【June 2026 Update】Has Awareness of Reliability and Usage in Research Changed? Survey on Generative AI Trustworthiness Vol.3

Nile Inc. (headquartered in Shinagawa-ku, Tokyo; President and CEO: Tobi Takahashi; hereinafter "Nile"), which operates DX and marketing businesses, conducted a survey targeting 3,000 men and women nationwide on the actual usage of generative AI for research purposes.

This time, we compare results with the second survey conducted in October 2025 to examine how users' awareness and behavior regarding the use of generative AI for research have changed over the past year.

Summary of the survey

42% of respondents use generative AI for research, indicating that the previous rapid growth has plateaued. While usage is declining among younger generations, it is expanding among older age groups.

The most common approach to using generative AI and search engines is "using them depending on the content being researched." While sequential combined usage is spreading, the proportion of users who do not consciously differentiate between the two has doubled.

The most frequently asked topics on generative AI are "procedures and methods" and "unknown words." In contrast, usage remains low in areas such as recommendations, reviews, and travel, where personal experience is highly valued.

Approximately 80% of users verify generative AI responses, consistent with the previous survey. However, the proportion of users who "always verify" has decreased, indicating a slight relaxation in verification frequency.

Verification methods are overwhelmingly dominated by "search engines" (over 90%). Meanwhile, the number of users following reference links provided by generative AI has gradually increased since the previous survey.

Looking ahead, "generative AI" (30%) is the most desired tool for increased usage, surpassing "search engines" (20.1%). However, combined use with search engines is expected to continue.

Survey Overview

Survey period: May 27 to June 1, 2026

Survey method: Internet survey (using Freeasy)

Target respondents: 3,000 men and women aged 20–60 across Japan

Do you use generative AI when conducting research?

42% of respondents answered that they use generative AI when conducting research, slightly down from 43.5% in the previous survey, remaining nearly flat.

The adoption rate of generative AI for research had rapidly increased from 28.7% in the first survey (March 2025) to 43.5% in the second (October 2025). This time, the growth has stabilized, plateauing in the 40% range.

Despite awareness of generative AI, many users still rely primarily on traditional methods such as search engines. Alternatively, some users may have tried generative AI and concluded that other methods are more suitable for research, leading them to reduce usage.

When broken down by age group, previous surveys showed higher usage among younger generations—62.9% among those in their 20s and 49.5% among those in their 30s. This time, usage dropped to 46.3% among 20-year-olds and 45.2% among 30-year-olds.

This suggests that the group who tried generative AI and then reduced usage is growing, particularly among early-adopter younger users.

In contrast, users aged 40 and above showed higher adoption rates than in the previous survey, indicating that older generations, previously less engaged with generative AI, are beginning to adopt it.

How do you differentiate between generative AI and search engines?

"Do not use search engines" and "none of the above" were options only available in this survey.

The most common response, consistent with the previous survey, was "using them depending on the content being researched" (40.9%). This confirms that most users continue to differentiate based on the strengths and weaknesses of each tool.

However, there were changes in the breakdown.

The proportion of users who "search the same content on both and compare" decreased from 23.9% to 19.5%. In contrast, the proportion of users who "first search using a search engine, then organize information using generative AI" increased significantly from 8.9% to 19.5%.

Combined with the 18.0% who "first ask generative AI, then verify via search," this indicates a growing trend of sequentially combining both tools.

Notably, the proportion of users who "do not consciously differentiate" between the two increased from 7.8% to 18.3%, more than doubling.

While some users consciously differentiate, the increasing adoption across age groups may mean that many users have not yet established personal usage rules.

What types of content do you most frequently ask generative AI about?

The types of content users ask generative AI about have not significantly changed from the previous survey.

The most common response was "confirming procedures or methods" (53.1%), followed by "unknown words" (38.5%).

These two categories stand out, suggesting that users value generative AI for breaking down complex procedures or quickly explaining word meanings.

Other categories such as "hobbies" (24.2%), "news" (21.6%), and "IT knowledge" (20.0%) follow at around 20%, indicating broad usage across diverse topics without strong bias toward specific areas.

In contrast, usage remains low in areas such as "recommendations/reviews for stores, services, or products" (18.3%), "travel" (15.6%), and "beauty" (8.0%), where personal preferences and experiences are highly valued.

In these fields, users strongly desire authentic user voices and reviews, and may find AI-generated summaries insufficient.

When using generative AI, do you verify information using other sources?

When asked whether they verify (fact-check) generative AI responses using other sources, 28.1% answered "yes" and 50.8% answered "sometimes," totaling approximately 80% who perform some form of verification.

This 80% verification rate remains nearly unchanged from the previous survey. However, the breakdown has slightly shifted: the proportion of users who "always verify" decreased from 35.1% to 28.1%.

Combined with a slight increase in "sometimes," this suggests a trend toward reduced verification frequency.

One possible reason is the expanding user base observed in Q1. As generative AI adoption spreads across broader age groups—including those previously less familiar with it—awareness of the possibility of AI errors and the need for verification may not be fully established.

The increase in "never considered it" from 7.6% to 11.1% supports this interpretation.

How do you verify generative AI responses?

"Search engines" (91.6%) were the overwhelmingly dominant method for verifying generative AI responses.

As in the previous survey, most users who use generative AI for research still rely on search engines for final confirmation.

In contrast, "books" decreased from 26.8% to 11.9%, suggesting a preference for immediately accessible sources such as search engines and reference links.

A notable change is the increase in users following "reference page links displayed by generative AI," rising from 21.5% to 27.7%.

This indicates growing user engagement with the sources provided alongside generative AI responses, although