How to write a scientific white paper

This guide is for researchers, labs, consortia, scientific societies and science or health agencies who need to turn a body of research into a document someone can act on. It is part of our series on white papers. For the basics, start with what a white paper is; for the general writing process, see how to write a white paper.

What a scientific white paper is (and is not)

A scientific white paper is a report that sets out the state of the research on a question and its implications, written for people who must make a decision. Those people might be a funder setting priorities, an agency drafting guidance, a hospital board choosing a technology, or a company deciding whether to invest in a research line.

It sits between three other documents, and it is easy to drift into each of them:

The scientific white paper borrows rigour from the review article and purpose from the policy brief. If yours has no recommendations, it is probably a review. If it has no account of evidence strength, it is probably a brochure. For the line between research writing and white papers in general, see white paper vs research paper.

Common reasons to write one:

• A research agenda: a consortium or society sets out the open questions in a field and what funding should target.
• The state of the science on a contested issue: an agency or society explains what is known about a risk, a technology or an intervention.
• A technology readiness case: a lab or company explains what the research shows about a method and what it would take to use it in practice.
• A position paper: a society states its view, with the evidence behind it, on a question of practice or policy.

Before you write: the question, the reader, the decision

Most weak white papers fail before the first sentence. Settle three things on one page and get your co-authors to agree to them:

1. The question. One sentence. "What does the evidence say about the effect of school-based physical activity programmes on attention in children aged 6 to 12?" is a question. "Physical activity and children" is a topic.
2. The reader. Name them. "Regional education officials deciding whether to fund a programme" leads to a different paper than "researchers planning trials".
3. The decision. What will the reader do differently after reading? Fund, regulate, adopt, wait, or study further. Your recommendations section answers this.

Write the reader and the decision into the introduction. It tells reviewers what to judge the paper against, and it stops scope creep.

A structure that works

This is the structure the White paper module in AI for Research uses for its "Scientific or research" kind, and it matches what most society and agency papers do.

Put the methods note near the end, not the start. Decision makers want the answer first; specialists will turn to it to check you. If you use a formal search, describe it the way you would in a review: databases, search dates, inclusion criteria. Our guide to systematic review search strategies covers how to document one.

Gathering the evidence

A scientific white paper is only as good as its evidence base. You do not always need a full systematic review, but you do need to be able to say how you found the studies and why you trust them.

1. Start from existing syntheses. Systematic reviews, meta-analyses and guidelines on your question save time and carry more weight than single studies. Our guide to levels of evidence and study types explains how they rank.
2. Search more than one index. Different databases cover different fields. A medical question needs PubMed; an engineering one needs arXiv and proceedings; a social one needs broader indexes.
3. Follow citations both ways. Look at what key papers cite and what cites them, to catch newer work and dissent. See citation snowballing.
4. Look for disagreement on purpose. Search for studies that found no effect or the opposite. A white paper that only cites supporting work is advocacy.
5. Record what you did. Keep a list of searches, dates and the studies you excluded and why. It becomes your methods note.

If you need help organising the reading, our guide to how to write a literature review applies directly to the "state of the science" section.

Grading the evidence and stating uncertainty

This is what separates a scientific white paper from every other kind. Readers need to know not only what the studies found, but how much weight each finding can bear.

Use a consistent scale. Pick one and say which in the methods note. Options:

• GRADE-style certainty (high, moderate, low, very low), used widely in health guidelines. It rates a body of evidence on risk of bias, inconsistency, indirectness, imprecision and publication bias.
• Calibrated language of the kind the IPCC uses, where words such as "likely" and "very likely" and levels of confidence have stated meanings.
• A simple three-level scale (strong, moderate, limited) with a one-line definition of each, for fields without an established system.

Match the verb to the evidence. "Shows" for consistent findings across good studies; "suggests" for a few or weaker studies; "is unknown" when the evidence is thin. Never "proves".

Give the denominator and the setting. "Reduced readmissions" says little; "reduced 30-day readmissions in three randomised trials in hospitals in high-income settings" tells the reader whether it applies to them.

Separate consensus from uncertainty in its own section. A table helps:

Say why uncertainty persists. Too few studies, short follow-up, different outcome measures, conflicts of interest, or a genuinely mixed effect. This is useful to funders: it tells them what research to pay for.

Writing for decision makers

Your reader is intelligent but not a specialist, and is short of time.

• Lead with the answer. Each section opens with its main finding in one sentence, then the detail.
• Plain words for technical ideas. Define each term once, in the text or the glossary. Keep units and numbers; drop jargon.
• One idea per paragraph, with the source cited right after the claim.
• Figures that carry one message each. A forest plot for a specialist, a simple bar chart for a board. Print the source under every figure.
• Make recommendations specific. "More research is needed" is not a recommendation. "Fund a trial of at least two years in primary care that measures [outcome]" is.
• Separate what the evidence says from what you advise. Readers should be able to accept your summary of the evidence and still disagree with your recommendation.

Peer input and review

A scientific white paper carries the authority of the people behind it, so the review process matters as much as the writing.

