AI-Driven Persona Generator

Context

As AI tools for UX research became widely available, persona generators quickly emerged as a common shortcut. While efficient, most outputs were structurally shallow, producing personas that were serviceable for documentation but ineffective for strategic design decision-making.

At the same time, I was actively deepening my understanding of AI prompt engineering and exploring how large language models could augment research workflows beyond surface-level automation.

This project originated as a self-initiated investigation into whether AI could support meaningful persona construction rather than generic synthesis.

Laptop screen displaying a sidebar with various AI and GPT tools, including ChatGPT, Sora, Video GPT by VEED, Web Browser, Presentation & Diagrams, Presentation Slides, and others, in a dark mode interface.

The Process

Smartphone screen displaying a chat conversation about creating a persona, with questions about demographics, location, education, marital status, children, language, and income, and a response indicating the person is about 28 years old.

01 / Problem Framing

Most AI-generated personas failed in three critical ways:

  • Homogenization. Outputs converged toward predictable archetypes.

  • Lack of narrative depth. Personas lacked backstory, motivation, and internal tension.

  • Low engagement value. Teams struggled to emotionally or cognitively connect with them.

As a result, personas functioned as compliance artifacts rather than decision-shaping tools.

The core question became:

How might AI-generated personas achieve human-level specificity and narrative richness without sacrificing scalability?

Screenshot of a mobile phone displaying a list of interview questions about a person's professional background, job experience, and skills.

02 / Approach

  • Research narrative psychology and character construction

    • Focused on how humans perceive uniqueness and credibility

  • Designed a structured question framework

    • Questions intentionally targeted:

      • Personal history

      • Contradictions and tensions

      • Motivations and fears

      • Contextual constraints

  • Applied prompt engineering principles

    • Sequencing

    • Constraint layering

    • Context preservation

  • Iterated through live testing

    • Compared outputs against standard persona generators

A smartphone screen displaying a list of technical questions related to digital habits and social media usage.

03 / Outcomes

Generated personas that were:

  • Distinct, memorable, and internally consistent

  • More engaging for design critique and ideation

  • Better aligned with empathy-driven UX practices

  • Validated that prompt structure directly influences perceived human realism

  • Established a repeatable method for using AI as a research amplifier, not a shortcut

While not client-facing, this work demonstrated a transferable system applicable to:

  • UX research

  • Product discovery

  • Early-stage concept validation

  • AI-assisted DesignOps workflows

Why This Matters

Screenshots of a mobile app displaying a detailed profile of Olivia Carter, a UX designer, including sections on her persona, demographics, goals, fears, and final thoughts.

This project is not about personas.
It is about how AI should be integrated into design practice.

It demonstrates that I:

  • Treat AI as a design material, not a magic box

  • Understand how to translate abstract capability into practical frameworks

  • Can operate in ambiguity and self-direct learning into usable systems

  • Anticipate second-order effects of tooling on team behavior and decision quality

For organizations adopting AI-forward workflows, this reflects a readiness to lead transformation rather than
follow trends. See the GPT here »