable to, or even higher than, a traditional IDI. The premium is justified less by cost efficiency than by what the technology enables: much faster fieldwork, the ability to run large numbers of in-depth conversations in parallel and the possibility to conduct international studies across multiple languages without the same dependence on local moderation teams. In other words, the value proposition is primarily about speed, scale and geographic flexibility, rather than simply reducing the cost per interview.
Traditional qualitative researchers will continue – quite legitimately – to argue that an AI-led conversation cannot reproduce everything that happens between an experienced moderator and another human being. They are right. But that may also be the wrong comparison.
AI-moderated research does not need to replace traditional qualitative research to become extremely useful.
Its strengths are different. It can conduct tens or hundreds of individual conversations simultaneously, probe dynamically, accelerate fieldwork and analysis, create consistency across markets and reduce some of the linguistic and organisational barriers involved in international qualitative research.
There may also be particularly interesting applications around sensitive subjects. In certain circumstances, respondents may feel more comfortable discussing a personal or potentially embarrassing issue with an AI interviewer than with another person whom they fear may judge them.
Equally, an AI interviewer may fail to detect cultural signals, emotional nuances or subtle changes in behaviour that an experienced moderator would immediately recognise. Again, the answer is not to decide which methodology “wins”. It is to understand where each approach produces the greatest value.
The human researcher is not disappearing. Their value is moving
AI will make many activities involved in research faster, more scalable and increasingly automated. Transcription, coding, repetitive probing, first-level synthesis, translation and basic reporting will require progressively less manual intervention.
But this does not make researchers less important. It makes different research capabilities more important.
Framing the right business question. Choosing the appropriate methodology. Understanding when not to trust the technology. Interpreting culture and context. Connecting different sources of evidence. Challenging an apparently obvious answer. Turning research into a decision.
The value of the researcher is moving upstream: from producing information to making sense of it.
And this brings us to what may be an even more important transformation.
Innovation needs to happen on both sides
The lesson we take from Valencia is therefore not that traditional research is disappearing, nor that every AI-enabled methodology should immediately replace the approaches we use today. It is that innovation now needs to happen on both sides of the research relationship.
On one side, we need to continue exploring how Digital Twins, Synthetic Data, AIMI, Qual at Scale and other AI-enabled methodologies can improve the way research is conducted – while remaining rigorous about where they work, where they do not and how they should be validated.
On the other, we need to understand an equally profound transformation taking place outside our industry: AI is changing how people discover, evaluate and choose.
Perhaps the biggest change AI brings to market research will therefore not come only from changing the tools researchers use.
It may come from recognising that the consumer we are researching is changing too.
We are changing how we research consumers. But consumers are changing too
Looking at the development of AI over the last couple of years, we could summarise the evolution in three questions:
- 2023: Are consumers using AI?
The first challenge was to understand adoption: who was using ChatGPT and other generative AI tools, how often, and for what purposes. As early as 2023, Eumetra began exploring how AI was starting to reshape people’s behaviours and decision-making.
- Esomar Valencia 2026: How should researchers use AI?
Much of the debate at Esomar was understandably centred on our own industry: Digital Twins, Synthetic Data, Qual at Scale, automation and new research workflows.
But there is a third question which, at Eumetra, we believe deserves just as much attention.
- Eumetra’s question: What happens when AI itself becomes a new intermediary between the consumer and the brand?
Because AI is no longer only sitting on the researcher’s side of the table. It is increasingly sitting on the consumer’s side too.
Consumers are using AI environments to discover products, understand categories, compare alternatives, solve problems and ask for recommendations. And when that happens, AI becomes more than another source of information or another touchpoint. It starts participating in the decision itself.
A consumer may no longer search for ten brands, visit multiple websites and compare all the options personally. Instead, they may explain their needs to an AI and simply ask: “Which one would you recommend for me?”
At that point, AI interprets the need, decides which attributes are important, identifies the relevant brands, creates the consideration set and potentially recommends a winner.
Eumetra’s The Future of The Customer Journey
That represents a fundamental change in the Customer Journey. This transformation is at the heart of Eumetra’s The Future of the Customer Journey research programme.
Our objective is not simply to measure how many people use AI. We want to understand how AI is beginning to mediate the relationship between consumers, categories and brands.
- What are consumers actually asking?
- At which stages of the journey are they turning to AI?
- Which parts of the decision are they willing to delegate?
- How important is AI becoming compared with traditional touchpoints?
- Which brands enter AI-generated recommendations and which do not?
- Does a brand’s strength in the real market translate into comparable visibility within AI responses, or are some brands disproportionately amplified while others are effectively filtered out?
- And perhaps most importantly: why?
- Which attributes and associations make an LLM more likely to recommend one brand over another and how do those criteria compare with the factors that actually drive consumer choice?
This creates a new layer of customer understanding: not only what consumers think about brands, but increasingly what the AI systems helping consumers make decisions “think” about those brands too.
This is where a new dimension of relevance emerges: brands are no longer competing only for people’s attention, but also for the attention of artificial intelligence.
Through international benchmarking, proprietary segmentation and prompt analysis across specific target audiences and tasks, The Future of the Customer Journey explores how AI is transforming needs, behaviours and decision-making criteria — and what this means for brand visibility and relevance.
We will explore this on 27th October at Eumetra International’s first webinar dedicated to The AI-Mediated Customer Journey.