Content Writing · Copy & messaging
The Best Prompt for Writing Headlines
The best prompt for headlines starts from the content, not from headline formulas. This prompt has the model state the most valuable thing the piece delivers, choose four different angles, write three options per angle and then test each for accuracy, specificity, clarity and length. You get a table of headlines with character counts, a reasoned top three and a suggested pair to test against each other.
Headline Writing Prompt
Generates headlines from several distinct angles, checks each against accuracy and length tests, and recommends three with reasons.
- Works with
- ChatGPT, Claude, Gemini
- Difficulty
- Beginner
- Complexity
- Single reply
- Variables
- 6
- Version
- 1.0
Customize this prompt
Fill in what you know. Anything left blank stays as a [bracketed placeholder]. Nothing you type leaves your browser.
6 variables · fields marked * matter most
# Role You are a headline editor. You write headlines that tell the right reader what they will get, and that the piece then delivers. # What needs a headline - Where it will appear: [Where the headline appears] - Audience: [Audience] - Voice: [Voice] - Character limit: [Character limit] - Words or phrases that must appear: [Words that must appear] <content> [The content] </content> # Task 1. Read the content and write down, in one sentence, the most valuable specific thing it gives this audience. Then list the two or three concrete details in it that a headline could be built on: a number, a named method, a surprising finding, a clear outcome. Use only details that are in the content. 2. Choose four distinct angles for this piece. Possible angles: the outcome the reader gets, the problem or mistake it addresses, the most surprising detail, the question the reader is already asking, a plain description of what it is. Pick the ones that fit; do not force all of them. 3. Write three headlines per angle. Within an angle, vary the structure, not just the synonyms. 4. Check each headline against the tests below and drop or rewrite any that fail. 5. Recommend your top three and say in one line each who it is most likely to work for and why. # Tests every headline must pass - Accurate: the content fully delivers what the headline promises. No implied results, numbers or stakes that are not there. - Specific: it could not be attached to a different piece on the same topic. - Clear on first read to someone outside the team, with no internal jargon and no wordplay that hides the subject. - Within the character limit, with the most important words early in case it is cut off. - Suited to the placement: a subject line reads like a message from a person, a page title describes the page, a social headline makes sense without the surrounding context. # Constraints - No curiosity gaps that withhold the subject, no manufactured urgency, no all-caps emphasis. - Include the required words naturally. If they make a headline worse, say so. # Output format The core value sentence and the details list. Then a table with columns: angle, headline, character count. Then the top three with reasons. Finish with one pair of headlines worth testing against each other, and what the result would tell me.
Use it when you are
- Titling a finished blog post or article
- Writing subject lines for a newsletter issue or announcement email
- Producing headline variants for a landing page test
- Getting unstuck when every title you write sounds the same
Why this prompt works
Reading comes before writing
Step 1 makes the model extract the core value and the concrete details from your content before it writes a single headline. Those details are the raw material; without them, headlines fall back on templates that fit any article on the subject. Limiting it to details that are in the content also guards against promises the piece cannot keep.
Angles produce range that a number alone does not
Asking for twenty headlines usually returns one idea reworded twenty times. Asking for four angles with three headlines each forces different ways into the piece: outcome, problem, surprise, question. You end up comparing approaches, which is the decision that matters when choosing a headline.
The tests are written out, so the model can self-check
Accuracy, specificity, clarity, length and fit for the placement are stated as pass or fail checks, and the model is told to drop or rewrite failures. That is more reliable than asking for good headlines, because each test says what to look for.
It ends with a test, not a verdict
No one can know in advance which headline readers will choose. The top three come with reasoning about who each suits, and the closing suggestion pairs two headlines that differ in angle, so a test between them teaches you something about your audience rather than about word order.
How to use it
- Finish the content first. Headlines written before the piece tend to promise something it does not deliver.
- Paste the full text if you can; otherwise a summary that includes its numbers, methods and conclusions.
- Set the character limit from where the headline will be shown, and count any that come back close to it.
- Look at the angles before the individual headlines and decide which approach fits the piece.
- Ask for five more options in the angle you prefer, or for a shorter version of a headline you like.
- If you can test, run the suggested pair and keep a note of which angle won for next time.
AI model compatibility
| Model | Status | Notes |
|---|---|---|
| ChatGPT | Expected to work | Character counts reported by a model can be off by a few characters. Check any headline close to the limit yourself. |
| Claude | Expected to work | The content is wrapped in XML-style tags as Anthropic's guidance suggests for pasted material. A summary works, but the full piece gives it more concrete details to build on. |
| Gemini | Expected to work | If the angles come back too similar, name the two you want replaced and ask for alternatives; the angle list in step 2 is a menu, not a fixed set. |
“Expected to work” means the prompt uses plain, model-independent structure and has been reviewed for that — not that it has been run on that model. A model is marked Tested only when a test record is published below.
Worked example
Scenario: A freelancer-focused blog has finished a practical article on getting invoices paid. Illustrative inputs, not recorded output.
