What the Clarivate 2026 Library Survey Really Says About AI Gaps

In Clarivate's Pulse of the Library 2026 news release, 1,876 surveyed librarians report sharply uneven AI adoption: 28% in Mainland China, 16% globally, and 11% in the U.S. The headline comes from a vendor that sells library AI products through its Academic AI line, so the numbers warrant a critical read rather than a simple regional scoreboard.

Reading the gap responsibly means checking what the survey counted as adoption, why library leaders say they lag, and what AI implementation actually costs before mapping it onto MLIS curriculum. For libraries and LIS programs, the harder question is whether librarian professional development can keep pace with vendor tooling.

What Is the Clarivate Pulse of the Library Report?

Clarivate runs the Pulse of the Library survey in partnership with Library Journal, and the 2026 edition is the third year of the series. It surveys academic, public, and national library staff and leaders worldwide to track technology adoption, budgets, and priorities. This year's findings center on AI adoption.

Survey basics

After data cleaning, the 2026 analysis included 1,876 responses. The online survey ran from April 22 to May 18, 2026 and was promoted through email, listservs, website pop-ups, and social media. Respondents could take it in English, Mandarin, Korean, or Japanese. U.S. respondents made up 57% of the cleaned sample, up from 46% in 2025, so the results lean heavily toward U.S. institutions even though the report highlights global gaps.

Qualitative interviews

Clarivate supplemented the questionnaire with 20 qualitative interviews. These conversations add context to regional differences, especially around AI adoption and perceived barriers.

Who publishes it

The report comes from Clarivate's Academia & Government division, the business unit behind Ex Libris, ProQuest, and Web of Science. The report is both research and vendor insight, since the publisher sells AI-related library products. That publisher relationship shapes the AI adoption findings.

Where to read it

Clarivate published the September 9, 2026 announcement on its news page, and the full report is available on Clarivate's Pulse of the Library page.

Regional AI Adoption Gaps: Mainland China, the U.S., and Beyond

Clarivate's Pulse of the Library 2026 survey of 1,876 librarians worldwide puts Mainland China at the front of AI adoption in libraries, reporting 28% implementation against a 16% global figure, while the U.S. trails at 11% and is described as the most cautious region. Standalone adoption percentages for Europe and other Asia-Pacific countries are not stated in the available source excerpt. The sample was heavily U.S.-weighted: 57% U.S. respondents, 5.3% Mainland China, 4% U.K., 3.5% India, 2.7% Australia, and 27.5% other regions.

How to Read the Data: Sampling, Definitions, and Vendor Bias

How to read this data starts with a simple question: who actually answered? The Pulse of the Library 2026 survey collected 1,876 usable responses and 20 qualitative interviews between April 22 and May 18, 2026. Academic libraries made up 65% of respondents, and 57% were based in the United States. That means country-level comparisons, including the report's Mainland China finding, rest on much smaller subsamples than the headline counts suggest. Treat regional rankings as directional, not as precise national measurements.

Definitions and weighting are not fully spelled out

The report groups AI progress into familiar stages such as initial, moderate, and active implementation, but the published methodology gives only shorthand descriptions, not complete definitions. It is also not clear whether responses were weighted to match any population. When a survey does not disclose weighting, its percentages describe respondents, not necessarily all libraries. The same caution applies to the 46% figure for academic libraries in initial, moderate, or active implementation, as well as the narrower 16% described as moderate to active.

Vendor context matters

Clarivate is a major library technology vendor. Its Alma and Primo integrated library systems are widely used, and its Nexus AI products are part of the commercial portfolio. That does not make the findings wrong, but it does mean the report should be read as a vendor's view of a market it serves. Independent surveys help triangulate.

Independent comparisons show how much library type and size matter

The OCLC Library Leaders AI Survey collected completed responses from 698 U.S. academic and public library directors, with a stated margin of error of plus or minus 3%. It found 24% used AI daily, 40% occasionally, and 36% rarely or not at all. But library type changed the picture sharply: 67% of ARL research directors used AI daily, compared with 14% of small public library directors, and 26% of small public directors reported no AI use. Only 13% of all leaders were more excited than concerned, while 40% of small public directors were more concerned than excited. A separate ARL quick poll, based on 39 completed responses from member research libraries, found more than one-third in active implementation, with most others exploring practical applications.¹ These surveys cover different populations and should not be treated as interchangeable, but together they suggest that size, type, and mission shape AI adoption more than a single global ranking reveals.

