Elsevier Leapspace.
For these additional features to be worth what universities are increasingly being asked to pay for them, however, they need to be trustworthy. The LeapSpace adverts now appearing in my personal social media feeds describe it as a “research-grade AI workspace”. That sounds reassuring, but does it stand up to scrutiny? In testing LeapSpace, I clicked through from one of its confident summaries to an article cited in support of a particular claim. From the title and abstract alone, it was immediately clear that the AI had drawn the opposite conclusion to that of the paper itself. In other cases, the drive to compress complex research into a single sentence stripped out most of the nuance. The summaries weren’t entirely wrong, but they often weren’t entirely right either. Crucially, they were delivered with such confidence that a time-poor student could easily take them at face value and cite the paper without reading it.
This, of course, is not entirely differ- ent from relying on a review article to summarise a field, and critical thinking is required in both cases. Yet I can’t help feeling that the slickness of platforms like LeapSpace actively encourages trust while discouraging a more critical eye.
Popularity contest
Another recurring concern with AI-surfaced content is the opacity of the algorithm itself. It is often difficult to know why particular articles are being shown and others are not. Many platforms are quite open about the fact that they prioritise popular and recent content by default. This can easily obscure older, but still valid and impor- tant, arguments within a field. When I searched for material on the early days of the open access movement, I was served recent reflective articles looking back, but nothing published at the time. I could force the system to surface older material, but this was not the default behaviour. These algorithmic choices are also likely to
April-May 2026
compound existing inequities in scholarly communication. We already know that Western and male voices are more likely to be amplified; AI risks reinforcing this by nudging researchers towards work that is already popular and frequently cited.
End of evidence
A more practical, but no less serious, issue for both publishers and acquisitions librarians is the challenge of recording usage when content is accessed through an AI middle layer. If an AI tool provides a summary, and a student simply incorpo- rates that summary and citation into their work, no download or view is recorded on the publisher’s platform. Librarians rely on these usage statistics to understand which journals are valued, and which are seeing changes in use over time. Without reliable data, acquisition decisions could increasingly be driven by cost alone rather than by evidence of need. If librarians do not know to buy subscriptions to certain journals, then it may be become increas- ingly difficult to access the work behind the AI summary.
Open access threat Open access is another area of librarianship that I spend a great deal of time thinking about, partly because I also spend a significant proportion of my budget making it happen. Read-and-pub- lish deals and article processing charges are expensive ways of ensuring content is openly available, but they are not the only routes.
Preprints, green open access through repositories, and Diamond open access models all offer alternatives at a much lower cost. AI is largely indifferent to these distinctions, although it argua- bly struggles most with the traditional publisher PDF. By removing the need for researchers to encounter the article “in the wild”, the rationale for paying for
open access begins to shift. If AI tools can access a green open access version as eas- ily as a publisher’s version, that becomes an attractive option for both researchers and libraries. For publishers who have built their financial plan on a pay to publish model, it could be a concerning development however.
The ability of AI to access subscrip- tion content, usually through deals with publishers, risks taking us back to an uncomfortable starting point. If most of the people you want to read your work are accessing it though AI, then why bother paying for it to be open access? We could easily see a backsliding to the days of majority subscription content. The open access movement arose in response to the gatekeeping of academic literature: without access to a university library, you simply could not read the research. Are we now setting up a new two-tier system? If you don’t have access to the right AI-based tools, it may become harder and harder to engage with the literature, particularly if researchers and publishers start optimising for machine rather than human access. LeapSpace can be purchased on an individual sub- scription, but at £250 a year it is hard to imagine this, or other paid for AI tools, being a realistic option for many students. Of course, the universities with healthier balance sheets are likely to provide tools, although with the current financial crisis this may not be possible for all institutions.
Impact of bypassing content As more content is consumed via AI-gen- erated summaries, the perceived need for open access may diminish further. Yet these tools rarely provide access to the full text itself. Researchers are left either to track down and pay for the original article or to trust the AI’s summary. I know which option I think is more likely to be happening. Databases like Scopus
INFORMATION PROFESSIONAL 27
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60
