#43 AI in User Research: Synthetic Personas and Generated Insights

AI-generated personas and insights are starting to pop up in my LinkedIn feed, often pitched as clever ways to speed up user research. But if you’re aiming to build something better than average, these shortcuts won’t save you time. They’re more likely to lead you in the wrong direction.

When we design systems and products, we’re searching for what makes them valuable for customers and users. Not just functional, but distinctive. We want to help users in ways that actually matter, and stand out from the competition.

If you’re entering a mature category and just need to hit the basics, you can often start from assumptions. Yes, users need to edit this data. Yes, they need to share that. You can probably ship version one without doing deep research.

But once the basics are in place, real differentiation begins. That’s when you need to understand not just what users do, but why, and what they actually need, even if they can’t express it themselves. That’s exactly where generative AI falls short.

Large language models are trained on vast amounts of text. They generate answers by predicting which words are most likely to come next based on patterns in that data. They don’t “understand” content, context, or meaning. They simulate knowledge, they don’t hold it.

So best case, your AI-generated personas or insights will reflect what’s already known in your domain. You’ll get the average answer, the most documented behavior, the default assumptions. That might be fine, but if that’s what you need, why not just buy or copy one of the competing products already built on those same assumptions?

Worst case, the model hallucinates. You get plausible-sounding details that are completely made up. And if you don’t catch them, you’ll waste time and money building features nobody asked for, or solving problems that don’t exist.

Generative AI can be incredibly helpful for tasks like translation, idea generation, and providing starting points. But when you’re trying to uncover unmet user needs, validate critical assumptions, or build something that truly stands out, it won’t get you there.

Synthetic personas and generated insights might feel efficient, but if your strategy depends on creating something truly valuable and unique, they’re more likely to steer you off track than help you move faster, and that will most definitely cost you more than it saves.