Moving Beyond Automation: Projecting the Place of Discovery Libraries

The future of library discovery lies beyond automation: making collections and research findable across the network, not only in a local search box.

By Cephas Tope

Published 11/10/2026

The future of library discovery lies beyond automation: making collections and research findable across the network, not only in a local search box.

Key takeaways

  • The future of library discovery lies beyond automation. Automation made local library operations more efficient; discovery now decides whether collections, research and expertise are found, used and trusted.
  • A discovery library is an institution, not a product. It makes its collections, metadata, services and expertise discoverable in its own systems and in the search engines, scholarly indexes and research tools where users already work.
  • Discovery often happens elsewhere. People may find a work on Google, Google Scholar, Wikipedia or Amazon and only then turn to the library for a usable copy (Dempsey, 2007, 2020).
  • Metadata, persistent links and the path from discovery to fulfilment are core discovery work. A result that cannot be reached is a failed discovery.
  • Institutional outputs such as theses, research data and digitised collections need to be disclosed into the external systems where they will actually be found.
  • AI-assisted discovery should keep human review, source transparency and information literacy in the design.
  • In Nigerian academic libraries, readiness, cost, infrastructure and local capacity shape what any discovery strategy can achieve.
  • Success should be judged beyond search logs: did readers reach a usable copy, and did they trust the result?

Introduction: from automated catalogues to discovery libraries

Library automation solved a local problem. Catalogues, circulation, and related files were made more efficient, and for a generation that work defined much of what it meant to modernise a library. That work still matters. It no longer, however, describes where finding actually takes place.

People now encounter books, articles, data, and institutional materials in many settings—search engines, scholarly indexes, citation tools, Wikipedia, and, increasingly, AI-mediated research services—and only sometimes in a system the library itself provides (Dempsey, 2007, 2020). To treat the future of library discovery as the next module of an automated local stack is to mistake a part of the picture for the whole.

This article starts from that shift. Automation made library operations more efficient; discovery now determines whether collections, research outputs, and expertise are found, used, and trusted. A discovery library, as the term is used here, is not simply a library that has bought a discovery layer. It is an institution that deliberately makes its collections, services, and knowledge discoverable across its own systems and across the wider network where so much searching already begins.

That claim draws on a line of thinking associated with Lorcan Dempsey, who argued that “discovery happens elsewhere” and that libraries therefore need to think about how they disclose resources into other environments (Dempsey, 2007).

In a later synthesis he noted that discovery often occurs in network-level services such as Google, Wikipedia, and Amazon, and in more specialist resources such as arXiv, and that people may find a work in one of those places and only then turn to the library to locate a usable copy (Dempsey, 2020). The pages that follow take that framework as a starting point.

They trace the path from the online public access catalogue to web-scale discovery services and library services platforms; distinguish those systems from one another; and then ask what follows for institutional knowledge, linked data, open access, and AI-assisted search. A later section turns to Nigerian academic libraries, where readiness, cost, infrastructure, and local capacity still shape what any future of library discovery can mean in practice. The aim is not to recommend a single product.

It is to project the place of discovery libraries as a strategy of visibility, fulfilment, and judgement in a networked information environment.

What is a discovery library?

The phrase needs a little care, because it can be heard in more than one way. In much professional discussion, a discovery system is the user-facing search environment on a library website: a single search box that provides a point of access to library resources (Oh & Colón-Aguirre, 2019). That is a useful definition of a product category. It is not, by itself, a definition of the institution.

A discovery library, in the sense proposed here, is broader. It is a library whose collections, metadata, services, and expertise are designed to be discoverable in more than one environment: in the local catalogue or discovery layer, certainly, but also in the search engines, scholarly indexes, course systems, repositories, and research tools where users already work. The distinction matters.

A library may implement a well-designed discovery layer and still leave unique institutional materials poorly represented on the open web, or leave the path from a Google Scholar result to a licensed copy unreliable. Conversely, a library with a modest local interface may still do important discovery work if it discloses holdings, resolves access, and makes its own research outputs findable where they are likely to be sought.

