Skip to main content
v2026.11,986 entries · CC-BY 4.0

Primo Research Assistant: What It Searches, Limits and Exclusions

Ex Libris Primo Research Assistant explained: what it searches, how to use it, the daily and query limits, content exclusions and administrator settings.

Written and maintained by CASRAI Editorial Board

Published

On this page

Primo Research Assistant is a generative AI feature that Ex Libris, a Clarivate company, built into its Primo discovery interface. A signed-in user types a research question in plain language, and the tool converts it into a Boolean search of the Central Discovery Index (CDI), re-ranks what comes back, and writes a short overview from the descriptive metadata of the top sources, with links to those sources. It is a layer over a library catalogue, not a separate database, and a library has to switch it on.

What Primo Research Assistant is

Ex Libris documents the feature in its Knowledge Center article “Getting Started with Primo Research Assistant” (read 2026-11-21), which covers Primo VE, the cloud edition of Primo, and its two interfaces, the Original UI and the NDE UI. The article describes a retrieval-augmented generation (RAG) design with the OpenAI model GPT-4.1 Mini. In five steps, it:

  1. asks the language model to convert the question into a Boolean query with several variations joined by OR (for a non-English question, variations in both the question language and English);
  2. sends that query to the CDI and, if the library has enabled it, to the library’s own Primo VE records;
  3. re-ranks the results to find the most relevant sources;
  4. asks the model to write an overview from the descriptive metadata of the selected sources;
  5. shows the overview together with the sources it drew on.

The step that matters most for evaluating an answer is the fourth. The overview is written from abstracts and other descriptive metadata, not from the full text of articles or books. Anything an abstract does not say, the overview cannot report.

Clarivate announced the launch on 2024-09-18, describing a beta that began in June 2024 with 18 institutions in 10 countries and regions. In that announcement Clarivate said the answer is based on the top five abstracts and that a Summon version, Summon Research Assistant, was planned for the first quarter of 2025.

Who runs it and what it covers

Ex Libris and Clarivate develop and host the feature; each library decides whether to enable it in a given Primo view. The Ex Libris article states that both access options (the Main Menu icon and the Brief Results widget) are off by default and are enabled in Configuration › Discovery › Display Configuration › Configure Views on the General tab. A third setting, Enable Research Assistant use of local data, controls whether the library’s own records are searched in addition to the CDI. The tool is not available to a user who is not signed in unless the library has changed that default; the documentation says sign-in is required by default.

Content in scope

By default the assistant draws on CDI records. When local records are enabled, it can also use Primo VE records that come from Alma physical, electronic and digital inventory, Esploro research assets and external sources indexed in Primo VE. For local records the library chooses up to five metadata fields in priority order, and the tool uses the first field that has a value. The default order is Description, General Note, Table of Contents, Local Notes, and blank. The Ex Libris article states that full text is not used and that metadata quality affects answer quality.

Content left out

The documentation lists CDI content that is excluded, and says the list will be updated regularly. As of the 2026-11-21 reading, the exclusions are news content, collections not marked “free for search” in the CDI collection list (with the exceptions Ex Libris names), records with insufficient metadata or abstracts, withdrawn or retracted documents and retraction notes, and collections from DataCite, Elsevier, JSTOR and Condé Nast, including their content received through aggregator collections. The practical consequence is that an overview can omit a relevant Elsevier or JSTOR article that the same library’s ordinary search would show, so the exclusion list belongs in any decision about relying on the tool.

Size claims vary by source and date. Clarivate’s 2024 announcement described the CDI as holding over five billion records, and the University of Florida’s library guide describes it as nearly six billion. The table below compares the operating limits that Ex Libris and one library state; it leaves the record count out because the sources give different totals.

How to use it

The following steps come from the Ex Libris documentation and the University of Florida guide (George A. Smathers Libraries). Interface labels can differ between institutions because libraries can rename them, and a library may have turned the feature off.

  1. Find the entry point. Sign in to your library’s Primo, then look for the Research Assistant icon in the Main Menu (available from any page) or the widget on the Brief Results page, which opens a dialog box. Where a library has set it up differently, the University of Florida guide describes it as a “Research Assistant” link in the top menu bar.
  2. Write a complete question. A query needs at least two words and at least ten characters, and the documentation gives a maximum of 250 characters. Follow-up questions are not supported, so each question must carry its own context. If a query fails, the tool shows “We couldn’t generate an answer for your question.”
  3. Optionally refine. Select the Refinement icon beside the search box to restrict by resource type (books, journal articles, peer-reviewed), by date (last 12 months, last 5 years, last 10 years or custom) or by the online availability check box. Defaults are all types and all dates.
  4. Read the Overview of Sources. The response has three parts: Sources, Overview of Sources, and Related research questions. The overview offers Copy and Try Again, and a rerun can give a different answer.
  5. Open the sources. Selecting a source shows its abstract with options such as Download PDF, Full Text Online and More Details and Request Options. Thumbs up and thumbs down buttons send feedback on a source or on the overview.
  6. Move to a conventional search. Choose View more results to run a standard Primo search and then use facets and filters. The CARLI webinar summary reports that this view shows the constructed search terms, which is useful for learning how the Boolean query was built.
  7. Manage your history. The left panel lists the first question of each topic; select Start a new topic to separate research areas. Saving of history is a personal setting, Allow saving Research Assistant search history, and the last 200 searches per user are kept.

