Ask an AI to describe "a nurse," and you get a cliché. Kind, tired, run off her feet, almost certainly a woman. Ask it for "a finance director," and you get another one. This is the quiet problem with building a persona from a job title and a couple of traits: what you get back is the stereotype you put in.
Most personas start life as a line or two. "Busy mum, 35, values convenience." "Cautious, time-poor, risk-averse." Hand an AI that little, and it has almost nothing to work with, so it fills the gaps with the most obvious version of that person. The result reads well and is, more often than not, a caricature. That is not the AI being careless, and it is not you asking the wrong way. A thin description can only ever produce a thin person.
The way out is the opposite of what you would expect. It is not fewer assumptions; it is more of them. Here is the strange part: every extra detail you add is itself a small stereotype. "Runs marathons" carries its own cliché. So does "grew up in the countryside", "studied law", "has two young children". But they pull in different directions. Add enough of them, and they start to cancel each other out, and what is left is no longer a type. It is a person. A stereotype is a fine place to start and a poor place to stop.
There is one condition. The details have to belong to someone who could actually exist. Real people are not a random bag of traits. What someone earns tends to go with what they studied, where they live, and the work they do. Give an AI a jumble that does not hang together, an eighteen-year-old retired surgeon on a tiny income with five children, and you do not get a person; you get noise. The extra detail only cancels the cliché when the pieces fit together the way they do in real life.
This is the whole idea behind how Cambium AI builds a persona. Instead of a job title and a guess, each synthetic person is built from hundreds of details that genuinely occur together in real public data: age, income, education, household, where they live, and a great deal more, in the combinations you actually find in the population. Enough real, connected detail that the person stops being a stereotype and starts being someone specific.
For a marketing team, that is the difference between a persona you show in a meeting and one you can plan a campaign around. A one-line persona quietly narrows your audience to its most obvious slice and hides everyone else, which is how most AI personas miss the audience they are meant to describe. A fuller one shows you the range of people you are actually trying to reach, including the ones the cliché leaves out.
And because each person is built from public data, you can check the work rather than take it on trust. That is the real point: a persona you can verify before you spend on a campaign, not a character you hope is right. There is a worked example of grounding a market decision this way.
The next time a persona feels a little too neat, that is the tell. A real audience is messier and more interesting than its stereotype. The way to see it is more detail, grounded in real people. Have a look in the app →