Whether you're a designer or a founder launching a product or service, creating personas is a critical step on your journey to successful outcomes.

The question we explore today is whether we can use ChatGPT — one of our favourite AI robots — to aid us in the process of creating user personas and getting know our target audiences better.

We will create personas using ChatGPT and assess the quality of the results. We based our experiment on a target audience that is very near and dear to us: designers!

What are user personas and what is their purpose?

UX designers are well versed in what personas are and why they matter. But there are still many people who are less familiar.

Personas are fictional representations of a target audience, capturing their key traits, behaviors, and goals. These representations help us to empathize with the person who will use a product or service.

This, in turn, enables designers — or developers or marketers or founders — to create tailored, user-centered products and services that can effectively address our audience's needs and expectations.

To say it's a pretty useful tool is an understatement. Especially for those people on a team who may not be able to talk to the users or customers directly themselves. That's the exact scenario that user personas were created for, by software developer Alan Cooper way back in the '80s.

A Quick Intro to ChatGPT, by ChatGPT

We asked ChatGPT4 to describe itself to a group of designers:

"Hi, I'm ChatGPT, an advanced AI language model developed by OpenAI. I've been designed to understand and generate human-like text. I'm here to help you with tasks like answering questions, generating content, and offering suggestions, making it easier for you, as designers, to work more efficiently and creatively in achieving your goals."

ChatGPT works by finding patterns and relationships in a huge amount of training text, using a method called deep learning. When you interact with ChatGPT, it generates a response based on its understanding of language and context. It selects the most likely words or phrases by essentially predicting what normally comes next in the sequence. In other words, it's playing the odds, going for a relatively average response.

If you want to learn more about ChatGPT and how it works from a technical perspective, computer scientist Till Musshoff covers the technical basics in an approachable 10 minute video.

How not to use ChatGPT to create user personas

Before we share our approach, we're going to share how not to use ChatGPT to create user personas.

Many tutorials out there tell you to straight up ask ChatGPT to "Generate 3 personas for ___", with little or no preliminary information.

The results are instantaneous and, on the surface, they do sound similar to traditional user personas.

However, there is a one massively important difference: while user personas are always fictional, they are meant to be created using research and analysis from actual user data.

In other words, they are meant to be developed from data gathered while interacting with real people.

As Nick Groeneveld of The Designers Toolbox said, “[UX] doesn’t work this way. You get requirements and build personas by talking to users and stakeholders.”

ChatGPT, on the other hand, completely makes up these personas if you don't give it any guardrails or guidance.

Given its training, it has likely seen personas related to your field of inquiry, so there may be grains of truth to them. There are, after all, busy moms like Sarah who don't have a lot of time to schedule doctor appointments for their families.

But these personas don't achieve their primary directive: to communicate insights about real people so that designers, developers and marketers can create solutions that will effectively solve these users problems and pain points.

Chris Kernaghan goes further saying, "using ChatGPT to generate personas can lead to inaccurate or unrealistic representations of users, which can ultimately harm the design process."

Danger Ahead

Whipping up personas in ChatGPT without guidance or vetting is a bad idea. If a persona is off-track, a lot of things can go wrong:

1. Design disasters: Product designers may create features that make users go "meh" or "huh?" instead of "wow!"
2. Ineffective marketing: Marketing messages and strategies may be developed that don't resonate with the real audience, leading to poor engagement and conversion rates.
3. Resource roulette: You'll be throwing time, money, and effort down a black hole by developing features, services, or products that don't address real user pain points or desires.
4. Customer dissatisfaction: Customers might find the product or service frustrating or irrelevant, leading to negative reviews, higher churn rates, and damaged brand reputation.
5. Hindered growth: When the offering doesn't align to audience needs, the product or business' growth may sputter and stall.

There are a few more reasons why using ChatGPT to create personas is problematic.  

1. ChatGPT lies. It's been known to make up statistics, disagree with established facts, or provide inconsistent answers with the aim of providing some natural variety in its responses.
2. ChatGPT is biased. Any biases in the text it's consumed are baked in; and any biases in its "human reinforcement learning" trainers will be reflected too.
3. ChatGPT is out of date. Its training material was cut off in September of 2021 and, unlike Microsoft's Bing, it is not connected to the internet.

Um, so why even consider using ChatGPT for user personas?

It does all sound pretty damning, doesn't it?

Any self-respecting designer would surely say, nope, uh-uh, not gonna do it.

Buuuut, personally, we like to test the boundaries. 😈 And we like finding efficiencies. 😇

Plus, we have had so many excellent outcomes — in business, in art, in design and in life — by being open to random, unexpected ideas and possibilities, that we are frankly addicted to at least giving it a try.

How we used ChatGPT to create personas...and liked it

First, we asked ChatGPT itself for its take on the best way to use it for creating user personas.

It gained our respect by saying it needed us to provide demographic and psychographic data about our target, such as interests, values, and behaviors. That's good advice, and a simple step that many tutorials we've looked at skipped right over.

