Cited or mentioned? What we learned about how research actually shows up in policy
Nathalie Cornée & Kat Hart
Citations aren’t the only way research shows up in policy. Sometimes it’s a named researcher, a nod to “the evidence,” or a reference tucked into the text with no formal citation at all. Kat Hart and Nathalie Cornée dug into Overton’s data to find out how often that happens, and what it means for how we understand research’s reach into policy.
When we want to know if research has influenced policy, we usually turn to citations. They’re easy to spot and easy to count, so it makes sense that they’ve become the go-to signal for research influence.
But citations are only part of the story. Policymakers don’t always follow academic conventions. Sometimes they name a researcher directly, reference a report rather than the paper behind it, or they simply gesture at “the evidence” without pointing to anything specific at all.
So, we set out to ask a simple question: if we look past formal citations, what else might be going on? Using Overton’s data, we compared the scale of formal citations against these more informal mentions and tried to get a sense of what the gap between them might look like. This is an exploratory pilot rather than a definitive study, as the sample is far too small to draw firm conclusions. We hope its value lies more in how it reframes the question of what counts as ‘research engaging with policy’ (and vice versa) in the first place.
Why this matters right now
Definitions of research impact are shifting. A recent Australian study published on LSE’s impact blog made a useful point on this: institutional definitions of impact often don’t match how researchers actually work (O’Connor et al., 2026). Instead of asking “did this have impact?”, the authors suggest asking “what value does this create, and for whom?” It’s a small shift in framing, but it moves the conversation from proving impact to understanding it.
If citation counts alone don’t capture the full picture, what other signals should we be paying attention to?
How we approached it
We used the Overton Index as our starting point. At the time of the study, Overton indexed over 27 million policy documents from more than 190 countries and 3,100 sources, spanning everything from intergovernmental organisations like the UN and WHO to NGOs, think tanks, national governments, as well as regional and local bodies like Canadian provinces and UK devolved nations.
For the pilot itself, we kept things deliberately focused. We took a random sample of 50 UK government policy documents from 2025 and used Overton’s AI classification to pull out document type, length, and authorship. Then came the manual part: we read and coded every single document by hand.
This was a conscious choice. Automated tools are good at catching formal citations, but they miss the less obvious ways research shows up in policy text. We wanted to capture all of it, so we built a coding framework covering two broad categories:
- Objects: the types of evidence referenced, such as peer-reviewed publications, policy evidence, or informal knowledge transmission material like blog posts and policy briefs.
- Actors: the people and organisations referenced, such as research organisations, named researchers, funders, or in-house policy experts.

