Synthetic users vs digital twins: why synthetic users are gaining ground in market research on Terapage

Synthetic Users vs. Digital Twins: Why Synthetic Users Are Gaining Ground in Market Research

Key takeaways

  • Synthetic users are independent AI personas built from selected attributes; a digital twin is an AI copy of one real person, built from that person’s own data.
  • A digital twin asks “what would this customer do?”, while a synthetic user asks “what would this kind of customer do?”
  • Synthetic users win on speed, scale, diversity, new products, privacy and cost, because they need no personal data and no recruitment.
  • In a real Terapage study, 20 synthetic shoppers kept a week-long journal about a brand-new AI shopping assistant that had no real users to copy.
  • Use digital twins to go deeper with customers you know, synthetic users to explore people you haven’t met, and validate key findings with real people.

Every product, retail and finance team now faces the same research problem. They need answers from diverse audiences before a launch, a campaign or a price change, but recruiting real participants across cities, ages and incomes can take weeks. Digital twins promise a shortcut, yet each twin still needs months of one real person’s data before it can answer a single question.

That’s why more research teams are turning to synthetic users, AI personas that represent realistic consumers without copying anyone real. With Terapage Synthetic Users, you can build a diverse synthetic audience in minutes and test ideas the same day. In this guide, we explain how synthetic users and digital twins differ and show, with a real study, why synthetic users are gaining ground.

What Are Synthetic Users?

Synthetic users are independent AI personas created from synthetic data and selected attributes such as age, location, income, habits and personality. They are not copies of real people, and they respond to research questions the way a realistic consumer with those traits would.

Groups of synthetic users form synthetic audiences, and each individual profile within them is a synthetic persona. Researchers choose the attributes, and AI creates a varied group that reflects the target market. Learn the basics in our guide to synthetic data for researchers, or see how behavioural and psychological attributes make synthetic personas more realistic.

A synthetic audience of 20 synthetic personas with different ages, locations and backgrounds in Terapage
Figure 1: A synthetic audience of 20 synthetic personas with different backgrounds, ready to take part in the same study.

What Is a Digital Twin in Market Research?

A digital twin is an AI copy of one real person, built from that person’s own data. IBM describes a digital twin as a virtual representation of an object or system that is updated with real-time data.

Think of it this way: a digital twin of you would think and shop like you. It would know your favourite brands, your usual budget and the time you normally order groceries, because it learned all of this from your real purchases, survey answers and behaviour.

This makes digital twins powerful for studying known customers in depth, but every twin needs a real person behind it. McKinsey notes that digital twins depend on rich, connected data to stay accurate.

Synthetic Users vs. Digital Twins: What’s the Difference?

Synthetic users and digital twins both use AI to stand in for consumers, so they are often treated as the same thing. They are not interchangeable, because they start from different places and answer different research questions. A digital twin asks “what would this customer do?”, while a synthetic user asks “what would this kind of customer do?”

Synthetic users vs. digital twins: 11 research factors compared
Research factorSynthetic usersDigital twins
Who it representsA realistic profile, such as a budget-conscious parent in Texas, who is not one real personA specific customer who already exists in your CRM or panel
Where it startsA research brief, ready-made research templates and a mix of attributes you chooseMonths of one person’s purchases, survey answers and app activity
Time to first insightSame day, as personas are ready in minutes and respond straight awayWeeks or months, because data must be collected and connected first through data integrations
Sample sizeGrows with one setting, from 20 personas today to 200 tomorrowGrows one real person at a time, each needing their own data
New markets and segmentsCan explore a city, age group or lifestyle you have never sold to, from technology and telecoms to healthcareCan’t model people you hold no data for
Products not yet launchedCan react to a concept, prototype or app with zero usersRelies on past behaviour, so it struggles with something completely new
Personal dataNone, as personas are built from attributes, not individualsDetailed personal data, which raises consent and data privacy questions
CostLow, using AI credits instead of recruitment fees and incentivesHigh, with ongoing data collection and upkeep
Accuracy depends onHow well the attributes and context files reflect the real market, supported by transparent AIHow complete and up to date the person’s data is
Best fit in the research cycleEarly stages: exploring ideas, screening concepts and shaping questions before participant recruitmentLater stages: predicting how loyal or high-value customers respond to a change
Biggest riskPlausible answers that drift from reality if never validated with real peopleOutdated data that makes the twin behave like the customer used to, not as they do now

