Plain Language Summaries: How to Write One

Patients, carers and journalists may read your summary and never open the paper. Reaching them takes a rewrite in everyday words, not a trimmed-down abstract.

Guide · 11 min read · Directive Publications Editorial Team

What a plain language summary is for

A plain language summary tells readers without specialist training what your study asked, what you did, what you found, and how far the findings can be trusted. Those readers include patients, carers, journalists, policymakers, and researchers from other fields.

The abstract is written for specialists and leans on technical terms and statistics; the summary reaches the same conclusions in everyday words. Cochrane's guidance draws that line by rule, not tone: its summaries drop the abstract's set structure, use conversational language, and leave the statistics out.

If you have written a cover letter, the one or two plain sentences on what your study found are a good seed for the summary's first key message.

Know which kind you are writing

"Plain language summary" covers several different documents. Length, audience and rules change between them, so identify yours first.

This guide focuses on the first type. For trials under the EU Clinical Trials Regulation, the sponsor's lay summary of results is a legal requirement with ten content elements set out in Annex V. It is generally due no later than 12 months after the end of an adult trial, or 6 months for a paediatric one, with some exceptions. The European Commission's Good Lay Summary Practice guidance covers that process.

Check what your journal asks for

Before drafting, open the target journal's instructions for authors and note:

• The name. Author summary, lay summary, lay abstract: use the journal's term so the text lands in the right field of the submission system.
• The length. Treat the limit as a ceiling, not a target.
• The audience. PLOS Neglected Tropical Diseases ideally aims its author summary at an undergraduate student; EU trial summaries are written for the general public.
• Voice and format. The PLOS Computational Biology submission guidelines ask for a 150–200 word non-technical summary in the first person, distinct from the abstract, and note that the text is subject to editorial change. Taylor & Francis places its text summary between the abstract and the introduction, with no references.
• Timing. Check whether the summary is due with the first submission or later.

If the journal offers a summary but gives little guidance, the Open Pharma recommendations make a sensible minimum standard: text-based, 250 words or fewer, consistent with the abstract's key points and conclusions, and ideally reviewed by a non-expert. Text also helps discovery, because PubMed's XML format lets publishers tag a plain language summary alongside the abstract.

How to write one in seven steps

1. Check the journal's instructions. Settle the name, word limit, placement and audience before drafting a sentence: all four change what you write.
2. List the five facts. From your abstract and results, write down the question, what you did, the main finding (including any harms), the main limitation, and what it means.
3. Write a plain title. Cochrane recommends phrasing the title as a question where possible, using words readers would search for. If a technical term cannot be avoided, explain it in the title.
4. Explain why and what you did. In two or three sentences, say why the study was needed, who took part, how many people, and for how long. Explain any study design term you use.
5. Report the findings in plain numbers. Give the main results as frequencies such as "10 out of 100 people". Include unwanted effects, or say they were not reported.
6. State the limits and what the findings mean. Say how confident you are and why, and what should happen next. Do not tell readers whether to use a treatment.
7. Test it on a non-specialist reader. Ask someone outside your field to read the summary on its own, then rewrite whatever they misunderstood.

Write every sentence fresh. The European Commission's expert group advises against simply copying text from the technical summary.

Choose a reading level

There is no universal target, so decide based on who will read it.

• Patients and the public. The European Commission expert group says a well-written lay summary of trial results would normally be accessible to young people from age 12. Cochrane suggests a reading age of around 11 for UK audiences or those with similar literacy levels.
• A mixed audience. PLOS author summaries are written for scientists and non-scientists alike. Write for the reader furthest from your field.
• Readers in a second language. Cochrane asks writers to keep summaries easy to translate: avoid regional terms and abbreviations such as "e.g." and "i.e.".

Readability formulas such as Flesch–Kincaid, available in Microsoft Word, give a rough check, but do not chase the score: the Good Lay Summary Practice guidance notes that short, meaningless sentences can still score well. Aim for an average of about 20 words per sentence, as Cochrane suggests, and prefer the active voice: "we measured", not "measurements were taken".

