Marshall Breeding's 2025 Library Systems Report

On May 1, Marshall Breeding released the 2025 Library Systems Report. For decades, this report has summarized the state of the library systems market, highlighting changes and trends. Over the past several years, the key trend has been consolidation of the marketplace as larger companies and investment firms have acquired smaller system vendors to expand customer bases and absorb technological innovations into their existing suite of offerings.

This year, however, the dominating trend seems to be stability. As Breeding points out, private equity firms have largely cashed out of the market, selling their products to larger, more stable companies that are interested in long-term development and growth over fast profits. Despite that, Breeding predicts at least some mergers in the next few years, especially as the smaller privately-held companies reach a point where founders and generational leadership wind down. Breeding also predicts acquisitions of companies that have developed technologies outside of libraries that library systems vendors wish to incorporate in their next-generation offerings.

Most of the new technologies being developed for inclusion in library systems, as Breeding points out, are related to AI. Examples include:

• Alma Specto: A digital collections management platform slated for release by early 2026 that Ex Libris is developing in partnership with other institutions.
• Primo Research Assistant: Ex Libris' discovery add-on that can combine results from their Central Discovery Index with AI-generated summaries and suggestions. 
• Natural Language Search and AI Insights: EBSCO's AI-assisted search interface and summary tools have completed extensive beta testing and are expected to be rolled out to all EBSCO customers by the end of the year.

Despite acknowledging the growing inclusion of AI in library systems, both in collections management and discovery, and admitting that AI tools could supplant existing technologies and give the competitive edge to the companies with the best implementations, Breeding concludes this year's report with a word of caution about these tools:

"Libraries will have to carefully parse which technologies can amplify their work and which may do harm. They are already working closely with vendors to ensure that products and services with AI features deliver appropriate results. Having the ability to automatically generate descriptive metadata at a speed and scale not previously imagined is an attractive prospect, but these tools must be accurate."

That is an astute point. AI is still prone to "hallucinations" and many users still misunderstand how AI works and blindly trust the results without double-checking, to their potential peril. A recent high profile story about a law firm that faced a $31,000 sanction for allowing hallucinations to slip into their filings demonstrates this point. It was not an isolated incident as other recent stories show.