Limits and exclusions at a glance

The data table on this page compares the documented limits, the number of sources behind an overview, and the content exclusions, each with the source that states it. Two points need care when reading it. First, the daily limits are described by Ex Libris as internal preventative measures, so they are an operating setting and not a contractual commitment. Second, the University of Florida guide says the tool analyses up to 30 top results before selecting the five abstracts it summarises, which means the overview represents a small part of the result list; the guide itself advises comparing the summary with the full result list.

Strengths and limits

What the sources report as strengths

  • It turns a natural-language question into a Boolean search and shows the sources, which can help a novice see how a search is constructed. The CARLI summary says the tool “does not attempt to do all the work for users”.
  • Answers link to records the library holds or licenses, so a user can move from an overview to the item and its full-text or request options.
  • Non-English questions are searched in the question language and English, and the overview is returned in the question language; the University of Florida guide says more than 50 languages are supported. Ex Libris notes that language support depends on the model version and may change.
  • Administrators can add a custom link to responses. The default label for it is “How to cite AI-generated content”, and the link target must be a full URL.

What the sources report as limits

  • Metadata only: overviews come from abstracts and descriptive fields, not full text.
  • No follow-up questions, and each question stands alone.
  • The CARLI webinar summary (2024-10-31) recorded that the tool could return sources a patron cannot access while omitting accessible ones, that a label reading “the 5 most relevant sources” might mislead users, that the scope of a search was unclear to users, and that of 26 survey responses most felt not ready. That summary dates from the early rollout, and the Ex Libris documentation has since added optional local records.
  • Ex Libris warns that large language models can be inaccurate and recommends checking responses against the sources. The known issues it lists include a latest question that may not be saved to research history after a page refresh.

The sensible reading for a researcher is to treat the overview as a pointer to records. Open the cited sources, check the abstract against the claim made about it, and use View more results for any search that must be complete. A systematic or comprehensive search still needs a documented strategy in the databases themselves; see the CASRAI guide to Boolean search operators for literature searching.

Library administrators: what to know before enabling it

The Ex Libris article lists several view-level requirements. The ISSN index must be enabled on the Advanced Search Configuration tab, or the CentralIndex scope must be active in a Search Profile Slot, so that View more results works. Labels are edited in Configuration › Discovery › Display Configuration › Labels, under Research Assistant Labels and Research Assistant Search Screen Labels. History saving is controlled by the customer setting save_users_RA_search_history, which defaults to true. A Resource Recommender banner can promote the feature on Brief Results, and the Action Usage subject area includes a ResearchAssistant group for analytics.

The CARLI summary adds that administrators can trial the feature with Test View, and recommends checking the institution’s own AI policy before enabling it. For the wider discovery platform this feature sits on, see the CASRAI guide to Alma from Ex Libris.

Primo Research Assistant is one of several AI-assisted literature search options. Only the entries below are described from sources read for this page; the linked CASRAI guides cover the others on their own terms.

Option What the sources say Source
Summon Research Assistant Clarivate’s 2024-09-18 announcement said a Summon version of the feature was planned for the first quarter of 2025, running on the same Clarivate Academic AI Platform. Clarivate press release
Standard Primo search (View more results) The Research Assistant hands the user to a normal Primo search with facets and filters; this is the route for any search that must be complete. Ex Libris documentation
Scopus AI A separate AI search product covered in a CASRAI guide, with its own scope and limits. Scopus AI guide
General AI research assistants Tools not tied to a library’s holdings; see the CASRAI overview of AI-powered research assistant tools and the dictionary definition of an AI research tool. CASRAI pages

The generic term “research assistant” in CASRAI’s dictionary refers to a person who supports a research project; the product on this page is unrelated to that meaning (see research assistant).

Sources

Data: Primo Research Assistant: documented limits, source selection and exclusions

Operating limits, the number of sources behind an overview and the content exclusions for Primo Research Assistant (Primo VE), compiled from the Ex Libris documentation and the University of Florida library guide on 2026-11-21.

Period covered: 2026.

Primo Research Assistant: documented limits, source selection and exclusions
ItemValueStated by
Minimum query lengthAt least two words and at least ten charactersEx Libris Knowledge Center
Maximum query length250 charactersEx Libris Knowledge Center
Questions per day per institutionUp to 10,000Ex Libris Knowledge Center
Questions per signed-in user per dayUp to 100Ex Libris Knowledge Center
Saved searches per userLast 200Ex Libris Knowledge Center
Follow-up questionsNot supported; each question stands aloneEx Libris Knowledge Center
Sources summarised in the overviewFive abstracts, selected from up to 30 top resultsUniversity of Florida Smathers Libraries guide
Content used for the overviewDescriptive metadata; full text is not usedEx Libris Knowledge Center
Default state in a viewIcon and widget access both off; sign-in requiredEx Libris Knowledge Center
CDI exclusion: content typesNews content; records with insufficient metadata; withdrawn or retracted documentsEx Libris Knowledge Center
CDI exclusion: providersDataCite, Elsevier, JSTOR and Condé Nast, including via aggregator collectionsEx Libris Knowledge Center

Sources

Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · full access

Ask about Primo Research Assistant: What It Searches, Limits and Exclusions

Full access: large monthly usage, up to 10 source passages behind each answer, and MCP and API access for Claude Code, Codex and other CLIs. Cancel any time.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →