ChatGPT Prompts for Copywriters: Briefs, Ads and Revisions

Effective ChatGPT prompts for copywriters specify the audience, the offer, the desired action and the claims you can support. Ask for distinct copy directions, inspect their reasoning and revise the strongest option against those constraints. The prompt should make it harder for a polished line to introduce an unsupported promise.

The recurring challenge is keeping approved claims and brand choices available when a campaign spans several sessions. fenbs can hold reusable prompts, Decisions and rules, and the review record outside a private chat. Start by collecting your actual inputs; the templates below are designed to work with a short, explicit set of project instructions.

What belongs in the input before you request copy?

Separate what the audience needs from what the business wants to say. A feature list describes the product. A useful copywriting input also describes the reader’s situation, the question stopping them from acting and the action this particular piece should invite. Name the placement because an email subject line has a different job from a product page.

• Audience: their situation and the problem relevant to this message.
• Offer: what is included, what is excluded and where readers can verify details.
• Evidence: approved facts, source references and permitted customer quotations.
• Action: one next step and the page or process it leads to.
• Brand boundaries: tone examples, prohibited phrases, required wording and review needs.

Label missing details explicitly. Do not let the assistant fill gaps with a plausible guarantee, imagined testimonial or assumed feature. OpenAI’s prompting guidance↗ (opens in a new tab) recommends clear, specific instructions and refinement after review. Our AI prompts hub offers more structures for separating input material from the job you want performed.

Which prompt exposes gaps in the project instructions?

Before drafting, ask the assistant to identify uncertainties that would change the message. Prioritize questions whose answers affect accuracy or positioning. A request for twenty questions often produces busywork. Five consequential questions can reveal that the audience is unclear or that the only attractive promise has no supporting evidence.

Review these project instructions before drafting. Summarize the audience, offer and requested action. Then identify up to 5 gaps that could change the copy. For each explain what would be unsafe to assume and ask one precise question. Separate confirmed facts from proposed positioning. Do not write copy yet.
Project instructions: [paste]
Approved evidence: [paste]
Brand examples: [paste]

Answer those questions with the client or decision maker. Keep their answers beside the evidence, including limitations that might make a headline less dramatic. For a hypothetical scheduling product, faster booking is a proposed benefit until you know what the product actually does and what supports that description. The assistant can surface a question; it cannot approve the answer.

Our work prompt examples use the same separation between supplied information and requested output. When a person settles a recurring claim boundary, record that choice in fenbs Decisions and rules so later drafts start from the approved position.

How do you generate distinct copy directions?

Ask for a few concepts based on different audience concerns, rather than a large number of surface variations. One direction might emphasize the reader’s current difficulty; another might demonstrate the workflow; a third might answer an objection. Each should use the same approved evidence. You can then compare the underlying message before polishing individual words.

Propose 3 distinct directions for this placement. For each provide the audience concern, central message, headline, short body and requested action. Identify the approved fact supporting each factual claim. Keep required wording intact. Do not add urgency, guarantees, awards, customer results or product features unless explicitly supplied. Explain how the directions differ.
Placement and length: [paste]
Approved inputs: [paste]

Remove any concept that requires evidence you do not have. Compare the remaining ones on clarity, relevance and whether the destination fulfills the promise. A clever headline that needs a paragraph of explanation may be wrong for the placement. Select a direction yourself, then ask for variants within that direction instead of mixing incompatible messages into one draft.

What prompt makes revisions easier to review?

Turn general feedback into an observable editing goal. If a reviewer says the copy feels too corporate, ask what that means in this context: fewer abstract nouns, shorter sentences or a clearer account of what the reader can do. Preserve approved facts and required language while changing only the named issue.

Revise this approved direction to address only this feedback: [paste]. Give one minimal edit and one alternative with a different opening. Preserve the offer, factual meaning, required wording and requested action. After each version list changes and identify any claim whose meaning or strength changed. Do not silently remove qualifications.
Current copy: [paste]
Approved constraints: [paste]

Read the change explanation, then compare the actual text yourself. A change from can help to will can materially strengthen a promise even when the sentence gets shorter. Review headlines and buttons alongside the body copy, since a qualification buried below may not repair a misleading first impression. Keep accepted feedback as an example for the next round.

Use a reusable instruction for recurring editorial boundaries, with examples of acceptable and unacceptable wording. Our system prompt examples show how to make persistent instructions concrete. A rule such as never invent customer outcomes is more useful than a vague request to sound trustworthy.

How should prompts handle claims and testimonials?

Give the assistant a claim list with source references and ask it to map the draft back to that list. Missing evidence should produce a question or a removed claim, not a fabricated citation. Keep proposed copy separate from material approved for publication, and require a person to review changes that affect meaning.

Audit this copy against the approved evidence. List explicit factual claims and likely implied promises. For each identify the supporting source I supplied, any missing qualification and a question for the reviewer. Flag invented testimonials, unsupported comparisons and artificial urgency. Do not decide legal compliance or approve publication.
Copy: [paste]
Approved claims and sources: [paste]

The FTC’s small-business advertising guidance↗ (opens in a new tab) says advertising must be truthful and nondeceptive, and advertisers need evidence for claims. It also discusses implied claims and the impression an ad creates. Use that guidance to inform a human review; an assistant’s checklist cannot establish compliance for your particular campaign.

Keep testimonial wording faithful to the approved source. Do not turn a customer’s limited experience into a general product result. Check names, permissions and any required disclosures through the relevant review process. The useful job for ChatGPT is spotting a mismatch you might otherwise miss, not supplying a confident approval in place of evidence.

How can reusable prompts stay useful across clients?

Save the method separately from the client’s facts. Your template can ask for an audience concern, evidence and a requested action; its examples should not contain another client’s private material. Replace all client-specific inputs for each engagement, then verify that the resulting draft has not carried over an old promise or phrase.

After review, save the chosen direction and the reason it won. Note what remains unapproved so another assistant does not treat an exploratory draft as final. For campaign planning beyond the copy itself, use the separate marketing prompt guide.

How to do this with fenbs

1. Create a free account and add a project for one client engagement or campaign.
2. Save a reusable input review prompt and a claim review prompt in Skills and prompts. Keep private client material out of general templates.
3. Record the approved audience, claims and brand boundaries in Decisions and rules, identifying the person who chose them.
4. Connect your assistant from Integrations and ask it to read project context before applying the templates.
5. Save progress with the current draft stage, reviewer feedback, source checks still needed and the next action. Let a person make publication decisions.

fenbs keeps the shared work record; it does not approve advertising claims or replace your publishing tools. If a single draft has no recurring guidance or handoff, a document may be enough. Use the saved record when the same evidence and decisions need to survive several rounds of work.