Step01
Synthesize Research in NotebookLM
Group the authorized research files by persona in NotebookLM, de-identify unnecessary participant details, and record the included source set, dates, and known sample limitations.
AdvancedProduct
How to Create Interactive AI Personas from Static Customer Research
Convert static customer research into clearly labeled synthetic personas whose responses stay grounded in cited source material, then use them for exploration without treating them as real customer validation.
Source episode · 24:00
How I AI: Automate Recruiting and Build Interactive Personas with Michal Peled of HoneyBookwith Claire Vo
Michal loads HoneyBook’s customer research into NotebookLM, asks it to create source-grounded persona instructions with citations, tightens the prompts to fit Custom GPT limits, and tests five interactive personas.
Before you start
What you need
- Authorized customer-research files grouped by persona
- NotebookLM with a defined source set
- A Custom GPT workspace and instruction limits
- A research owner who can validate persona claims
- Representative test questions and known evidence
What you’ll make
A set of disclosed synthetic persona assistants that answer from the research, cite supporting material, state uncertainty, and help teams explore hypotheses.
Tools used
- ClaudeVisitClaude
Anthropic AI assistant
- NotebookLM
Google's AI-powered research assistant
- ChatGPTVisitChatGPT
OpenAI conversational AI
Step by step
The workflow
Follow the sequence once, then adapt the prompts, checks, and handoffs to your own setup.
4 steps
Step02
Prompt NotebookLM to Act as a Prompt Engineer
Ask NotebookLM to produce one instruction draft per persona from the selected sources. Require citations for identity, needs, constraints, decision patterns, technology, journey, and communication preferences.
Example prompt
Create Custom GPT instructions for each research persona using only the selected sources. For every defining trait, include a source citation in the working notes. Cover context, goals, constraints, decision criteria, technology, journey, and communication preferences. Do not add a trait that is not stated or strongly implied; mark gaps as unknown.
Step03
Refine and Add Guardrails
Tighten each draft to the platform limit and add behavioral boundaries: stay in persona, answer only from supported evidence, disclose uncertainty, avoid sensitive speculation, and never present the simulation as a real customer.
Example prompt
Compress these instructions to under [limit] characters without removing evidence-backed distinctions. The assistant must state when research is insufficient, avoid political, religious, racial, gender, health, and other sensitive inference unless essential and explicitly supported, and remind users that it is a synthetic research persona.
Step04
Create and Test Your Custom GPTs
Create the Custom GPTs and test each with the same messaging, onboarding, pricing, and product questions. Compare answers with cited research, capture unsupported claims, and tell users when direct customer research is required.
Example prompt
Respond as this synthetic research persona using only its approved source material. Question: [question]. Give the answer, the research evidence or citation behind it, confidence, and what would need direct customer validation. Do not claim to be a real customer.
What good looks like
- Persona traits and answers trace back to the approved research set.
- The assistant refuses to invent beliefs when the source material is silent.
- Different personas produce meaningfully different, evidence-based responses.
- Teams know synthetic responses are prompts for research, not replacement customer evidence.
Build your next product with ChatPRD
Turn an idea into a PRD, user stories, and a plan.
After the steps
Runbook notes
How to recover when the loop fails and where human judgment helps.
Recover
If it goes sideways
- Evidence from several segments is blended into one inconsistent persona
- Separate source sets, label overlaps, and require every defining trait to cite the intended segment.
- The persona fills research gaps with plausible stereotypes
- Instruct it to say the evidence is insufficient and ask the research owner to add or clarify sources.
- Teams quote the synthetic persona as if a customer said it
- Label every interface and export as synthetic, preserve citations, and route material decisions to direct customer validation.
- Research files or outputs expose personal or sensitive participant data
- Minimize and de-identify inputs, restrict access, and remove details not needed for the persona’s decision context.
Keep building
Related workflows
beginnerPersonalClaude
Create an Automated Workout Tracker with a Simple Claude Voice Note
Log your workouts effortlessly without ever opening a spreadsheet. This workflow uses the Claude mobile app to parse a simple voice note and automatically update your workout tracking spreadsheet in the background.
Open
advancedOperationsClaude
How to Rebuild Your Gmail Inbox Inside Claude to Manage Email
Create a custom email management interface within Claude to read, triage, and draft replies without opening the Gmail tab. This workflow helps you reclaim focus and compound your learnings by training your AI on every interaction.
Open
intermediateSalesClaude
Automate Client Proposals and Onboarding with a Claude 'Pipeline Operator'
Build a fully automated system in Claude to generate personalized, interactive HTML proposals and client onboarding materials, saving hours on administrative work and creating a high-touch experience.
Open
Start shipping
better products.
Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.
Free to start
No credit card
SOC 2 certified
Enterprise ready