AI Personality Analysis: How an AI Personality Report Actually Works

Three kinds of “AI personality analysis”

The first kind is a research area of its own, and some commercial tools use it to profile people at work. The second is what you get if you ask a chatbot to “analyse my personality” after a few messages. The third separates two jobs that should be separate: measurement, which a questionnaire does well, and explanation, which a language model does well. The rest of this guide is about the third kind.

How a scored AI report is built, step by step

1. You answer a fixed questionnaire

The items are the same for everyone and are scored the same way. In our case the core is about 80 questions. Built on the public-domain IPIP item pool for the Big Five, with attachment items adapted from the ECR-R and career items from Holland's RIASEC model. Some items are scenarios or sliders, and a few optional boxes let you add a sentence in your own words.

2. The answers become scores

Ordinary code, not AI, turns your answers into numbers: five trait scores from 0 to 100, plus results from any add-ons you took, such as an attachment pattern or a three-letter RIASEC career code. You see these scores for free before anything is written.

3. The model gets a tightly specified brief

The language model receives your scores and answers along with detailed instructions: which sections to write, roughly how long each one is, what to cover in each, and what not to do. Ours, for example, is told to read the five traits together rather than one at a time, to name a specific behaviour rather than a vague quality, and to present MBTI-style and Enneagram labels only as rough approximations derived from the Big Five, not as measured types.

4. The model writes for your exact combination

This is the step that justifies using AI at all. A traditional report is written in advance, one block of text per type or per score band, and stitched together. That cannot handle the way traits interact: high Openness reads differently next to low Conscientiousness than next to high Conscientiousness, and the number of combinations across five traits, an attachment pattern and a career code is far too large to write by hand. A model can write for the combination in front of it. Our article on trait interactions shows why this matters.

5. The output is checked and formatted

The model’s output is parsed into fixed sections and checked for completeness. If a report comes back incomplete, the request is tried again with another model. The finished report is then available as a page and a PDF.

Where AI personality reports go wrong

Fluent Barnum statements

In 1949 Bertram Forer showed that people rate vague, flattering descriptions as highly accurate even when everyone receives the same text. Language models are very good at producing that kind of text. “You value deep connections but also need time to yourself” is true of almost everyone, and it sounds wise.

Invented specifics

The opposite failure is a confident detail with nothing behind it: a claim about your childhood, or a prediction that does not follow from any score. A good brief tells the model to tie each insight to a pattern in your scores and a behaviour it produces, but you should still notice when a sentence seems to come from nowhere.

Flattery

A report that only finds strengths is pleasant to read and useless to act on. Look for an honest account of costs, such as what high Neuroticism actually feels like, or how high Agreeableness turns into quiet resentment.

Overclaiming the measurement

No amount of good writing makes a short self-report questionnaire into a clinical instrument. These are our own short forms. They are not independently validated, and this is a self-reflection tool, not a clinical one. A report that hides this is a warning sign. The guide on how accurate personality tests are covers the limits in more detail.

Five checks before you pay for an AI personality report

1. Can you see your scores? If the product will not show you the numbers the writing is based on, you cannot tell whether the writing follows from them.
2. Is there a complete sample? Not a teaser paragraph: a full report you can read, so you can judge the writing yourself.
3. Does the text depend on you? Pick five sentences. Would each be wrong for someone with opposite scores? If not, it is filler.
4. Does it say what it is built on? A named questionnaire or item pool, and an honest caveat about what self-report can and cannot show.
5. What happens if it misses? A refund policy tells you whether the seller expects the report to land.

How the Depth Profile report applies this

The free assessment shows your Big Five scores and archetype first. The Full Depth Report is written from your scores, your scenario choices and anything you wrote in the optional boxes, and adds a long-form portrait, a written facet reading for each trait (descriptive, not facet scores), attachment and career chapters if you took those free add-ons, a 30-day growth plan built around three experiments, and 50 prompts you can paste into ChatGPT, Claude or Gemini so the assistant knows how you think. It costs $19 once, includes a PDF, and comes with a 30-day money-back guarantee. Details are on the pricing page.

If you plan to use the prompts, the guide to giving ChatGPT your personality profile shows how to set them up. And if you would rather have a trained person interpret your results, a practitioner-administered assessment with a debrief is a better fit than any automated report.