In short, digital twins help you understand the customers you already know, and synthetic users help you explore the ones you haven’t met yet.

Why Are Synthetic Users Gaining Ground in Market Research?

Most research questions are about people a brand doesn’t fully know yet. A digital twin can only answer for customers already in the data, while synthetic users can answer for anyone you can describe. Here is why more teams are choosing them.

  1. Speed: Synthetic personas are ready the same day, helping enterprise teams and research agencies meet launch deadlines.
  2. Scale: Synthetic audiences grow from 20 to 200 personas with one setting, making larger qualitative and quantitative studies easy.
  3. Diversity: Synthetic users reach new audiences across consumer goods, financial services and the public sector.
  4. New products: Synthetic users can test ideas with no real customers yet, ideal for concept evaluation and product testing and de-risking a launch.
  5. Privacy: Synthetic personas are built from attributes, not real people, on a secure, compliant platform.
  6. Cost: No recruitment or incentives to pay, so check the incentive calculator and test ideas within your plan.

How We Built Synthetic Users for an AI Shopping Assistant Study

To put these advantages to the test, we ran an open journal study on Terapage called the AI Shopping Assistant Adoption Study. Synthetic shoppers across the US were created for the study and recorded their experience over a week. Here is how the study was built in four steps.

Step 1: Set Up the Study

We named the study and added a cover image, then described the research in plain language. AI turned that short description into a full study set-up, including the journal questions, which we could edit before moving on.

Basic Settings in Terapage naming the AI Shopping Assistant Adoption Study and adding a cover image
Figure 2: Naming the AI Shopping Assistant Adoption Study and adding a cover image in Basic Settings.
Researcher describing a synthetic user study in plain language so Terapage AI suggests the full research set-up
Figure 3: Describing the study in plain language so AI can suggest the full research set-up.

Step 2: Build the Synthetic Audience

Next, we chose who the synthetic users should be. Terapage offers 13 attribute categories, from demographics and location to lifestyle and personality, so the synthetic audience reflects a realistic mix of US shoppers rather than one average customer.

Selecting synthetic persona attribute categories and the number of personas in Terapage, with estimated token usage
Figure 4: Selecting attribute categories and the number of synthetic personas, with the estimated token usage shown before launch.

Step 3: Ground Synthetic Personas With Study Data, Not Personal Data

Instead of feeding in real customer records, as a digital twin would need, we uploaded a study context file describing the app, its features, pricing and the shopping situations to explore. This keeps responses realistic without copying anyone’s personal data.

Uploading a study context file in Additional Data to ground synthetic users in the research scenario
Figure 5: Uploading a study context file in Additional Data to ground synthetic users in the research scenario.

Step 4: Review and Launch

Finally, we reviewed the full set-up, from activity details to persona attributes, and launched the study. With no participants to invite, the synthetic personas started responding straight away.

Review screen in Terapage with the activity overview and synthetic persona attributes before creating the study
Figure 6: Reviewing the activity overview and synthetic persona attributes before creating the study.
Synthetic activity created in Terapage with 20 AI personas and no participants to recruit
Figure 7: The synthetic activity created with 20 AI personas and no participants to recruit.

Testing a Product With No Real Customers

This is the situation where digital twins can’t help. The AI shopping assistant in our study was brand new and had no real users, so there was no one to copy. Synthetic users, however, could still show how it might fit into everyday life, from comparing prices to deciding which brands to trust.