Why Are Libraries Slow to Adopt AI?

Some libraries describe slow AI adoption as a budget problem; others describe it as a governance and trust problem. The Clarivate Pulse of the Library data suggests both are true, but at different moments and for different library types.

Survey evidence: budget, privacy, and skills

• Budget: In the 2025 Pulse, 62% of libraries named budget constraints as a top concern, a familiar small library technology budget issue, and 47% called budget the greatest challenge. Clarivate noted budget overtook lack of expertise from 2024 to 2025.
• Privacy and security: 57% of libraries listed privacy and security as a barrier in 2025. The 2026 report describes a shift toward privacy, security, misinformation, and bias concerns, but Clarivate has not released comparable percentages for those items.
• Skills and training: 56% of libraries reported no structured learning support for AI in 2025. Only 12% had formal AI training or onboarding, while 28% of more active implementers had AI literacy embedded in training, 23.3% had dedicated time, and 24.2% had manager encouragement.

Survey evidence: governance and confidence

• Governance: In 2026, only 16% of libraries reported moderate or active AI implementation, and one-third were still exploring or evaluating. In 2026, 46% of libraries had a defined strategic position on AI and 72% used structured decision-making.
• Confidence: Senior staff reported 43% confidence in AI at a 4 or 5 level in 2025, compared with 36% for junior staff. Clarivate notes 2026 confidence stalled after a marginal rise in 2025.

Inference: academic vs public is directional, not settled

Public libraries in 2026 showed the clearest governance gap: 67% had no defined AI strategic position and 37% had no structured decision-making process. The available Clarivate figures do not supply matching academic library percentages, so the academic vs public librarian contrast should be read as directional, not as a precise comparison. Across the data, the common thread is that libraries are not simply rejecting AI; many are stuck between exploration and formal structure.

What AI Implementation Actually Costs Libraries

What does it actually cost a library to move from a pilot to routine AI use? The honest answer is that institutional pricing is thin, but the budget pressure points are consistent: vendor licensing or add-on fees, staff time, training, and hidden work like metadata cleanup, privacy review, and integration.

Vendor pricing is mostly custom, not list-priced

Major library-platform vendors, including Ex Libris Alma, OCLC, Innovative, and SirsiDynix, list AI capabilities but generally do not publish AI add-on prices. Alma and Wise, for example, direct buyers to contact sales. It is not public whether vendors charge per institution, branch, staff account, search volume, record, API call, or percentage of license. Consumer-style subscriptions such as those listed in AI Tools for Librarians , Jotform AI Agents at $34 per month, Connected Papers at $6 per month, and Microsoft Forms at $6 per user per month , are useful for small experiments, but they are not evidence of what a large library system pays. The Microsoft Forms figure reflects a broader Microsoft 365 subscription, not a stand-alone library AI add-on.

Grant budgets make staffing and training visible

Grant-funded projects offer more concrete numbers. The COBIMET "AI Literacy for All Librarians" project received $98,976 from an IMLS grant over 24 months to train about 250 librarians across 53 libraries, which works out to roughly $396 per librarian or $1,867 per library if divided evenly.¹ The LIBRAI proposal requested $149,886 and separated personnel, a graduate assistant, travel, honoraria, and indirect costs, showing that staff time and training are often bigger line items than software.²

Consortial funding stretches the dollars

Libraries also pool resources. The Library of Michigan offered an AI capacity-building series for up to 24 public-library staff³, and Columbus Metropolitan Library planned AI-fluency training for 25 employees through an IMLS award.⁴ OCLC WebJunction received $299,807 for a two-year project to update the Competency Index and design credentials, a national shared-resource model that avoids duplicating local training costs. IMLS reported $4,184,695 across eight AI literacy projects in FY2025⁵, with National Leadership Grants ranging from $25,000 to $1,000,000 depending on the category.⁶

Related Articles