Discovery, on this view, is not simply a search box. It includes the quality and consistency of metadata; the links that connect a citation to a copy the user is entitled to see; the arrangements that take a reader from finding to fulfilment; and the presence of library resources in external systems that the library does not itself control.

Dempsey has argued that a full appreciation of discovery and discoverability now involves thinking about much more than the discovery layer, as attention has shifted from the evolution of a single library service—from catalogue, to metasearch, to discovery layer—towards user behaviour in a complex network of library and third-party offerings (Dempsey, 2020).

He has also pointed to what he calls “full library discovery”: environments that can surface not only collections but website content, staff expertise, resource guides, and events, so that what is made visible is more of the library’s capacity, not only its inventory (Dempsey, 2020).

The definition is therefore institutional rather than technical. The local system remains important. It is one setting among several. The rest of this article asks how libraries arrived at that situation, how the main system types differ, and what it would mean to organise discovery as a strategy rather than as a purchase.

From library automation to networked discovery

The path is familiar, even if the labels have accumulated. For many years the online public access catalogue was the principal public face of the automated library. Wells (2016), in a study of discovery at Curtin University Library, notes that the OPAC became the principal tool from the late 1980s.

It made bibliographic records searchable, and it remained, in most implementations, closely tied to the local collection and to the record as the unit of description. That was a genuine advance on the card catalogue. It was also a local, collection-centred view of finding.

What followed was an attempt to make the public interface more like the rest of the web: relevance ranking, faceted navigation, and a less rigid search experience. A sharper break came with web-scale discovery services.

Vaughan (2011), in a Library Technology Reports issue devoted to the first wave of these products, described them as tools able to search quickly and seamlessly across a wide range of local and remote, pre-harvested and indexed content, and to return relevancy-ranked results in an interface closer to what users had come to expect. Locally held material—physical holdings, digital collections, institutional repositories—could sit alongside remotely hosted licensed content and material from open-access repositories.

The report reviewed four services then taking shape: OCLC WorldCat Local, Serials Solutions Summon, EBSCO Discovery Service, and Ex Libris Primo Central. The first of these services, Vaughan wrote, debuted in late 2007, with most reaching initial public release in 2010 (Vaughan, 2011).

In that sense, web-scale discovery was a response to fragmentation. Dempsey (2020) traces a service path from the catalogue, through metasearch, to the discovery layer, and notes that the discovery layer was, in some ways, an answer to the inefficiency of searching distributed databases on the fly. A central index could not remove every boundary, but it could hide some of them from the user.

Empirical work from the same period shows both the promise and the limits of this model. Wells (2016) examined Primo at Curtin University, where version 2 was launched in October 2009 and version 3 in November 2010, and where Alma replaced Aleph as the underlying library system in February 2014.

Using transaction logs (five-week samples between 2013 and 2015) and a 2014 user survey (204 started, 152 completed, in an institution of about 60,000 students and staff), he found that a single-search-box design was appropriate in that setting, and that users valued traditional library tasks over social-media-style features. Those social features were little used and were removed in early 2015 (Wells, 2016). The lesson is not that discovery layers failed.

It is that they are a stage in a longer reconfiguration, and that what users actually do with them still has to be watched.

Discovery layer vs ILS vs library services platform

The terminology is easily blurred in procurement conversations, so it is worth keeping the roles distinct. An integrated library system is first of all an operational system: it supports the management of collections and workflows. A discovery layer, or web-scale discovery service, is first of all a user-facing finding environment. A library services platform is a later, broader category of management software. The three can be connected, and are sometimes sold together, but they are not the same thing.

Breeding (2015) described library services platforms as products that help libraries manage collections and automate operations by addressing a wider range of resources, and by using more current technical architectures, than the integrated library systems that had previously dominated. An LSP, in his account, enables a library to acquire and manage physical and electronic content and to support more than one kind of procurement.