Tracking Synthetic Users Through a Week-Long Journal

All 20 synthetic users were tracked in one dashboard, with their status and latest submissions visible in one place. There were no scheduling reminders, drop-outs or missed entries to chase.

Study dashboard in Terapage tracking 20 synthetic users and their journal submissions
Figure 8: Tracking all 20 synthetic users and their journal submissions in one study dashboard.

Daily Entries in Text, Photos, Video and Voice

Just like real participants, synthetic users recorded shopping moments in written diaries and shared photo, video and voice responses. By the end of the week, the study held 60 media files alongside the written entries.

Synthetic users’ written diary entries and photos about shopping with a new AI shopping assistant
Figure 9: Synthetic users sharing written diary entries and photos about shopping with a new AI shopping assistant.
Video and voice note journal entries from synthetic shoppers using an AI shopping app in everyday situations
Figure 10: Video and voice note entries showing how synthetic shoppers used the app in everyday situations.
Photos shared by synthetic users of everyday shopping moments, from farmers’ markets to eco-friendly product shelves
Figure 11: Photos shared by synthetic users, capturing everyday shopping moments from farmers’ markets to eco-friendly product shelves.
Video journal entries recorded by synthetic users describing their experience with a new AI shopping assistant
Figure 12: Video entries recorded by synthetic users, showing how each persona described their experience with a new AI shopping assistant.

What Synthetic Users Revealed About AI Shopping Assistants

Terapage’s AI-powered insights turned a week of synthetic entries into sentiment, topics and themes within minutes. This early view of adoption is the kind of input that helps teams build products people will buy.

How Did Synthetic Users Feel About AI Recommendations?

Sentiment analysis groups responses by mood and shows how strong each feeling is. Feelings were mostly positive, led by satisfaction and excitement, and trust grew when the app filtered products by what shoppers cared about. A result this positive should still be checked with real shoppers.

Sentiment analysis chart of moods synthetic users expressed toward AI shopping recommendations, led by satisfaction and excitement
Figure 13: Sentiment analysis showing the moods synthetic users expressed toward AI shopping recommendations.
Key sentiments in synthetic users’ own words ranked by emotion, strength and weight
Figure 14: Key sentiments in synthetic users’ own words, ranked by emotion, strength and weight.

What Mattered Most to Synthetic Shoppers?

Synthetic shoppers valued brands that matched their values over the cheapest price, an early sign of the types of consumer insights that drive adoption.

AI excerpts ranking topics synthetic shoppers discussed most, led by sustainability and personalised experience
Figure 15: AI Excerpts ranking the topics synthetic shoppers discussed most, led by sustainability and personalised experience.

The Themes Behind Trust and Adoption

Quotes analysis ranks themes by how often they appear in synthetic users’ own words. Shoppers trusted the app when it helped them shop the way they already wanted to, a “why” that is hard to get from a digital twin and a natural next step for a consumer empathy study with real people.

Quotes analysis ranking the themes behind trust and adoption by how often synthetic users mentioned them
Figure 16: Quotes analysis ranking the themes behind trust and adoption by how often synthetic users mentioned them.

Comparing Views Across Synthetic Audience Segments

Demographic distribution shows whether a theme is shared across the whole synthetic audience or held mainly by one group. This helps researchers spot which segments to explore further before running real fieldwork.

Demographic distribution comparing shopping themes across male, female and other synthetic users
Figure 17: Demographic distribution comparing how shopping themes were shared across male, female and other synthetic users.

The Words Synthetic Shoppers Used Most

The word cloud shows the words synthetic shoppers used most, each linked back to its responses, so AI-driven data analysis is easy to check.

Word cloud data listing the most frequent words in synthetic shoppers’ journal entries
Figure 18: Word cloud data listing the most frequent words in synthetic shoppers’ journal entries.
Journal responses with a frequent word highlighted, showing the context behind the word cloud
Figure 19: Journal responses with a frequent word highlighted, showing the context behind the word cloud.

