Lost in Translation: The Reality of Synthetic Research Tools in Japan and Korea

Synthetic Research has become one of the most discussed topics across the insight industry, with synthetic data, digital twins and AI personas as the typical examples. Since these methods utilize AI to generate consumer data, we as a panel provider have been paying attention to this area, given its potential implications for a shifting role of human data.

As definitions and terminology surrounding synthetic research are developing, we  wanted to understand how marketing and insights professionals in Japan and South Korea perceive synthetic research tools and how much they currently utilize them.

The Research

We surveyed professionals in Japan and Korea in marketing or consumer-insight roles

as well as executives with at least three years of experience and already aware of synthetic research tools.

The analysis covers:

・145 respondents in Japan
・155 respondents in Korea

from client-side roles (60 and 63%) and supplier-side roles at agencies, research firms, and consultancies (15 and 24%).

Because response tendencies can differ between the two markets, we focus on each country rather than on head-to-head comparison.

The Findings

“Not sure” - everyone has a slightly different understanding

All respondents were already aware of synthetic research tools. The level of understanding could be evenly divided into two groups: 48.3% in Japan and 54.8% in Korea said they understood them well or fairly well, versus 51.7% and 45.2% who said they have limited knowledge or have only heard of the names.

There was no shared picture of what the following tools actually mean. Interpretations spread widely with a notable share of "not sure" with none exceeding 50% for any tool. This implies that mainstream definitions are not yet standardized and might be defined differently across markets.

• Synthetic Data was interpreted as "AI-generated artificial data" most often in Japan (33.1%), but "a recreation of a specific individual" in Korea (49.7%). "Not sure" ran 26.2% and 10.3%.
• Digital Twin was the least understood term. "Not sure" was the top answer in Japan (43.4%); Korea’s interpretation mainly split across “a recreation of a specific individual(33.5%)", “a model mirroring the whole market (28.4%)”, “AI-generated artificial data (25.2%)”.
• AI Persona was split most evenly. Three interpretations tied at 20.0% each in Japan; "a recreation of a specific individual" led in Korea (30.3%). The percentage of “Not sure” is still significant in both markets (37.2% in Japan and 28.4% in Korea).

Leaning positive toward synthetic research tools, but with reservations

"It's hard to trust it 100% at this stage, but I think it's necessary if we want to see major scientific and economic progress."

Sentiment leans positive in both markets, with positive responses far outweighing negative ones. When asked about their future intent to use synthetic research tools in their work going forward, 44.1% in Japan and 80.6% in Korea said they were inclined to.

Trust in the output followed a similar pattern (47.9% and 82.4% positive).

In Japan, however, a "neither" group of around 39%, nearly as large as the positive group, points to a substantial wait-and-see segment.

Several respondents in both markets mentioned their concerns on accuracy and means of validation in the open-ended question asking about their current point of views, expectations or concerns. As one Korean respondent put it, "It's hard to trust it 100% at this stage, but I think it's necessary if we want to see major scientific and economic progress."

AI vs. human data, by dimension

Respondents also assessed synthetic research tools against human collected data across seven dimensions, including: 
・accuracy
・speed
・cost efficiency
・emotional and cultural nuance
・access to niche or rare segments
・privacy protection
・reproducibility

AI was rated favorably among all dimensions with speed drawing the widest margin (61.3% versus 4.2% and 61.8% versus 11.5% in Japan and Korea respectively).

Emotional and cultural nuance was the only exception where human data was valued higher in both markets (38.7% versus 27.7% in Japan; 42.7% versus 30.5% in Korea).

Adoption of synthetic research tools is still in its early stages

Current usage and intended use cases

Only 17.9% in Japan and 35.5% in Korea have actually used these tools, indicating that actual adoption of synthetic research tools remains limited.

When asked about preferred applications for synthetic research tools in their work, Marketing strategy planning and simulation led in both markets (44.8% in Japan, 61.3% in Korea), followed by consumer insight collection and analysis (38.6 and 42.6%), concept testing for products and services (24.1 and 32.3%) and competitive and market trend analysis (35.9 and 22.6%).

Timeline and conditions for adoption

A clear timeline emerges on how soon synthetic research tools could substitute human data in respondents' own work. Within one to three years, 63.0% in Japan and 67.2% in Korea expected some replacement to be practical; over three to five years, that rises to 74.8% and 76.3%, with full replacement jumping from 11.8% to 36.1% in Japan and 18.3% to 42.7% in Korea.

Respondents mentioned multiple reasons to accelerate future use. Accuracy proof compared to human data was the highest in both markets (49.0% in Japan, 56.8% in Korea) with clearer regulatory and ethical guidelines, successful case studies, lower cost, and tool usability close behind.

Why Real People Still Matter: An Asian Panel Provider’s Perspective

While Japan and Korea mainly use traditional research methodologies, respondents in these countries mentioned high usage intent of synthetic research tools in the near future. To motivate further adoption, there is an importance of human data as a source of validation as accuracy proof compared to human data was the most cited factor for adopting these tools.

Respondents in these countries continue to rate human data ahead on emotional and cultural nuance as well. If human data is to be the benchmark, researchers should understand the norms of each country differently. For example, in a high-context society in Japan, where people do not tend to express themselves directly. If responses are only interpreted on a surface level, the deeper insight might be missed.

Researchers must move beyond simply trusting AI outputs and adopt concrete validation frameworks on a country level. Teams should employ parallel testing ensuring that synthetic data should be analyzed together with human research to assess if the outputs are inline with the norms of that market.

With AI generating insights in a blink of an eye, incorporating high quality data collected from genuine respondents while keeping the cultural norms per market intact will matter even more than ever.