Put numbers and certainty into plain words

The European Commission's lay summary guidance sets out these rules for numbers, and Cochrane's guidance agrees on the core points:

• Use frequencies, not ratios. The EC expert group tells sponsors to describe outcomes as "x out of xx people (xx%)" and to refrain from terms such as number needed to treat, odds ratio and confidence interval, or explain them simply if they must appear. Cochrane asks its authors not to report summary statistics and confidence intervals at all.
• Keep denominators and units consistent. Good Lay Summary Practice lists this among its health numeracy principles, so "1 in 10 people felt sick and 2 in 10 felt dizzy" beats "1 in 10" and "1 in 5".
• Prefer whole numbers, and do the arithmetic yourself. The same guidance favours whole numbers over decimals and says calculations should not be left to the reader: state the difference between groups rather than making readers subtract. Round only as far as you can without contradicting the paper, then check each rounded value against it.
• Write numbers as numerals. Cochrane asks for "3260 people" rather than the words, but never to open a sentence: recast it instead.
• Explain differences that could be chance. The EC expert group says a difference that is not statistically significant must be explained: the gap between groups is likely chance rather than an effect of the treatment.
• Signal certainty consistently. Cochrane's suggested wording says a treatment "reduces" an outcome when certainty is high, "probably reduces" it when moderate, and "may reduce" it when low. It also asks you to avoid the phrases "low-certainty" and "high-certainty evidence", which readers find hard to follow. Pick your qualifiers and use them consistently.

Cochrane's guidance gives this example from a review of e-cigarettes for stopping smoking:

A reader can picture 100 people and see how many benefit. An odds ratio cannot do that.

Replace jargon rather than deleting it

Plain language does not mean removing every medical term. Patients often know their condition by its technical name, so Cochrane suggests the plain phrase with the term in brackets, as in "blood thinners (anticoagulants)". Define a term once, then use the same word every time.

Take particular care with "significant". To a lay reader it means "important", which is not what a p-value tells you.

Mistakes that undermine a summary

AI policies differ. Taylor & Francis permits AI help with a summary of the authors' own work, but on conditions: keep the original and revised versions to share on request, and declare that the content is original and accurate, who checked it, the tool's name and version, that you read its terms of use, and that the authors take full responsibility for the content, references included. ICMJE says AI writing assistance should be reported in the acknowledgments.

Test it on a real reader

Cochrane, Open Pharma and Taylor & Francis all recommend that someone outside the field reads the draft. Cochrane names three candidates: a consumer who co-produced the review, someone with an interest in the topic such as a patient or carer, or someone who does not know much about it. Colleagues make poor testers because they already know what you meant.

Give the reader the summary alone, without the abstract, and ask what the study was trying to find out, what it found including any unwanted effects, and how sure the researchers are and why. Where an answer differs from what you intended, the text is the problem, not the reader. Rewrite that part and, if you can, test it on a second reader. Two answers matter most: a reader who repeats your finding but cannot say who it applies to has lost the population in a subordinate clause, and one who sounds more certain than your evidence warrants has met a qualifier too weak to survive a first reading.

• Confidentiality. Share only the summary, and ask the reader to keep unpublished findings to themselves until the article is out.
• Credit. Acknowledge lay reviewers who improved the summary. The ICMJE recommendations advise written permission from everyone you acknowledge, because acknowledgment can imply endorsement. Our guide to ORCID, affiliations, and authorship explains where acknowledgment ends and authorship begins.
• Pushback. If a reader wants a claim stated more strongly than your data allow, keep the accurate version and find a clearer way to say it.

Before you submit

1. The summary uses the journal's name for it, sits where the instructions say, and is within the word limit
2. Every number still matches the paper after rounding, and results use frequencies rather than ratios or confidence intervals
3. Unwanted effects are reported, or the summary says they were not reported
4. Certainty words such as "probably" and "may" are used consistently
5. No new findings, no treatment advice, no promotional language
6. No unexplained technical terms or acronyms
7. A non-specialist has read it and answered the three test questions correctly
8. Any AI assistance is declared the way the journal requires

Add the summary to your manuscript submission checklist so it gets the same attention as the abstract.

Where extra time pays off

The opening carries more weight than anything after it. Cochrane's guidance warns that the key messages, 2 to 3 bullet points at the top of the summary, might be the only part some people read, which is why it asks that they represent the evidence fairly rather than pick the most quotable result. Journal author summaries have no such section but the same problem: the first few sentences are what most readers take away. Make them carry the answer to your question, how confident you are, and who the finding applies to.

Directive Publications keeps its author guidelines and article-type requirements together on its For Authors page. Check the requirements for your article type there, and if you are unsure whether to include a plain language summary, contact the editorial office before you submit.