Session purpose
This session explored a practical problem at the heart of think tank credibility: even when research is rigorous, useful and relevant, it may not be visible, accessible or trusted by the people — and increasingly the machines — that need to use it.
The session asked how think tanks can build a stronger trust infrastructure around their work. This included questions of research visibility, funding transparency, peer review, content verification, digital trust marks and the role of AI in determining what knowledge is surfaced, cited and believed.
Main conveners and contributors
This was Parallel session 2E: “Your Research Is Good, But Does Anyone Know That? Building a Trust Infrastructure for Think Tanks.”
It was convened by:
- Toby Green, Publisher, Policy Commons
- Nick Cleveland-Stout, Research Associate, Quincy Institute
The programme framed the session around a growing problem: many think tanks produce rigorous research, but policymakers, publics and AI systems often struggle to know which organisations and outputs are trustworthy. This can favour already visible institutions and leave smaller or less prominent think tanks overlooked, even when they produce high-quality work.
Main presentation and framing
The session began from the observation that think tank research faces two linked challenges.
First, visibility: research is often produced but not read, reused or discovered by the right audiences. Large reports may be downloaded but not absorbed. Non-English outputs may remain invisible internationally. Younger audiences may search through AI tools rather than traditional databases, websites or newspapers.
Second, trust: users often lack clear signals about whether a think tank output is independent, reviewed, transparent and reliable. Unlike academic journals, think tank outputs are part of the grey literature ecosystem, where quality can be very high but review and verification processes are often opaque.
The session therefore proposed that the sector needs better systems to make quality, transparency and provenance visible — not only to human readers, but also to AI tools.
Main debates
1. Good research needs curation and translation
A key discussion focused on the gap between producing research and making it usable. The World Bank example showed how large, technical reports can be repackaged into companion outputs, multilingual podcasts and more accessible formats to “walk the last mile” between publication and use.
This was not presented as a lowering of standards, but as a necessary part of research impact. If people cannot access, understand or use research, its quality alone is not enough.
2. Language and visibility shape whose knowledge travels
Participants raised the problem of non-English research, including Portuguese-language research from Brazil’s IPEA. Even where there is significant high-quality output, international visibility can be limited by language, indexing and dissemination channels.
This has implications for global policy debates. If AI systems and international knowledge platforms mainly surface English-language or already dominant sources, they risk reinforcing existing inequalities in whose knowledge counts.
3. Funding transparency is central to credibility
The session discussed the damage caused by opaque funding. One presentation noted that 40% of top US think tanks do not disclose donors, including some organisations that frequently engage in policy debates and congressional testimony.
This lack of disclosure undermines public trust and makes it harder to distinguish independent research organisations from lobbying or advocacy vehicles with hidden agendas. The discussion also raised concerns about funding from defence contractors, foreign governments and politically aligned donors.
4. Peer review practices are uneven and often invisible
A major trust gap concerns review. Think tanks may have strong internal quality assurance processes, external reviews, seminars, editorial checks or advisory committees. But these practices are rarely standardised or clearly communicated.
This creates a problem for users. A policymaker, journalist, academic or AI system may not know whether a report was lightly edited, internally reviewed, externally peer reviewed or formally approved by a research committee.
The session proposed a possible four-level review framework to describe different levels of scrutiny and make review intensity more transparent and machine-readable.
5. Grey literature needs better trust signals
Think tank outputs often sit in the category of grey literature. This can lead academics and other users to treat them as less credible than journal articles, even when they are rigorous, timely and policy-relevant.
The session challenged this assumption. The issue is not that grey literature is necessarily weak, but that its quality signals are inconsistent. Better metadata, review disclosure, funding transparency and provenance could help users assess think tank outputs more fairly.
6. Digital trust marks could help users and AI assess reliability
A central proposal was the creation of a digital trust mark: a clickable, machine-readable token attached to think tank content. This would certify key information about the organisation and the output, including governance, funding transparency, review level, authorship, provenance and correction mechanisms.
The trust mark would travel with reports, blogs, podcasts and other outputs wherever they are shared. Clicking on it would reveal an independent certificate with relevant trust information.
The session also introduced the idea of a Content Object Identifier, similar to a DOI, to provide persistent identifiers for think tank outputs and reduce the problem of broken links or disappearing content.
7. AI makes trust infrastructure urgent
The discussion connected these issues to AI and large language models. Current AI tools often struggle to distinguish between high-quality, reviewed research and weak or unreliable content because trust signals are not consistently embedded in metadata.
This creates risks: AI may cite poor sources, miss strong research, hallucinate evidence or treat all web content as roughly equivalent. A trust mark system could help AI tools identify and prioritise verified content, improving the quality of AI-generated answers and reducing misinformation.
8. Governance will determine whether trust marks are trusted
Participants recognised that a trust mark system would only work if its governance is credible. It cannot simply be a badge sold by a platform or claimed by organisations themselves.
The proposal included an independent board involving libraries, publishers and think tank representatives, with separate oversight and implementation functions. This governance question is central: the trust infrastructure itself must be trustworthy.
Main takeaways
- Research quality is not enough. Think tanks also need to make their work discoverable, accessible, understandable and verifiably trustworthy.
- Visibility is uneven. Language, platform access, dissemination formats and search behaviour shape which research is found and used.
- Funding transparency is a foundational trust signal. Organisations that do not disclose donors contribute to wider scepticism about the sector.
- Think tanks should be clearer about how their outputs are reviewed. Review processes do not need to be identical, but they should be visible.
- Grey literature needs better quality signals so that think tank research is not unfairly dismissed or blindly trusted.
- AI increases the urgency of trust infrastructure. If AI systems cannot distinguish credible content from unreliable content, think tanks risk losing control over how their knowledge is interpreted and used.
- Digital trust marks could help both humans and machines assess authenticity, provenance, review level and organisational credibility.
- Trust marks will only work if they are governed independently and adopted widely enough to become meaningful.
- The future of think tank credibility may depend not only on producing better research, but on building the digital, institutional and governance systems that allow that research to be recognised as trustworthy.