Limitations: Why Synthetic Users Don’t Replace Real Research

Synthetic users are best for exploring ideas, spotting patterns and narrowing options early. They can reflect bias in their training data and may miss real-world surprises, and a strongly positive result like ours is exactly the kind of finding to check with real shoppers.

That’s why key findings should be validated with real participants through hybrid research. Our framework for using synthetic data responsibly explains how to combine both.

When Should You Use Digital Twins vs. Synthetic Users?

Digital twins help you go deeper with customers you already know. They are built from purchase history and survey answers from consenting customers, so they suit questions like how loyal subscribers will react to a price rise, which members might cancel, or which offers to personalise for your best customers. Their limit is that they only cover people already in your data.

Synthetic users help you explore people you haven’t met yet. They are AI personas built from target audience traits and a research brief, so they suit testing an app before it has users, exploring demand in a new city or hearing from a new age group. Synthetic audiences can represent anyone you can describe, and you can hear from them in a single day. Whichever you choose, validate key findings with real people before making the final decision.

Infographic on when to use digital twins for existing customers and synthetic users for new audiences, markets and products
Figure 20: When to use digital twins to go deeper with existing customers, and synthetic users to explore new audiences, markets and products.

Run Synthetic User Research on Terapage

Terapage brings synthetic users and real participants together on one platform. Build synthetic audiences in minutes, run journals and AI-moderated interviews, analyse every response with AI-powered reporting, and move straight to real people or a long-term insight community when you are ready to validate.

Terapage’s five research products, including Synthetic Users and Data, with shared platform features
Figure 21: Terapage, an AI-powered market research platform that brings synthetic users and real participants together.

Start your free trial or request a demo to run your first synthetic user study.

Ready to run your first synthetic user study?

No credit card required. Need help designing your study? Our Co-pilot Research Services team can help.

Frequently Asked Questions About Synthetic Users and Digital Twins

What is the difference between synthetic users and digital twins?
A digital twin is an AI copy of one real person, built from that person's own data. Synthetic users are independent AI personas built from selected attributes, not copies of anyone real. A digital twin asks “what would this customer do?”, while a synthetic user asks “what would this kind of customer do?”
What are synthetic users in market research?
Synthetic users are independent AI personas created from synthetic data and selected attributes such as age, location, income, habits and personality. They respond to research questions the way a realistic consumer with those traits would. Groups of synthetic users form synthetic audiences.
What is a digital twin in market research?
A digital twin is an AI copy of one real person, built from that person's purchases, survey answers and behaviour. It is powerful for studying known customers in depth, but every twin needs a real person behind it and rich, connected data to stay accurate.
When should you use digital twins instead of synthetic users?
Use digital twins to go deeper with customers you already know, such as how loyal subscribers will react to a price rise or which members might cancel. Use synthetic users to explore people you haven't met yet, like testing an app with no users or exploring a new city or age group.
Can synthetic users test a product that has no customers yet?
Yes. Because synthetic users are built from attributes rather than past behaviour, they can react to a concept, prototype or app with zero users. In Terapage's AI Shopping Assistant Adoption Study, 20 synthetic shoppers kept a week-long journal about a brand-new app.
Do synthetic users use personal data?
No. Synthetic personas are built from attributes, not individuals, and grounded with study context files such as a description of the app and its pricing. Digital twins, by contrast, need detailed personal data, which raises consent and data privacy questions.
How quickly can you get insights from synthetic users?
The same day. On Terapage, synthetic personas are ready in minutes and respond straight away, with no recruitment, scheduling or drop-outs. Digital twins usually take weeks or months because each person's data must be collected and connected first.
Do synthetic users replace research with real people?
No. Synthetic users are best for exploring ideas, spotting patterns and narrowing options early. They can reflect bias in their training data and miss real-world surprises, so key findings should be validated with real participants through hybrid research.