What the models are built from
Data can record what people say, who they are, or what they do. Each provider uses a different combination.
Surveys and panels
What people sayAnswers to known questions from known people. Limited to the questions that were asked.
Interviews
What people sayA long conversation with one person, in their own words. Deep for each person, slow and costly to collect at scale.
Public text
What people sayReviews, posts and forum threads, written without a researcher present. Large and in people’s own words, but limited to what people choose to post, and attributes have to be inferred. Kosinski et al., 2013
Census and official statistics
Who people areThe shape of a population by age, income, place and household. Says little about attitudes.
Behavioural records
What people doTransactions, store visits, app use and clicks. Records what people do rather than what they say, and is usually licensed.
Experiment responses
What people doHow people responded in controlled studies, which records treatment effects as well as answers. Kolluri et al. (Simile), 2025
A customer’s own data
Any of theseResearch, support conversations or panels from the organisation asking the question. The exact audience, but often a small sample.
Only what a language model learnt in training
Nothing collectedAny audience can be described in a prompt, and the answers reflect the model’s stereotypes of that audience. Wang et al., 2025