Integrated library systems, he added, remain viable for many libraries and continue to be developed along their own path. Management and discovery may be integrated; discovery is not a mere side-effect of back-office automation (Breeding, 2015).

A quick comparison of the four kinds of system:

Integrated library system

  • Primary role: Operational management of collections and workflows
  • Typical scope: Local collections and processes
  • Sourced examples: Aleph at Curtin, later replaced by Alma (Wells, 2016)

Online public access catalogue

  • Primary role: User-facing catalogue
  • Typical scope: Mainly local bibliographic records
  • Sourced examples: Principal public tool from the late 1980s (Wells, 2016)

Discovery layer / web-scale discovery service

  • Primary role: User-facing search over a pre-harvested, ranked index
  • Typical scope: Local and remote print, electronic, digital, and sometimes open-access content (Vaughan, 2011)
  • Sourced examples: WorldCat Local, Summon, EBSCO Discovery Service, Primo Central (Vaughan, 2011); Primo at Curtin (Wells, 2016)

Library services platform

  • Primary role: Broader management on a more current architecture, often linked to discovery
  • Typical scope: Physical and electronic collections and related procurement (Breeding, 2015)
  • Sourced examples: Discussed as a product genre by Breeding (2015)

Oh and Colón-Aguirre (2019) define the library discovery system as a website service offering a single search box over library resources. Vaughan (2011) adds the web-scale qualifier: the index is typically central, pre-harvested, and wider than the local catalogue. Wells (2016) shows the practical separation at Curtin, where Primo was the public discovery environment and Aleph, later Alma, the system underneath.

Buying or replacing an ILS or an LSP does not, by itself, answer the discovery question. Nor does a discovery layer exhaust it, once users begin their search somewhere else.

Discovery happens elsewhere: the network-level turn

The local discovery layer remains a visible part of many library websites, often treated as a shop window. It is not, however, where a growing share of finding begins.

In 2007 Dempsey introduced the phrase “discovery happens elsewhere” to name a change that was already becoming ordinary: no single website is the sole focus of a user’s attention, and people encounter library resources in a variety of other places, including network-level services such as Google and personal tools such as an RSS reader.

The practical consequence, he argued, was a shift towards disclosure of metadata, URL structures, feeds, and other traces that allow resources to be found from somewhere else. He credited Steve Rubel’s phrase “traffic happens elsewhere” as a prompt (Dempsey, 2007).

A few years later he set the same observation in a question of scale. Libraries were organised as institutional responses to information management, and the institution remains the right scale for many activities. Researchers, however, are often drawn to services that aggregate supply or attention at network or disciplinary scale: among the examples he gave were arXiv, SSRN, RePEc, ICPSR, Google Scholar, and Twitter. On discovery he was explicit.

“Discovery happens elsewhere: in Google, in Google Scholar, in Arxiv, in Mendeley, in Twitter, …” Two directions followed. One was to build single-search access to broad data wells—the discovery-layer solutions already discussed. The other was to syndicate library resources into external environments through APIs, data sharing, and related techniques, including the “find in a library” link in Google Book Search and Google Scholar’s work with institutional resolvers (Dempsey, 2011).

Merely making the library resemble popular network tools is, he suggested, at best a partial response. A more appropriate one is to consider library resources in the wider environment, and to invest in their visibility there—for example by ensuring that an institutional link resolver is properly configured for Google Scholar (Dempsey, 2011).

The 2020 foreword restates the pattern with more institutional coolness. Discovery often happens in network-level services that are now part of everyday online life, such as Wikipedia, Google, and Amazon, and in specialist resources such as arXiv. A user may find a work in a non-library service and only then turn to the library to locate a particular copy (Dempsey, 2020).

