Artificial intelligence seems to have arrived everywhere in user research.
AI can now write research plans, generate discussion guides, moderate interviews, analyse transcripts and even create synthetic users that mimic customer behaviour. Depending on who you listen to, it's either the future of research or the end of it.
The reality, as ever, is more nuanced.
I recently sat down with user research consultant and founder of Drop In Research, Nikki Anderson, on the Web Usability Podcast to discuss where AI is genuinely adding value and where we need to be more cautious. One thing became clear very quickly.
AI is a powerful tool.
It is not a replacement for understanding real people.
Why Researchers Are Embracing AI
It’s easy to see the appeal.
Research involves plenty of repetitive and time-consuming tasks. Creating recruitment screeners, reviewing secondary research, drafting plans and analysing large datasets all take time.
For busy teams, AI offers the chance to speed up those activities and free researchers to focus on higher-value work.
Nikki highlighted several areas where AI is already proving useful, including desk research, drafting research materials, identifying gaps in thinking and supporting quantitative analysis.
I would add that AI can also be a useful sounding board. It can challenge assumptions, help researchers spot alternative perspectives and pressure-test ideas before taking them into the field.
Used well, these tools can save time and improve efficiency.
The challenge is ensuring efficiency doesn’t come at the expense of quality.
Faster Doesn’t Always Mean Better
One of the biggest misconceptions about AI is that it can replace thinking.
In reality, the quality of any AI output depends heavily on the quality of the person guiding it.
I recently found this when working on a discussion guide. My first instinct was to ask AI to draft it for me. Very quickly, I realised that I was skipping the most valuable part of the process: thinking.
Writing a discussion guide isn’t simply about creating questions. It’s about understanding business objectives, stakeholder priorities and participant motivations. It’s about deciding what needs to be learned and how best to uncover it.
That’s where researchers add value.
I still use AI in the process, but more often to critique, challenge and refine my thinking rather than replace it.
AI can accelerate execution.
It cannot replace judgement.
The Risk of False Confidence
One of the most interesting points Nikki raised was the danger of AI-generated confidence.
AI is exceptionally good at producing outputs that look polished, complete and convincing.
The problem is that polished doesn’t necessarily mean correct.
Experienced researchers can usually spot weak outputs because they know what good research looks like. The bigger risk is for less experienced practitioners who may struggle to recognise when something sounds plausible but lacks rigour.
Poor quality evidence leads to poor quality decisions.
In user research, that can mean wasted investment, missed opportunities and products that fail to meet user needs.
As researchers, we need to remain critical consumers of AI outputs, not passive recipients of them.
Can AI Moderate Research?
AI moderation is one of the most talked-about developments in the industry.
These tools can conduct interviews, ask follow-up questions and generate transcripts without a human moderator present.
It’s an intriguing concept, but there are limitations.
Great moderation isn’t simply about asking questions. It’s about building trust, reading the room, recognising emotion and knowing when to dig deeper or change direction entirely.
These are deeply human skills.
Current AI moderators often struggle with nuance. Conversations can feel repetitive and rigid, with participants reporting frustration when questions are repeated or context is missed.
That doesn’t mean AI moderation has no place.
I can see value in certain situations, particularly for large-scale studies, multilingual research or short exploratory interactions where depth is less important than reach.
But I don’t see it replacing the rich, nuanced conversations that come from speaking to real people.
The Role of Synthetic Users
When I first heard about synthetic users, I was sceptical.
The idea of replacing research participants with AI-generated personas felt uncomfortable. After all, the foundation of user research is understanding real human experiences.
Over time, however, I’ve started to see where these tools can play a useful role.
Synthetic users can help researchers pressure-test assumptions, refine research questions and prioritise areas for investigation before investing in fieldwork.
The key is understanding what they are and what they are not.
Synthetic users are not replacements for real participants.
They are another research tool.
Like any methodology, they have strengths, limitations and appropriate use cases.
The question isn’t whether synthetic users are good or bad.
It’s whether they are the right tool for the problem you’re trying to solve.
Throughout my conversation with Nikki, one theme kept resurfacing.
Human experience is difficult to replicate.
AI can identify patterns. It can generate responses. It can help us process information faster.
What it cannot do is live a human experience.
That becomes particularly important when researching accessibility, inclusion, emotional decision-making or complex life experiences. These are often the moments where the richest insights emerge and where researchers add the greatest value.
For organisations looking to explore AI, my advice is simple: start small and be intentional.
Use AI to support desk research, planning, competitor reviews and quantitative analysis.
But don’t stop talking to your users.
Because the goal of research has never been simply to collect data.
The goal is to understand people.
AI can help us move faster. It can help us work more efficiently. It can even help us ask better questions.
But understanding why people behave the way they do remains one of the most valuable things any organisation can do.
And for now, that still starts with listening to real humans.