Figure 1. Workflow for identifying and classifying attention to actors and objects in UK government policy documents.
For each one, we recorded whether it was a citation (a formal, traceable reference) or a mention (an informal acknowledgement with no formal citation attached). Where a document referenced several things together, we logged each one individually rather than lumping them into a single entry. Across just 50 documents, that added up to over 2,200 individual references, which gave us plenty to work with despite the small sample size.
What we found
Policy talks to policy more than it talks to academic research
Across our sample, we identified 2,244 objects and actors in total. Policy actors (think tanks, government bodies, IGOs) and policy evidence (past reports, transcripts, consultation outputs) were by far the most common categories in our data, together with in-house research accounting for around 75% of everything referenced.
Peer-reviewed publications, by comparison, made up just under 5% of all references. Generic, non-specific nods to “experts” or “previous research” outnumbered them, at around 9%. In other words, plenty of documents acknowledge that evidence exists somewhere, without pointing to anything specific.
Citations and mentions tell different stories depending on what’s being referenced.
Peer-reviewed articles were the type of evidence most likely to receive a full, formal citation. Commissioned research told a different story: it often got absorbed into the narrative (“the data suggests”, “interviews found”) with no formal attribution at all. Policy outputs sat somewhere in between, appearing as both citations and mentions in meaningful numbers.
Across the whole sample, mentions were consistently more common than citations.
Who wrote the document matters as much as what it says
Referencing patterns varied a lot depending on who authored the document. Our five most prolific authors in the sample were the Department for Education, the Law Commission, UK Government more broadly, the Ministry of Justice, and the Department for Environment, Food and Rural Affairs (DEFRA) together with the Animal and Plant Health Agency (APHA). Each had a distinct pattern of what they referenced and how often.
A working paper from the Foreign and Commonwealth Office, written by commissioned academics, leaned heavily on peer-reviewed citations, as you’d expect. A similar paper from Technopolis, a consultancy commissioned by the Department for Science, Innovation and Technology (DSIT), leaned instead on funder acknowledgements and commissioned research. Same document type, same broad government context, very different citation behaviour.
Longer documents don’t necessarily cite more
We also tested the intuitive assumption that longer documents would contain more citations. They don’t, at least not cleanly. Some of the highest citation counts did appear in longer documents, but plenty of short documents cited heavily too, and plenty of long ones cited very little. There’s no straightforward relationship between page count and citation volume in this sample.
Even formal citations don’t always make it into a bibliography
Most documents in our sample didn’t have a formal bibliography at all, relying on footnotes instead. And even where a reference qualified as a formal citation, it only had roughly a 50/50 chance of actually appearing in a bibliography rather than just the body text. Citation habits clearly vary a lot from one policymaker to another.
The research being cited mostly comes from a small number of places
Of the peer-reviewed research we identified, half came from UK universities alone, 43% came from international institutions, and the rest reflected UK and international collaborations.
What this means
A few things stood out consistently across the sample:
- Policy actors and policy evidence dominate what gets referenced in policy documents, well ahead of peer-reviewed academic research.
- Some documents and document types are far richer sources of citations than others, and a handful of outliers account for a disproportionate share of all references.
- Mentions turned out to be the more common form of attention across nearly all reference types.
Taken together, these early findings suggest that citation counts alone may only capture a part of how research reaches policy.
Where we think this work is genuinely useful is in how it reframes the question. Right now, most tools only ask whether a policy document cited a piece of research or not. Splitting that into citations and mentions lets us ask a more useful question: how is research actually showing up, even when there’s no formal reference at all? That feels like a real gap worth exploring.
The coding framework itself, sorting references into different types of evidence and different types of actors, and tagging each one as a citation or a mention, is a first step towards making that visible. It gives us a way to describe the kinds of engagement that citation counts alone may miss, so it’s a starting point for more work.
This also fits with a broader pattern often noted in research impact literature: that scholarly expertise tends to reach policy indirectly, through reviews, consultations, and expert commentary, rather than through direct citation. That raises a genuinely open question: how much of what we’re seeing is a passing reference, and how much reflects real engagement with the research itself?
Being able to tell the difference matters if we want to understand not just whether research is being used, but how and why.
Where we go from here
This was a small, exploratory pilot, and we’re upfront about its limits. Fifty documents from one country over one year is a start, not a conclusion.
Next, we’d like to scale up the sample across a wider, more representative set of countries using an LLM to help with the coding. We’d also like to track whether the balance between citations and mentions is shifting over time and compare document types more systematically to understand whether citation behaviour is driven more by document function or by the authoring body. Beyond that, we want to develop ways to assess the depth of engagement behind a reference, not just whether it exists, and look at whether these patterns differ across policy domains, comparing something like health policy against education policy.
If you have thoughts on this work, or you’re doing something similar, we’d love to hear from you. Get in touch at nathalie@overton.io or kat@overton.io.
References
- O’Connor, R., Malde, S., Russell, W., & Haviland, M. (2026, March 23). We should focus less on research impact and more on research value. LSE Impact Blog. https://blogs.lse.ac.uk/impactofsocialsciences/2026/03/23/we-should-focus-less-on-research-impact-and-more-on-research-value/
About the Authors

Nathalie Cornée helps the research and policy community around the world get the most out of Overton. She joined in 2023, bringing more than 20 years of experience across the pharma, research, and academic library sectors, always focused on highlighting the significance of research and supporting the researchers behind it, including through bibliometric analysis. She takes a practical, hands-on approach to helping users find the data they need and make sense of research impact.

Kat Hart is a Senior Analyst at Overton, where she curates policy data and uses bibliometrics to explore evidence-use patterns worldwide. Alongside this, she’s a PhD researcher at the University of Nottingham, studying UK water governance. She previously worked as a knowledge broker in UK higher education, focused on ‘what works’ in connecting academic research with public policy.
Unless it states otherwise, the content of the Bibliomagician is licensed under a Creative Commons Attribution 4.0 International License.