In a world of information scarcity, learners and researchers had to build their work around the library. In a network world, many will expect library resources to fit into existing workflows. The discovery layer, in this light, “addresses a part of the discovery need” and “seems very much like a service category in transition” (Dempsey, 2020).

Few libraries have taken the further step that Dempsey describes at Utrecht University Library, which, on the evidence of the university’s own guidance as retrieved in September 2019, dispensed with a conventional local discovery layer and treated several network-level resources as the primary entry point. The stated reason was that researchers and students often begin with a wide search, for instance in Google Scholar.

The library offered support for major scientific search engines and a mix of open and licensed network-level resources (including Scopus and Web of Science); a search of the library’s own collection, through the Dutch union catalogue via WorldCat and a resolver for journal titles, came second, alongside services for discovering research data (Dempsey, 2020).

Dempsey reads this not as a model for every institution, but as a clear case of scoping, scaling, and sourcing: a judgement about what the library should do itself, at what level, and through which mix of services. That triad is more useful than any single product decision.

Inside-out and outside-in: making institutional knowledge discoverable

Most discussion of library discovery has been about the acquired collection: the books, journals, and databases a library buys or licenses and makes available to a local audience. Dempsey (2020) calls this an “outside in” collection, because the materials originate outside the institution. That work remains central, and it is the work that conventional catalogues and discovery layers were built to support.

In a print world of scarce resources, assembling such a collection close to the user made sense, and “goodness” was often associated with its size (Dempsey, 2020). The habit is deep, and it still shapes how success is described.

A network environment changes the picture, because institutions are also producers. Research no longer issues only in the published book or article. Software, methods, preprints, and research data appear across the research life cycle, encouraged by funder and institutional mandates and by a wider interest in reproducibility (Dempsey, 2020). Universities maintain expertise and research profiles, and many researchers already manage their visibility through services such as Google Scholar or ResearchGate.

Special collections and archives, often unique or rare, are increasingly digitised not only for local teaching and research but so that they can be found by a wider audience and, where useful, placed alongside related materials from elsewhere (Dempsey, 2020). In these cases there is a growing interest in discoverability outside the institution as well as inside it. “In this sense,” Dempsey writes, “these are ‘inside out’ collections,” a label he takes from Dahl (2018) (Dempsey, 2020).

The discovery dynamic is different. It still makes sense to represent such materials in local systems. It may matter more that they are effectively represented in the external systems used by potential audiences: that they are crawled and indexed well, that metadata is supplied to scholarly or cultural aggregators, that links are added where people actually look, including Wikipedia (Dempsey, 2020). The library’s role, in this orientation, includes improving the discoverability of local scholarship and distinctive cultural materials, not only improving the local search box over content bought from elsewhere.

This is easy to under-resource, because attention to it is often diffuse and has no single organisational home (Dempsey, 2020). Yet for a discovery library it is not a side project. Theses, research data, digitised archives, and locally created records are first-class discovery assets. If the library’s public value is increasingly tied to what the institution itself produces, then inside-out disclosure is as much a part of the future of library discovery as the refinement of access to licensed content.

Linked data, BIBFRAME, and machine-readable discovery

If resources are to be found outside the local system, the data that describe them have to travel. For decades that data has travelled, with remarkable success, in MARC. The Library of Congress, setting out a new bibliographic framework in 2011, was careful to recognise that success, and to say that MARC 21 would need to be maintained for many years while a transition was designed.

It was equally clear that a carrier “grounded in the current and expected future shape of data interchange” was needed. The new work, it said, would focus on the web environment, on linked-data principles, and on the Resource Description Framework as a basic data model, with a view to replacing MARC as the common exchange currency in due course, not overnight (Library of Congress, 2011).

That initiative is now known as BIBFRAME, the Bibliographic Framework Initiative, and is described by the Library of Congress as grounded in linked data (Library of Congress, n.d.). It has not replaced MARC, and it would be misleading to speak as if it had.