Asking ChatGPT for the best way to work with it is a surprisingly effective technique, and we encourage you give it a try any time you are considering how to use it for design.

Next, it provided us with a handful of tips for getting good results:

Tips for Using ChatGPT to Create Personas

Here are some tips to help you get the most out of ChatGPT when creating personas:

Be specific: The more specific you are about your target audience, the more accurate the persona generated by ChatGPT will be.

Ask open-ended questions: To get the most detailed responses from ChatGPT, ask open-ended questions that encourage the persona to share their thoughts and feelings.

Use multiple prompts: Don't rely on a single conversation with ChatGPT to create your persona. Instead, use multiple prompts to gather a range of insights about your target audience.

Over the past few weeks, in preparation for launching our design newsletter Unicorn Snacks and our parent company Unicorn Fuel, we talked with designers about the problems they are facing, their awareness of AI-enabled design tools, and their concerns about how AI might impact their job security.

These interviews gave us a baseline understanding of our audience, but also allowed us an opportunity to provide ChatGPT with detailed user profiles.

So we provided ChatGPT with the insights gleaned from these conversations, along with some demographic and psychographic data, and then asked it to generate personas.

The results were fascinating

We were delighted by the results of the experiment—as designers, as business owners and as endlessly curious people.

Does this mean that we think the data should be used as fact? Hell no. But that didn't mean it wasn't a valuable exercise that we would recommend to designers and founders alike. The results aren't what we would call true user personas in the traditional sense, but as one more way to think through ideas we feel it fits nicely into the design thinking universe.

Here are the top three aspects that we appreciated the most:

1) Perfectly Formatted Personas

Instead of giving us back walls of text, ChatGPT chose to create three distinct personas and formatted its response accordingly. They were well-written, clear, and similar to what you might produce yourself — but just in a few seconds. This definitely saved us time.

And because we seeded it with real interview data, these personas reflect our knowledge of the audience quite well.

2) Extended Interviews

Now that ChatGPT had developed Sam, David and Michelle, we could ask them questions.

In the interest of science, we also asked these personas the very same questions we'd already asked our real-world designers in interviews. And what we got back was quite close!

For instance, we asked designers on a scale of 0 to 5 how concerned they were that AI would 1) threaten job security in general for designers and, 2) threaten their own position.

Interestingly, ChatGPT accurately predicted that designers across the board were less worried about their own job than those of designers in general.

There was one aspect where ChatGPT's personas were consistently off, however. ChatGPT's personas consistently ranked their concern level about the threat of AI on their job security as at least a full point higher across the board than the real designers we spoke to.

We found that both amusing and a bit disturbing.

Once we felt the answers were coming back generally in the same ballpark as our real conversations, we went beyond and asked additional new questions, like, "What are their annual salaries?"

In general, questions about money can be tricky and this is where ChatGPT came in handy. While careful to provide caveats for the results, ChatGPT explained that it took into account factors like industry standards, regional differences, and the various job titles. For a fictional persona, this information was valuable.

Our light internet search showed numbers ranging from $63k to $79k for David's demographic of a freelance designer located in Seattle, so ChatGPT's $70k was reasonable.

3) Role-Playing! (oh, do we sound too excited about that? 😳)

At the risk of being a broken record, these personas aren't real people. However, unlike traditional personas, once created we were able to literally interact with them. This is a more than a bit mind-blowing.

We could ask them questions that we might want to ask in the future of our designer audience, such as "What might motivate you to purchase a group mentorship program?" and "What would you be willing to pay for such a program?"

David, for example, came back with a specific range of what he would be willing to pay. His answer may or may not reflect market value but, more importantly, David laid bare his decision-making process about how he came up with his answer.

He said he did market research, but also took a number of other criteria into consideration. David said he would pay more for a smaller group with more direct attention, networking opportunities, online resources, and would consider the experience level and reputation of the instructors.

When we told David about the real world qualifications that Kirk and Suzanne have, the bonus 1-on-1 private sessions, and the user-centric approach to topics that we plan to offer in our upcoming real life mentorship program, the price he would be willing to pay doubled!

From our perspective, while this type of interaction isn't the same as hearing it from actual people, it did give us lots of ideas for follow-up interview questions and a good reminder about the information to include on our mentorship program sales page!

Conclusion: To ChatGPT or Not To ChatGPT

The answer is obvious from our perspective. ChatGPT is a great contributor to the creative process in general and a valuable asset when creating user personas.

Like working with any human collaborator, it can remind you of the obvious, or point out things you haven't thought of yet. It's like having a (very knowledgeable) "sounding board" to bounce ideas off of. Or an ideation machine that can kick-start your own thought process.

Just remember, at the end of the day, ChatGPT cannot replace real research with real people. You still need to do your own research — ahead of time, to seed it properly; and after the fact, to vet what it tells you.

If you keep those caveats in mind, we believe you'll find ChatGPT to be a solid creative partner that can enhance your persona development process.