Synthetic Personas: How to Use AI-Generated Personas

A synthetic persona is a customer profile generated by AI without studying real people, so treat it as a hypothesis to test, not as research. The User Persona Generator builds one from a business description and a target audience.

What are synthetic personas?

Nielsen Norman Group defines a synthetic user as an AI-generated profile that attempts to mimic a user group, with findings produced without studying real users. It is built on large language models trained on vast amounts of data about people.

A classic persona is different. NN/g says personas must be based on user research, such as interviews, surveys and field studies. A synthetic persona skips that step.

A synthetic persona is a static profile. Some "synthetic user" platforms go further and let you chat with the profile, which is a separate product from a persona.

How to generate a synthetic persona

With the User Persona Generator, which is free and needs no sign-in:

1. Open the generator.
2. Enter the Business description: what your product is and how it works.
3. Enter the Target audience: the segment of customers you want to reach. Be specific.
4. Choose a Language for the result.
5. Click Generate for FREE. The persona appears on the same page in about ten seconds.

The result has eight parts:

• Name: a label you can use in your marketing plan.
• Overview: a high-level description of the persona's life.
• Problem: what bothers the persona.
• Pains: how the unsolved problem affects them.
• Goal: the change they want.
• Benefits: the positive outcomes they want.
• Trigger: the event that pushed them to look for a solution.
• Barriers: the objections that slow them down.

Copy or download the result. The more detail you put in the two text inputs, the sharper the persona. [FounderPal](/) also has other generators that use the same description, such as the target audience generator.

What synthetic personas are good for

• A first picture of a segment you do not know. NN/g notes that a synthetic user can synthesize a large amount of available information about a user group and present it in a digestible way. When NN/g asked a synthetic user to describe a typical day, the answer was very similar to what real users had said in their study.
• Hypotheses and interview questions. NN/g's guidance is to use synthetic users for desk research and generating hypotheses, and to treat the output as a hypothesis that guides later research.
• A first draft of ad and landing-page angles. Pains, triggers and barriers give you claims to test against real customers.

Academic work points the same way. In Out of One, Many, Argyle and colleagues conditioned GPT-3 on thousands of socio-demographic backstories from real participants in US surveys. They found it could emulate the response distributions of human subgroups, a property they call "algorithmic fidelity". That result rests on backstories taken from real people, not on a profile typed in one line.

Synthetic personas vs real customer interviews

NN/g gives examples of the honesty gap. Asked whether they had finished all the courses they took, synthetic users said yes, while real users admitted they had not. Jeff Sauro and colleagues found that ChatGPT vastly outperformed most real people in tree testing, so it was not an acceptable replacement for them.

The rule from NN/g: supplement, do not substitute. Use a synthetic persona to prepare for real research, not to replace it.

How to check a synthetic persona against real customer signals

Go through the persona field by field and look for evidence you already have or can get cheaply.

• Problem and Pains: compare them with support emails, product reviews, sales-call notes and community posts where your audience talks.
• Goal and Benefits: compare them with the words customers use in testimonials and reviews.
• Trigger: ask five recent customers what happened right before they started looking for a solution.
• Barriers: list the objections from lost deals and the questions that go unanswered before purchase.

Then sort every claim. Keep what at least two real sources confirm. Rewrite what contradicts them. Mark the rest as unverified until you hear it from a customer.

If the persona reads generic, run the generator again with a tighter target audience. To gather the real signals, follow how to do market research for a startup. NN/g explains why personas need research behind them to represent a product's users accurately.

Mistakes to avoid

• Deciding pricing or product scope from one synthetic persona. NN/g warns against using synthetic users for final decisions.
• Writing a broad audience such as "everyone". The persona comes out generic and nobody recognizes themselves in it.
• Never revising the persona after real conversations. Update it each time customers confirm or contradict a field.

To see what finished personas look like, browse the user persona examples. Turn the fields you confirmed into a message with how to define a value proposition.

FAQ

What is a synthetic persona?

A synthetic persona is a customer profile generated by AI without studying real people. It describes who the customer is, what bothers them and what holds them back, based on patterns in the data a language model was trained on.

Are synthetic personas accurate?

They can be a useful starting point, but they are unreliable for final decisions. NN/g found that synthetic users give shallow or overly favorable feedback and tend to please. Use them to form hypotheses, then check those against real customers.

What is the difference between a synthetic persona and a synthetic user?

A synthetic persona is a static profile you read. A synthetic user, as NN/g describes it, is a profile that is "alive", meaning you can chat with it and run a simulated interview. Both are produced without studying real users.

Is the FounderPal User Persona Generator free?

Yes. It is free, needs no email, and you can copy or download the result. It returns a persona with eight parts, from the problem and pains to the trigger and barriers.