Wiley has released new findings showing that AI adoption among researchers has risen sharply while expectations about its capabilities have become more grounded and evidence-based. The company has shared results from its second ExplanAItions study, which indicates a significant course correction in how researchers perceive and use AI.
The comprehensive survey of more than 2,400 researchers worldwide has revealed that enthusiasm for AI remains strong, with 85% reporting that it has improved their efficiency and nearly three-quarters noting enhancements in both the quantity and quality of their work. Overall, the proportion of researchers using AI tools has increased from 57% in 2024 to 84% in 2025. Use of AI specifically for research and publication tasks has also grown markedly, from 45% to 62%.
At the same time, researchers have adjusted their expectations as hands-on experience has deepened their understanding of AI’s current limits. Whereas in 2024 researchers believed AI could outperform humans in more than half of potential use cases, this year that perception has declined to less than one-third.
Jay Flynn, Wiley’s Executive Vice President and General Manager of Research & Learning, has highlighted that the findings reflect a major shift in mindset as researchers’ growing experience with AI has prompted a recalibration of expectations. He has added that Wiley remains committed to equipping researchers with clear guidance and purpose-built tools to help them use AI effectively and confidently.
The study shows that as researchers gain more direct experience with AI, they have also developed a more cautious and informed perspective. Concerns about potential inaccuracies and hallucinations have risen from 51% to 64%, while privacy and security concerns have increased from 47% to 58%. This change signals a maturing research community that is balancing optimism with a realistic understanding of AI’s strengths and limitations.
Researchers continue to favor general-purpose AI tools over specialized ones for scientific tasks. Around 80% of respondents report using mainstream platforms such as ChatGPT, compared with only 25% using dedicated AI research assistants. Limited awareness contributes to this disparity—on average, only 11% of researchers have heard of the specialized tools surveyed. Most respondents (70%) rely on freely accessible tools, even though nearly half (48%) also have access to paid solutions. The findings suggest that many researchers are relying on a mix of free and paid tools to meet their research needs.
Corporate researchers have emerged as more confident and capable AI users, facing fewer barriers to successful implementation. They benefit from greater institutional access, with 58% reporting that their organizations provide AI tools, compared to 40% across all sectors. They are also less constrained by the absence of training or clear guidelines—44% cite these as barriers, compared to 57% among all respondents.
Corporate researchers are more likely to view AI as highly capable, believing it currently surpasses human performance in 50% of potential use cases, compared to the overall average of 30%. Collectively, these insights indicate that enhanced organizational support and access to AI resources can help researchers move from cautious experimentation to more productive integration of AI into their daily work.
The study underscores a continued disconnect between researcher needs and institutional support. Only 41% of respondents feel adequately supported by their organizations. To bridge this gap, a majority of researchers (73%) are turning to publishers for clear guidance and best practices to avoid potential pitfalls. More than half (57%) cite a lack of training and guidelines as a key barrier to expanding AI use.
Wiley has emphasized its ongoing commitment to supporting researchers as they adapt to the evolving AI landscape.
Click here to read the original press release.