# Role You are a headline editor. You write headlines that tell the right reader what they will get, and that the piece then delivers. # What needs a headline - Where it will appear: Blog post or article title - Audience: Freelance designers who invoice clients monthly - Voice: Plain and informative - Character limit: 60 characters - Words or phrases that must appear: late invoices <content> Article summary: A step-by-step process for freelance designers to deal with late invoices. Covers: stating payment terms in the proposal, sending the invoice the day work is approved, a three-email reminder sequence (templates included) sent at 1, 7 and 14 days overdue, when to pause work, and when to charge a late fee. Based on the author's own practice as a freelancer. </content> # Task 1. Read the content and write down, in one sentence, the most valuable specific thing it gives this audience. Then list the two or three concrete details in it that a headline could be built on: a number, a named method, a surprising finding, a clear outcome. Use only details that are in the content. 2. Choose four distinct angles for this piece. Possible angles: the outcome the reader gets, the problem or mistake it addresses, the most surprising detail, the question the reader is already asking, a plain description of what it is. Pick the ones that fit; do not force all of them. 3. Write three headlines per angle. Within an angle, vary the structure, not just the synonyms. 4. Check each headline against the tests below and drop or rewrite any that fail. 5. Recommend your top three and say in one line each who it is most likely to work for and why. # Tests every headline must pass - Accurate: the content fully delivers what the headline promises. No implied results, numbers or stakes that are not there. - Specific: it could not be attached to a different piece on the same topic. - Clear on first read to someone outside the team, with no internal jargon and no wordplay that hides the subject. - Within the character limit, with the most important words early in case it is cut off. - Suited to the placement: a subject line reads like a message from a person, a page title describes the page, a social headline makes sense without the surrounding context. # Constraints - No curiosity gaps that withhold the subject, no manufactured urgency, no all-caps emphasis. - Include the required words naturally. If they make a headline worse, say so. # Output format The core value sentence and the details list. Then a table with columns: angle, headline, character count. Then the top three with reasons. Finish with one pair of headlines worth testing against each other, and what the result would tell me.
What a good response should contain
- A core value sentence about a repeatable process and reminder templates for getting paid
- A details list that includes the three-email sequence and the 1, 7 and 14 day timing
- Four angles, for example outcome, problem, specific detail and reader question, with three headlines each
- No promises of guaranteed payment or statistics about late payment, since the article contains none
- A suggested test between a detail-led headline and a problem-led one
This is an illustrative scenario showing how to fill in the prompt and what to look for. It is not a recorded model output. Recorded outputs will be published here once this prompt has a test record.
Testing and editorial review
Status: Untested. Reviewed for structure and clarity by the editorial team. No recorded model test yet.
No test record has been published for this prompt yet. When it is tested, this section will show the model and version, the date, the inputs, the actual output and notes against our seven-point rubric. We do not publish numerical scores.
| Author | TheBestPromptFor Editorial Team — The editorial team writes, reviews and maintains the prompts and guides on TheBestPromptFor. Prompts are drafted with AI assistance and reviewed against published editorial standards. |
|---|---|
| Published | November 15, 2026 |
| Last reviewed | November 15, 2026 |
| Last tested | Not yet tested |
| Version history |
|
Limitations and things to check
- The model cannot predict which headline will perform. Its recommendations are reasoned guesses until you test them.
- Reported character counts may be slightly wrong. Verify any headline near your limit.
- It has no knowledge of what your audience has responded to before unless you tell it.
- For search results, a search engine may display a different title from the one you set. This prompt does not handle title tag optimization.
Alternative prompts
SEO
Title Tag and Meta Description Prompt
Writes unique, accurate title tags and meta descriptions for a batch of pages, with options, character counts and a check against what each page says.
- ChatGPT
- Claude
- Gemini
Content Writing
Conversion Copywriting Prompt
Works out what a page must make the reader believe, then writes the copy in that order using only the proof you supply.
- ChatGPT
- Claude
- Gemini
Marketing
Ad Copy Prompt
Generates ad variations across four distinct angles, counted against your character limits, with every claim traced to proof you supplied.
- ChatGPT
- Claude
- Gemini
Content Writing
Blog Post Writing Prompt
Turns your own notes and experience into a focused blog post with one clear argument, and marks every spot where it needs a real example or a fact check.
- ChatGPT
- Claude
- Gemini
Related resources
Guide
How to Improve AI Responses
A diagnostic approach to weak AI answers: name the fault, trace it to the gap in your prompt, then follow up precisely or rewrite and start again.
Technical guide
How to Test AI Prompts
A seven-criterion rubric and a nine-field test record for evaluating prompts, plus why this site describes results in words and publishes no scores.
Frequently asked questions
How many headlines should I generate before choosing?
Enough to see real alternatives. Twelve across four angles is usually sufficient to find one or two approaches worth refining. Past that point, more options from the same angles add little; asking for a new angle is a better use of another request.
Does this work for email subject lines?
Yes. Choose email subject line as the placement and set a short character limit, since many inboxes cut subject lines off. The placement test in the prompt tells the model a subject line should read like a message from a person, not like an article title.
What is wrong with curiosity-gap headlines?
A headline that withholds the subject can earn a click, but the reader arrives without knowing whether the piece is for them, and feels misled if it is not. Over time that costs trust and open rates. This prompt keeps curiosity that comes from a specific detail and excludes the kind that hides the topic.
Sources and references
- Prompt engineering guide — OpenAI
- Influencing title links in Google Search — Google Search Central
Was this prompt useful?
No account needed. Do not include personal information.