Methodology
What synthetic audiences are, how Tesemble uses them, and where they stop.
This matters more for Tesemble than for most software. Here is exactly what we claim, and what we don't.
Methodology
Simulation, not synthetic certainty.
Tesemble is built for generating and prioritizing hypotheses. It does not replace conversations with real customers.
Synthetic profiles are not real people. They give you a structured, repeatable way to compare how a defined audience is likely to react to competing ideas, so the expensive part of GTM, talking to real customers, starts from a shorter list.
What we're building toward
- Confidence ranges
- Repeated simulations
- Prompt and version tracking
- Model diversity
- Experiment reproducibility
- Calibration against real-world outcomes
How a simulation runs
- 1Input context
- 2Audience definition
- 3Controlled messages
- 4Many simulated responses
- 5Aggregation
- 6Segment comparison
- 7Prioritized hypotheses
Why not just use Claude or ChatGPT?
A simulation is more than asking an LLM a question 100 times.
You can run the same method yourself. We open-sourced it as icp-gtm-sim: a free skill for Claude, ChatGPT, and Codex that builds a defined audience and reports segment-level patterns.
Tesemble runs that workflow when you need it often — with visual reports, comparisons, and faster runs at scale.
Get icp-gtm-sim on GitHubInstall guides: Claude · ChatGPT & Codex
Can I run this in Claude or ChatGPT instead?
What is a synthetic audience?
Are synthetic profiles real people?
Does Tesemble replace customer interviews?
Where do the profiles come from?
How accurate are simulations?
Can I create my own audience?
Can I upload my own customer research?
Which AI models does Tesemble use?
Is my company data used to train models?
Can I compare several messages at once?
You've got more GTM ideas than you can test.
Test them with Tesemble first.
Try for free10,000 free credits at launch. No credit card.