Skip to content
What is a synthetic persona?
A synthetic persona is a decision-oriented profile generated by artificial intelligence using real data, designed to simulate the behavior, objections, and reasoning of a target segment.
Unlike the traditional marketing persona — often static and descriptive — a synthetic persona is interactive. It can be questioned, confronted with an offer, exposed to a value proposition or a strategic shift, and produce responses consistent with the segment it represents.
When properly built from CRM data, insights derived from the sales cycle, and behaviors observed in analytics, it becomes a strategic exploration tool.
Why the concept is gaining maturity
The evolution of large language models has made it possible to move beyond content generation. It is now possible to simulate:
- purchase objections
- reactions to repositioning
- budget trade-offs
- internal tensions between marketing, finance, and leadership
Recognized players in UX and marketing research have begun documenting the phenomenon.
Nielsen Norman Group: synthetic users
Nielsen Norman Group published a detailed article on “synthetic users” and their use in UX research. Their position is clear: these profiles are useful in exploratory phases but do not replace real users.
Source: https://www.nngroup.com/articles/synthetic-users/
NielsenIQ: synthetic respondents
NielsenIQ refers to “synthetic respondents” to describe AI-simulated respondents in quantitative research. The central issue raised is methodological: validity depends on bias control and statistical rigor.
Source:
The rise of synthetic respondents in market research:
Bluetext: marketing application
Bluetext documents how synthetic personas can accelerate customer research and test messaging before launch.
Source:
Synthetic Personas: How AI-Generated User Models Are Changing Customer Research
Academic research (ACM)
A paper published through ACM evaluates the use of large language models for generating synthetic personas. It highlights, among other concerns, the risk of bias amplification if no methodological framework is applied.
Source: https://dl.acm.org/doi/10.1145/3750069.3750142
Traditional persona vs synthetic persona
| Criteria | Traditional persona | Synthetic persona |
|---|---|---|
| Nature | Descriptive document | Interactive model |
| AI-based | No | Yes |
| Simulation capability | No | Yes |
How a synthetic persona is built
1. Structured data
- CRM, sales cycles, closing rates
- GA4 and behavioral data
- Call summaries
- Industry studies
- Competitive analysis
2. Methodological framework
Prompt engineering must incorporate a precise segment, company size, digital maturity, budget constraints, and decision-making structure.
3. Calibration
Responses must be confronted with conversion rates, customer acquisition cost (CAC), sales cycles, and observed friction points.
Advanced B2B example
Simulate a Director of Operations at a B2B manufacturing company with 120 employees facing margin pressure. They must choose between industrial automation investment and commercial expansion into a new North American market.
This type of simulation helps test the credibility of projected ROI, the robustness of an expansion plan, and objections related to operational risk.
Structural limitations
Synthetic personas do not replace real research. They accelerate exploration but must be validated with real-world data.
Bias can be amplified if the model is not properly structured and controlled.
Conclusion
The synthetic persona is an advanced exploration tool at the intersection of marketing, advanced data analysis, and artificial intelligence.
Used with discipline, it becomes a strategic lever. Used without structure, it becomes an attractive but fragile artifact.