What is at stake for discovery is not a cataloguing fashion. Linked, entity-based data can connect works, agents, places, and subjects across systems in ways that a self-contained bibliographic record does not. Dempsey (2020) noted that libraries already hold a substantial apparatus of authority files, and that bibliographic infrastructure is evolving towards graphs of works, authors, and places rather than only shipping bundles of data about titles.

As links grow and resolve to descriptions of “things,” it becomes more possible to match and merge data at scale, and to build richer navigation and context into interfaces. He placed this work in a wider pattern that includes Google’s knowledge graph and Wikidata, and he described it, in 2020, as still at an early stage, though promising for both discovery and discoverability (Dempsey, 2020).

That caution still holds as a warning against over-claiming: adoption remains uneven, and many systems will carry MARC for the foreseeable future (Library of Congress, 2011).

For AI-assisted discovery, which is taken up below, the same point recurs. Models and ranking services work better when the underlying data are consistent, related, and machine-readable. BIBFRAME and related linked-data work are therefore better understood as discovery infrastructure than as a purely internal cataloguing standard. The strategic question is whether a library’s metadata can be understood, connected, and reused beyond the local catalogue—not whether a single new format has already won.

Open access, repositories, and equitable discovery

Open access improves the potential availability of research. It does not, by itself, guarantee that the work will be found. A paper that sits in a repository without adequate metadata, persistent identifiers, or a route into the systems people actually search may be available in principle and invisible in practice. Availability and discoverability are related; they are not the same.

The range of what needs to be found has also widened. Dempsey (2020) notes that discovered resources may now include software, research data, and learning materials as well as books and journals, and that a strong interest in open access has brought forward services that can be integrated with a discovery layer, a knowledge base, resource sharing, or a browser plugin.

Facilitating integrated access to open-access materials, he argues, improves delivery: the path from discovery to fulfilment is shorter when an open copy can be reached without friction (Dempsey, 2020). The same logic sits behind the earlier, more practical advice that libraries should make sure their link resolvers work well with Google Scholar (Dempsey, 2011). Open content that cannot be reliably reached from the places where discovery begins is only half-published.

Repository infrastructure is therefore discovery infrastructure. Institutional repositories, preprint servers, open indexes, and persistent identifiers are part of the same environment as the local discovery layer, not a parallel world. That is an especially sharp issue where repositories have been built at some cost and then little used.

A systematic review of institutional-repository use in Nigerian academic libraries concludes that usage is very low, even given the investment universities have made in creating them (Temim et al., 2025). The finding is not an argument against repositories. It is an argument against treating deposit as the end of the work.

Without disclosure into search engines and scholarly indexes, without advocacy and user support, and without a clear place in teaching and research workflows, an institutional repository can remain a store rather than a discovery service.

For a discovery library, the implication is straightforward. Open access should be connected to the same questions already raised about inside-out collections: where will this material be encountered, how will it be identified, and how will a reader reach a usable copy? Equity of access depends on those answers as much as on the licence.

AI-assisted discovery: promise, limits, and governance

The next change in the future of library discovery is already being felt in the tools researchers bring with them. Interest has moved from the single search box towards natural-language questions, conversational assistance, summarisation, citation recommendation, and maps of a research field. This is not an entirely new aspiration.

Dempsey (2020) already noted that discovery is data-intensive and likely to draw on machine learning for summarisation, recommendation, ranking, and personalisation, and that this raises issues of privacy, algorithmic transparency, authority, and appropriate use, with libraries well placed to advocate for users. What has changed is the speed and visibility of generative systems.

A useful snapshot of professional discussion is the OCLC Research Library Partnership and LIBER facilitated conversation of 17 April 2024, synthesised by Bryant (2024). Participants from 31 institutions in twelve countries described curiosity mixed with apprehension.

Under discovery and content evaluation, they were interested in searching with natural-language phrases as well as keywords, and they noted researchers’ use of tools such as scite, Consensus, ResearchRabbit, Perplexity, and Semantic Scholar to summarise findings, recommend citations, and visualise a literature.

