Wiley’s 2025 survey reported that 84% of researchers used AI, while Elsevier’s 2025 survey reported that 58% used AI for work. Different samples and definitions explain why these figures should stay separate. Researchers reported using AI to find and summarize papers, review literature, draft documents and analyze data, not necessarily to replace scientific judgment.
TL;DR
Wiley and Elsevier measured AI use differently in 2025; their adoption figures are not interchangeable.
Elsevier’s listed uses start with finding and summarizing research, followed by literature reviews.
Researchers want clear citations, current evidence, accurate answers and human checks.
Choose AI research tools that let you inspect sources, protect permitted uploads and export usable references.
How many researchers use AI
AI use is common in both surveys, but each measures it differently. Wiley’s ExplanAItions survey asked 2,430 researchers about AI in August 2025. Its October 7, 2025 announcement reports 84% using AI overall and 62% using it for research or publication tasks.
Elsevier’s Researcher of the Future survey covered more than 3,200 academic and corporate researchers across 113 countries. Its November 4, 2025 release reports 58% using AI tools in their work. The headline rounds the sample to 3,000; the release’s body gives the fuller description.
Wiley, ExplanAItions 2025
Sample: 2,430 researchers
Survey and release dates: Survey: August 2025. Release: October 7, 2025.
Headline findings: 84% overall AI use; 62% research or publication use.
Elsevier, Researcher of the Future 2025
Sample: More than 3,200 academic and corporate researchers across 113 countries
Survey and release dates: Release: November 4, 2025. The release does not specify fieldwork dates.
Headline findings: 58% using AI tools for work.
The distinction is straightforward: overall AI use is broader than AI use at work. The surveys also draw on different samples, so even their work-related figures are not directly comparable. Do not average them into a global adoption rate or label them as 2026 survey results.
Neither release identifies who answered every question behind every percentage. This guide therefore attributes each finding to its survey rather than assuming every task figure describes “AI users.” If you are new to the category, the AI research assistants explainer explains the workflow; the survey releases supply the adoption evidence.
What researchers use AI for
Once researchers adopt AI, what do they do with it? Finding and summarizing research leads the activities listed in Elsevier’s 2025 release, followed by literature reviews. Drafting and data analysis also appear in the findings.
Find and summarize the latest research: 61%
Perform literature reviews: 51%
Draft grant proposals: 41%
Analyze research data: 38%
Draft research papers or reports: 38%
These figures describe separate activities, not shares of a single total. Do not add them together or assume they use the same response base as the overall work-use figure.
Wiley reports a broader category: 62% used AI for research or publication tasks. That finding does not isolate literature reviews, so it cannot serve as a direct comparison with Elsevier’s literature-review figure.
Together, the findings describe where researchers bring AI into their work. They do not show whether the resulting work is correct. You still need to compare a summary with the paper, check a draft’s claims and validate an analysis’s inputs and methods.
General chatbots or research assistants
The task is only part of the story; the type of tool matters too. Wiley’s 2025 survey reports 80% using mainstream AI tools compared with 25% using AI research assistants. Researchers can use both, so these are not competing shares of one total.
Wiley also reports that only 11% of researchers, on average, had heard of the specialized tools surveyed. “On average” refers to awareness across those tools. It does not mean only that share had heard of any research assistant at all.
This awareness finding adds context to the use figures, but it does not explain every researcher’s choice. Neither popularity nor familiarity proves that a tool finds better papers or produces more accurate answers. To judge those qualities, you need to inspect its work.
The trust gap: accuracy, hallucinations and privacy
Researchers can use AI and still question its reliability. In Wiley’s 2025 survey, 64% reported concerns about inaccuracies and hallucinations, while 58% raised privacy and security concerns. A hallucination is false or unsupported content presented as an answer. These percentages measure concern, not actual error or security-incident rates.
Elsevier reports a similar gap between adoption and confidence: 22% considered AI tools trustworthy, compared with 39% who called them unreliable. Those responses describe attitudes, not a test of any product’s accuracy. The release does not specify how to classify the remaining respondents.
For your own work, separate evidence checking from data handling. First, open the cited paper and compare its passage with the answer. Then ask whether your institution, contract or consent arrangements allow you to send the material to that service.
