Writing a Headline Your Editor Will Keep

Key Takeaways

• Editors keep headlines that identify the genre of a piece and promise only what the body supports.
• Length, page layout, and institutional branding do not classify a publication; the publisher's product label plus your own functional reading do.
• AI rules belong to the outlet. Contributor guidelines commonly expect a record of prohibited uses, permitted uses, disclosure location, required detail, and the date of the policy.
• Draft the headline after you can state the opening move, central claim, evidence pattern, uncertainty treatment, and reference format.
• Outlet fit decides register. A newsbrief headline, an op-ed headline, and a journal headline follow different rules.

1. Introduction

Most headline drafts die in the same place. The editor reads the headline, decides it describes a different kind of article than the one in the file, and rewrites it. The writer then spends a day arguing about word choice when the mismatch was structural.

Writers publishing into policy and research outlets run into a second problem. Each publication applies its own rules on submission format, on evidence, and on whether an AI tool may be used at all. Those rules rarely match what a tool vendor advertises. A headline written under one outlet's assumptions can fail at another for reasons that have nothing to do with the words.

This article covers the sequence that produces headlines editors keep: classifying the output before writing, building the headline from a claim the body can defend, applying outlet-specific rules, and matching register to the venue. It draws on a curated research collection about think-tank publishing, with material collected up to 16 September 2026.

2. Classify the output before you write the headline

A headline is a genre label that readers and editors process first. Get the genre wrong and no wording survives review.

Publishers label their products loosely. A program page can list reports, briefs, commentary, testimony, interviews, and book reviews side by side, and those labels do not travel well between institutions. Store the publisher's own label, then add a functional classification of your own: does the piece make an argument, report a result, interpret someone else's work, or record a conversation?

Reading cards do the work here. For ten recent pieces from a target outlet, record the opening move, the central claim, the evidence pattern, how the author handles uncertainty, the conclusion, and the reference format. Ten cards give you the outlet's real conventions, which is more reliable than imitating the wording of one article. The cards also show what the headline has to carry. If most pieces open with a policy decision and then examine its consequences, a headline built around an abstract concept will read as foreign.

Three classification traps catch experienced writers:

• Length alone does not classify a piece. A long commentary can remain an opinion essay, and a short note can report a substantial empirical result.
• Visual design and an institutional logo say nothing about the review process behind a piece.
• An interview stays an interview. Converting it silently into a research article changes what the editor is evaluating and what the headline must promise.

Before drafting, write one sentence for yourself: "This is a [genre] that [claim]." If both slots cannot be filled, the genre is not settled and the headline will keep changing.

3. Make the headline a claim the body can defend

Every noun, number, and verb in a headline is a promise the editor tests against the text. Headlines get cut when the body cannot pay.

Run the element test. Take each part of the headline and point to the sentence in the body that supports it. A figure in the headline needs the figure, its source, and its date in the body. An actor named in the headline needs that actor's decision or statement. Where the body hedges, as in "early indications," the headline cannot assert.

Verification is the author's job regardless of tooling. An author verifies references and keeps control of the argument whether or not a particular tool is permitted. A disclosure statement does not cure an invented citation or an unsupported quotation, and editors check both before they check your prose.

Keep a private process log while preparing the manuscript: tool, version where available, date, task, material supplied, and how the result was checked. Two things come out of it. You can make an accurate declaration when the outlet asks for one, and you can trace a factual claim or a citation back to a specific step if it is challenged.

Outlets also distinguish generating analysis from translating, editing, or checking language. A tool's marketing description does not decide how an editor classifies its use. Describe the task you performed, and let the editor place it in the outlet's categories.

Three checks before you send:

1. Claim match: can you underline the sentence that supports each headline element?
2. Hedge match: does the headline's certainty match the body's certainty?
3. Evidence match: does the reference list support the headline's strongest noun or number?

4. Apply outlet-specific rules before you finalize the headline

AI and submission rules sit with the publication and change over time. For each prospective submission, record five things: prohibited uses, permitted uses, disclosure location, required disclosure detail, and the date of the policy. The date matters because a policy captured two years ago tells you little about the current one.

Use a source hierarchy when you research submission rules. Current publication-specific guidance comes first. Institutional pages serve for missions, programmes, and stated policies. Financial documents are read against their accounting periods. External sector resources supplement primary material for comparisons. No inferred email address and no guessed submission policy should be treated as confirmed. If the outlet has not published a submission route, you do not have one.

Build a submission file per outlet:

Refresh the file when you return to an outlet after a few months. If the outlet restricts generated text, the headline counts as part of the submitted artifact, so the disclosure decision covers it.

5. Outlet comparison and headline fit

These comparisons come from a curated research collection covering mainly European and US institutions and their publication channels. It is a curated collection, not a census or a ranking, and depth varies because institutions disclose different amounts of information. Check the current contributor page before relying on any entry.

Headline register follows the same split. The table below maps output type to what the headline has to do.

Verb choice carries genre. "Reports" belongs to an empirical note, "argues" to commentary, and "warns" commits you to a source's warning. Pick the verb that matches what the piece does, then check the outlet's recent headlines for length and punctuation.

6. FAQ

Q1. How long should a headline be?

Outlet and channel decide. Some publications state a character cap, others leave length to the editor, and the same piece may need a shorter headline for a newsletter than for the site. Read the last twenty headlines the outlet published and match that range. When you have to cut, remove qualifiers and the second clause before you remove the subject.

Q2. Can I use an AI tool to draft or polish a headline?

It depends on the outlet's policy, which is why the five-item record matters: prohibited uses, permitted uses, disclosure location, required detail, and the policy date. Generating analysis, translating, editing, and checking language are treated differently at many outlets. Whatever you use, verify the references and keep the argument yours. A disclosure does not cover an invented quotation.

Q3. My editor replaced my headline with a flatter one. What went wrong?

Common causes: the headline promised a claim the body did not make, it used a genre signal from a different output type, or it overstated the certainty of the evidence. Compare your version against the body sentence by sentence. If the editor's version is flatter and still accurate, the flatness came from the evidence rather than from editorial taste.

Q4. Should the headline include a number?

Include a number when the piece reports a discrete result that the body documents with a source and a date. Leave it out when the figure is illustrative or borrowed from someone else's summary. A number in a headline is the easiest element for a fact-checker to test.

7. Conclusion

Editors keep headlines that survive two tests: they match the genre of the piece, and they promise only what the body supports. Length and polish matter far less.

The working sequence runs from classification to compliance. Classify the output, note the publisher's label and your own functional reading, build reading cards from recent pieces, draft the headline from the central claim, then check the outlet's current submission and AI rules with the policy date attached. Keep the process log while you write. It takes minutes and settles most disclosure questions before they reach an editor.

A submission file for three target outlets, each with a dated policy snapshot and twenty recent headlines, removes most of the guesswork from the next draft.