1. Name a writing group and decide authorship and contribution statements early.
2. Consult outside experts on the scope and the evidence, especially people likely to disagree.
3. Run an internal review of the full draft against the question and reader page you wrote first.
4. Run an external review for a society or agency paper, and say in the methods note who reviewed it.
5. Declare conflicts of interest and funding for every author and the paper as a whole. A white paper funded by a company with a stake in the answer must say so on the first page.
6. Check every citation against the source, not against your notes. See how to check a claim against the research.

A scientific white paper outline: example

Example (invented for illustration). A research consortium on urban heat writes for city health and planning officials.

Title: Urban heat and health: what the evidence says and what cities can do
Subtitle: A review of the research on heat risk, vulnerable groups and interventions

Executive summary (one page)
  - The question and why it matters now
  - Three main findings, each with its strength of evidence
  - Five recommendations for city officials, and three research priorities

1. Introduction and scope: who this is for; heat-health only, not energy
2. The state of the science
   2.1 How heat affects health (strength: [rating])
   2.2 Who is most at risk (strength: [rating])
   2.3 What interventions have been tested (strength: [rating] by intervention)
3. Where the evidence agrees and where it is uncertain (table)
4. Implications for city planning and public health
5. Recommendations (numbered; each with who should act)
6. How this paper was prepared: databases, dates, inclusion criteria, reviewers
7. Glossary
References

Every rating and number in a real version would come from the cited studies. The structure is the reusable part. For more finished examples across kinds, see white paper examples, and for the page layout itself, the white paper format guide.

Using AI for a scientific white paper, carefully

AI can save time on structure, first drafts and summarising papers you have read. It cannot be the source of your evidence. General chatbots have been shown to invent references and figures, and in a scientific white paper a single invented study can discredit the whole document. Our guide to how to write a white paper covers the risk in more detail.

AI for Research has a White paper module in the Writing Studio, built to keep the evidence honest:

• Choose "Scientific or research white paper" and the draft follows the structure above: executive summary, the state of the science with a strength-of-evidence statement per theme, where the evidence agrees and where it is uncertain, implications, recommendations, how the paper was prepared, and a glossary.
• It searches first. The module plans several searches from your topic, across sources you choose (all papers, medical, preprints, trials, datasets, web or your Library), and cites only what it found, as numbered markers you can open. Markers that point to no real source are removed.
• Your own data stays yours. Paste or upload your results, and they are used exactly as given. Missing numbers become [placeholders], never estimates.
• Write a section at a time. The executive summary, the evidence section or the recommendations can be redrafted on their own, at 3 credits each; a whole paper is 15 credits. The free plan has 45 credits a month. High quality mode, on the Pro plan, uses a stronger model for three times the credits.
• Build on a Deep Research result, so the white paper starts from a cited answer you have already checked.
• Finish in the editor, with your citation style and Word or PDF export.

What it does not do: it does not grade evidence with a formal GRADE assessment, run a systematic review, or replace expert review. It says how strong the evidence looks from what the sources report; your writing group must confirm each rating. See the scientific white paper page for how it works, or try the free white paper generator for a short draft without an account.

Checklist before you release it

• The question, reader and decision are stated in the introduction.
• The executive summary fits on one page and states how sure each finding is.
• Every finding has a citation and a strength-of-evidence rating on one stated scale.
• Consensus and uncertainty are in a section of their own.
• Recommendations are numbered, specific and name who should act.
• The methods note says how evidence was found, when, and who reviewed it.
• Funding and conflicts of interest are declared.
• Every citation was checked against the source itself.
• A non-specialist has read the summary and could explain it back.

FAQ

What is a scientific white paper?

A scientific white paper is a report that sets out what the research says about a question and what follows from it, written for people who must act on it: funders, agencies, health systems, companies or policy makers. It summarises the state of the science, says how strong the evidence is, separates consensus from uncertainty, and ends with recommendations for research or action.

How is a scientific white paper different from a review article?

A review article is written for other researchers and judged by peer review; its job is to synthesise a literature. A scientific white paper is written for decision makers and judged by whether it helps them decide. It is shorter on method, longer on implications, uses plain language, and ends with recommendations. Many white papers rest on a review, but they are not the same document.

How long is a scientific white paper?

Most run 3,000 to 8,000 words, or about 8 to 20 pages, plus references. Consortium and agency papers that set a research agenda can be much longer. Whatever the length, the executive summary should fit on one page, because many readers will stop there.

Who writes scientific white papers?

Research groups, consortia, scientific societies, science and health agencies, think tanks, and companies with R&D programs. A paper usually has a lead author or writing group, input from named experts, and a review round before release. Societies and consortia often publish the paper under the organisation's name with a list of contributors.

Can AI write a scientific white paper?

AI can draft the structure, summarise papers you supply and propose wording, but it must not supply the evidence. General chatbots can invent studies and numbers. Use a tool that searches a real index and cites what it found, keep your own data separate, and have experts in the field check every claim and the strength-of-evidence statements before release.

Next steps

If your paper explains how a system or method works rather than what the research says, read how to write a technical white paper. If it proposes government action, read our guide to the government white paper. To start from a blank structure, copy the white paper template.