They also observed AI features being built into existing indexes such as Scopus and Dimensions, in some cases combining retrieval-augmented generation over a local index with generative models, a pattern described as a way of reducing hallucinations (Bryant, 2024). That characterisation should be handled carefully. Retrieval-augmented generation can ground answers in a selected corpus; it does not eliminate the risk of inaccurate or misleading output.

The same discussion raised a wider set of concerns: energy use, privacy, the use of copyrighted material in training data, misinformation, the dominance of English in large language models, and information overload. Academic integrity loomed large in institutional contexts. Upskilling was often ad hoc. Participants stressed the need for transparency, accountability, and inclusivity, and they saw a leadership role for libraries in AI literacy as an extension of information literacy (Bryant, 2024).

NISO’s Open Discovery Initiative, in a 2024 white paper, Generative Artificial Intelligence and Web-Scale Discovery, likewise treated generative AI as a matter for the discovery community, based on a survey of perceptions of its impact on library discovery services (National Information Standards Organization, 2024).

A vendor-sponsored survey offers a market signal, not a census. Clarivate’s first Pulse of the Library report, based on 1,527 responses gathered between April and June 2024 (76% from academic libraries, 47% from the United States), found that over 60% of respondents were evaluating or planning for AI integration, and that, among those evaluating or implementing it, content discoverability was a stated goal for 45% (Clarivate, 2024). Those figures describe interest and intent in a particular sample. They do not measure successful, well-governed adoption.

The position that follows is deliberately modest. AI may augment discovery; it should not be asked to replace metadata work, curated collections, access services, or human judgement. Any local experiment ought to show its sources, keep a person in the path where advice or citation is at stake, and treat literacy as part of the service, not as an afterthought.

Discovery libraries in the Global South and Nigeria

The argument so far has been general. It has to be tested against conditions in which connectivity, funding, staffing, and local systems are not those assumed by much of the North American and European discovery literature. Nigerian academic libraries offer a documented case, not a complete picture of the Global South, and the studies cited here should be read with that limit in mind.

Adeyemi and Omopupa (2020), in a mixed-method study published in Cataloging & Classification Quarterly, examined Nigerian librarians’ knowledge of, and readiness for, a move from OPACs to discovery systems. Their conclusion was blunt. The recent shift observed elsewhere had not been observed in Nigeria; librarians were not ready to make that move in the nearest future; and there was little or no knowledge of library discovery systems among those studied (Adeyemi & Omopupa, 2020).

The finding is now several years old, but it names a gap that cannot be closed by enthusiasm for network-level strategy alone. If the people who would implement and teach a discovery service do not yet know the systems, procurement is not the first problem.

A later state-level survey adds texture. Njuku, Nwachokor, Okonkwo, and Okwudibe (2025) studied 113 librarians in universities, polytechnics, and colleges of education in Delta State. Traditional cataloguing tools, including manual catalogue cards and card indexes, remained the most widely available and most used resource-discovery mechanisms. Modern digital and metadata-driven discovery tools showed limited availability and low use.

Systems such as the OPAC, KOHA, Google Scholar, and EBSCOhost Discovery Service were beginning to appear, but usage was still minimal. The challenges reported were familiar and concrete: inadequate funding, poor internet connectivity, fragmented or low-quality metadata, limited search skills among users, and low technological familiarity among staff (Njuku et al., 2025).

The authors describe libraries in the state as still in transition between traditional and modern discovery, and they point to sustained funding, infrastructure, metadata quality, interoperability, information literacy, and staff digital skills as the work required (Njuku et al., 2025). This is one state’s evidence, in a professional journal, not a national census.

It is consistent, however, with the earlier readiness study, and with the finding that institutional-repository usage in Nigerian academic libraries is very low despite the cost of establishing those repositories (Temim et al., 2025).