A well-sourced answer does not make every upload appropriate. Before sharing unpublished or restricted material, check who processes it, how long the service retains it and which rules apply to your project.
What would make AI tools more trustworthy
The trust findings lead to a practical question: what would increase researchers’ confidence? Elsevier’s respondents point to citations, current scholarly evidence, accuracy safeguards and human review. Its 2025 release lists these factors:
Transparency
Automatically cites references
Reported finding: 59%
Recency
Training data includes the most up-to-date scholarly literature
Reported finding: 55%
Safety
Explicitly trained for factual accuracy and safety
Reported finding: 55%
Quality
Trained on high-quality peer-reviewed content
Reported finding: 55%
Validation
Outputs regularly reviewed by human experts
Reported finding: 49%
These are requested safeguards, not proof that a product provides them. A citation can identify a real paper without supporting the sentence attached to it. Likewise, a vendor’s training claim needs evidence before you can treat it as a verified quality measure.
Researchers also want guidance on using AI responsibly. Wiley reports that 73% looked to publishers for clear guidelines and help avoiding pitfalls. That demand does not imply that publishers and institutions follow one policy.
Before starting a project, check which tasks your institution and target publication allow, what you can upload and what assistance you must disclose. Keep the source passages behind your claims so a colleague can follow the same evidence. Clear rules and checkable sources address different parts of the trust problem.
How to choose an AI research tool
Choose a tool by the work it helps you check, not by how many researchers use its category. Neither survey names a best-overall product. Start with a question and papers you know, then test how the tool handles them.
Search scholarly papers. Check whether it retrieves relevant studies, including familiar papers. A large collection does not guarantee coverage of your topic.
Check citations for each claim. Look for evidence behind each factual statement, not just a reference list at the end.
Open the supporting passage. Compare the population, outcomes, figures and qualifications with the answer’s wording.
Test reference exports. Confirm that exported records work in your reference manager and retain accurate details.
Use only material you have permission to share during the test. A useful result does not settle whether the upload met your data-handling obligations.
AnswerThis is an AI research assistant for researchers who want cited literature search, passage-linked answers and reference-manager exports. Its citations jump to the exact passage. The Free plan lists search across more than 250 million papers, cited answers and summaries, projects to organize your papers, and no card is required; Pro adds unlimited basic queries, systematic reviews and the AI Writer.
These are AnswerThis capabilities and plan terms, not Wiley or Elsevier findings.
Use the comparison of AI tools for research to explore options. To practice checking an answer against a document, try Chat with PDF and read the cited passage in context. Keep a tool only if you can verify its evidence and use its outputs in your research workflow.
FAQs
What would make researchers trust AI tools?
Elsevier’s 2025 Researcher of the Future survey identifies references, current scholarly literature, factual accuracy and safety, peer-reviewed training content and human validation as confidence factors. Its release covers more than 3,200 researchers across 113 countries. Wiley’s 2025 ExplanAItions survey of 2,430 researchers also highlights demand for publisher guidelines.
Is AI reliable for research?
These surveys measure reported use and attitudes, not the accuracy of individual answers. Assess reliability by checking each claim against its source passage, appraising the study itself and following your institution’s data and disclosure rules. A citation alone does not establish that the source supports the claim.
Do researchers use general chatbots or AI research assistants?
Wiley’s 2025 ExplanAItions survey of 2,430 researchers reported 80% using mainstream AI tools compared with 25% using AI research assistants. Researchers can use both categories, and the figures do not establish which performs better.
Which AI is best for research?
The best fit is a tool that supports your research task, makes source evidence inspectable and meets your data-handling requirements. Test scholarly search, claim-level citations, passage access and exports with papers you know. Neither Wiley’s nor Elsevier’s survey establishes a best-overall tool.
Is there a free AI for research?
Yes. AnswerThis’s Free plan includes scholarly search across more than 250 million papers, cited answers and summaries, and projects to organize your papers, with no card required. Pro adds unlimited basic queries, systematic reviews and the AI Writer.
Survey sources for the FAQ: Wiley’s ExplanAItions release and Elsevier’s Researcher of the Future release.
Disclosure: AnswerThis publishes this guide. Wiley and Elsevier conducted their own surveys; this guide does not imply a partnership or endorsement.