The strategic conclusion is not that Nigerian libraries should wait, nor that they should simply import platforms designed around different budgets and infrastructures. It is that a discovery library in this context has to be built in a way that is affordable, interoperable, and locally sustainable.

That means investment in metadata and connectivity as much as in a search product; attention to local content and institutional outputs that will otherwise remain invisible; and serious work on staff and user capacity. Network-level disclosure still matters here, perhaps especially here, because a library that cannot offer a full commercial discovery layer can still improve the visibility of theses, repositories, and digitised local collections in the open systems that students already use.

The future of library discovery, if it is to mean anything outside well-resourced research libraries, has to be designed with those constraints in view.

Librarians and information professionals in Nigeria who want to build the digital skills these changes call for can start with this Handmework guide on how Nigerians can learn AI, cybersecurity and data analytics.

Building a discovery library: a practical agenda

What, then, should a library actually do? The useful answers are less about a preferred vendor than about a small number of recurring choices.

First, find out where users already encounter library content. Dempsey’s work points consistently to Google, Google Scholar, Wikipedia, and other network venues, and to the unglamorous work of making resolvers, APIs, and metadata syndication function in those places (Dempsey, 2007, 2011, 2020). An honest audit of those paths is a better starting point than a redesign of the local home page.

Second, treat metadata, persistent links, and the path from discovery to fulfilment as core discovery work, not as technical residue. A result that cannot be reached is a failed discovery (Dempsey, 2020).

Third, give institutional outputs the same attention long given to purchased content. Theses, research data, digitised collections, and faculty work need to be disclosed into the external systems where they will actually be found (Dempsey, 2020).

Fourth, decide consciously what the library will do itself, what it will do with partners, and what it will source from network-level services. Dempsey’s triad of scoping, scaling, and sourcing is a practical planning language for those decisions (Dempsey, 2020). It also implies that discovery is a cross-departmental service, involving cataloguing, systems, research support, teaching, and communications, not only a systems-office purchase.

Fifth, if AI tools are piloted, keep human review, source transparency, and literacy in the design (Bryant, 2024; National Information Standards Organization, 2024). Experiment is useful; unexamined automation is not.

Sixth, watch what users actually do. Wells (2016) found a single search box appropriate at Curtin, and also found that social features, little used, could be removed. Evaluation has to continue after implementation.

None of this requires every library to follow Utrecht in setting aside a local discovery layer. It does require that a discovery layer, where one is used, be understood as part of a wider environment rather than as the whole of the library’s discovery strategy.

Measuring discovery beyond search logs

How, then, should success be judged? Search logs are useful, and they are not enough. They can show what was typed and what was clicked. They cannot, by themselves, show whether a reader reached a usable copy, trusted the result, or was well served compared with the other tools they also use.

Two studies make the last point difficult to ignore. Oh and Colón-Aguirre (2019) surveyed 975 users at more than twenty public research universities in the United States. The ways participants used Google Scholar and library discovery systems were similar, and both were seen as highly accessible and useful. Library discovery systems scored higher on perceived comprehensiveness, subjective norm, loyalty, and intended use. Google Scholar scored higher on perceived ease of use, system quality, and satisfaction.

Google Scholar, they concluded, is not a replacement for library resources, but it is a popular tool and one that shapes expectations (Oh & Colón-Aguirre, 2019). Any local evaluation that ignores that comparison will misread its own results.

Wells (2016) combined transaction logs with a user survey and, as already noted, found that a single search box was valued while social features were not. Logs described use; the survey spoke to perceived value. The pairing is the method worth copying.

What else a library should count will depend on its mission, and the following are recommendations, not findings of those studies: conversion from discovery to access or fulfilment; use of full text; deposits and reuse in institutional repositories; user trust and task completion; and whether access is equitable across groups of users. The point is simply that the future of library discovery will be poorly steered if the only dashboard is a rising or falling number of searches.

Conclusion: the future of library discovery is a strategy, not a product

The arc of this article can be stated without much ornament. Library automation made the management of local collections and workflows more efficient. Web-scale discovery services and library services platforms then widened what could be searched and managed. Those were necessary steps. They do not settle the place of the library in a network environment where finding so often begins somewhere else.

A discovery library, as defined here, is an institution that treats discoverability as a strategy: disclosing what it holds and what it produces, connecting discovery to fulfilment, and remaining present in the environments its users already inhabit.

That includes the acquired collection and the inside-out collection; open repositories as well as licensed databases; machine-readable metadata as well as a public interface; and, now, a careful stance towards AI-assisted search that augments judgement rather than pretending to replace it. In Nigerian academic libraries, and in other settings where cost, connectivity, and capacity are tight, the same strategy has to be built with tools and partnerships that can actually be sustained.

The future of library discovery will not be decided by a single product category. It will be decided by whether libraries can make their collections, research, and expertise findable, usable, and trusted—locally, and on the network where so much of the work of finding now takes place.

References

Frequently asked questions

What is a discovery library?

As used here, a discovery library is an institution that makes its collections, services, and expertise discoverable in its own systems and on the wider network. It is broader than a discovery system, which Oh and Colón-Aguirre (2019) define as a library-website environment with a single search box over library resources.

What is the difference between a discovery layer and an ILS?

An integrated library system is primarily an operational system for managing collections and workflows. A discovery layer is a user-facing search environment, often over a pre-harvested index of local and remote content (Vaughan, 2011; Breeding, 2015). They may be connected, as at Curtin University (Wells, 2016), but they are not the same.

What is a library services platform?

Breeding (2015) describes library services platforms as products that manage a wider range of resources, including physical and electronic content, on more current architectures than traditional integrated library systems. ILSs, he notes, remain viable for many libraries.

What is web-scale discovery?

Vaughan (2011) describes services that search quickly across local and remote, pre-harvested and indexed content and return relevancy-ranked results. Early products included WorldCat Local, Summon, EBSCO Discovery Service, and Primo Central. The first debuted in late 2007; most reached public release in 2010 (Vaughan, 2011).

How do discovery systems differ from OPACs?

The online public access catalogue, the principal public tool from the late 1980s, was typically centred on local bibliographic records (Wells, 2016). Web-scale discovery systems search a broader index that may include licensed, digital, and open-access material as well as the local catalogue (Vaughan, 2011).

What does “discovery happens elsewhere” mean?

Dempsey (2007) used the phrase to name a simple fact: people often encounter library resources in search engines, personal tools, and other network services, not only on the library website. Libraries therefore need to disclose metadata and access points into those environments (Dempsey, 2007, 2011).

What are inside-out and outside-in collections?

In Dempsey’s (2020) account, an outside-in collection is acquired from outside the institution for local use. An inside-out collection, a label he takes from Dahl (2018), is produced by the institution and often needs to be discovered by external as well as local audiences.

How will generative AI change library discovery, and can it replace discovery systems?

Emerging uses include natural-language search, summarisation, and citation recommendation, alongside concerns about hallucination, privacy, and literacy (Bryant, 2024; National Information Standards Organization, 2024). AI may augment discovery. This article does not claim that it can replace metadata, curated collections, or human judgement.

What are the barriers to discovery systems in Nigerian academic libraries?

Adeyemi and Omopupa (2020) found little knowledge of discovery systems and a lack of readiness to move from OPACs in the nearest future. A Delta State survey of 113 librarians found manual catalogue tools still widely used, and modern discovery tools little used, with barriers including funding, connectivity, weak metadata, and limited skills (Njuku et al., 2025). These studies should not be stretched to all African libraries.

What role should libraries play when discovery begins in Google or Google Scholar?

They should make library resources findable and fulfilable from those starting points, including through metadata exposure and link resolvers (Dempsey, 2011, 2020). The local system remains part of that work. It is no longer the whole of it.