GPT for marketers

Start from what customers actually say and the numbers you already have, then let GPT draft the options, critique the page and write the plan you test next.

4 min readUpdated

Marketing copy works when it starts from what customers say and gets tested against real numbers. GPT.me helps at both ends: sorting feedback into themes, drafting variations and turning results into a plan. It's free, needs no account and runs OpenAI's GPT models in your browser. It reads images too, so you can show it an ad or a landing page. GPT-6.1 Terra suits the analytical work, while GPT-6 Luna is quick for drafts. The chat has no internet access, so it can't check competitors or this week's figures for you; paste what you have. Seven uses follow.

Research and messaging

Four uses that help you decide what to say before you spend money saying it. They work best with real input, so collect a page of customer comments or your current copy before you start.

Sort customer feedback into themes

Reviews, survey answers and support tickets already hold your best copy. Paste them in and let the chat group them.

Here are [number] customer comments from [source], one per line: [paste comments]. Group them into themes, name each theme in a few words and count how many comments fall under it. Quote two comments word for word for each theme, and tell me which theme is the strongest selling point and which is the most common complaint.

Find the themes. Counts over long lists can be off, so spot-check them.

Draft positioning options

Before you write headlines, decide what you want to be known for. Ask for several clearly different positions to choose between.

My product is [product] for [audience]. The main alternatives customers consider are [alternatives], and our real advantages are [advantages]. Write three different positioning statements, each built on a different advantage. For each, describe the customer it would attract, the claim we would need to prove and the risk of choosing it.

Compare positions with GPT-6.1 Terra and test the favourite with a few customers.

Write subject lines within a limit

Field limits matter in email and ads. Ask for the counts, then verify them, because language models often miscount characters.

Write ten email subject lines for [campaign] aimed at [audience], each under [number] characters, with a mix of plain benefit, curiosity and urgency. Show the character count after each line, and mark the two you would test first and why.

Get the subject lines and read them on a phone screen, where most emails are opened and long lines get cut. For posts, try the Social Media Post tool.

Critique an ad or landing page

A first impression is easier to judge from the page itself. Attach a screenshot and ask for a stranger's reading of it.

I've attached a screenshot of our [ad or landing page] for [product]. Imagine you are a [target customer] seeing it for five seconds. Tell me what you think we sell, who it is for and what you would do next. Then list the three changes most likely to make the offer clearer.

Open the critique and attach the screenshot. A phone-sized capture shows what most visitors really see.

Campaigns and reporting

Three uses for running the work and explaining what it achieved. Here the thinking model earns its extra seconds, because the questions involve numbers, assumptions and trade-offs.

Plan an A/B test properly

A test without a hypothesis and a stopping rule tends to prove whatever you hoped. Let GPT-6.1 Terra write the plan.

I want to test [change] on [page or email] to improve [metric]. Our current numbers are [traffic and conversion figures]. Write a test plan with a clear hypothesis, the primary metric, guardrail metrics, a rough estimate of how long the test must run and the rule for stopping it. Explain any assumption you make.

Plan the test with GPT-6.1 Terra, and confirm the duration with a proper sample-size calculator.

Repurpose one article for several channels

One well-researched article can feed a week of channels if each piece is cut to fit.

Here is a blog post we published: [paste post]. Turn it into a short newsletter section, three social posts of different lengths and one script for a 30-second video. Keep our tone, which is [tone], and make each piece work on its own without the article.

Repurpose the post and adjust each piece to your brand voice. Post the pieces a few days apart rather than all on the same morning.

Explain a month's results in plain words

Stakeholders want to know what happened and what you'll do next. Paste the numbers and let GPT-6.1 Terra draft the story behind them.

Here are our marketing results for [month] compared with [previous month]: [paste figures by channel]. Write a short report for [audience] that explains the three biggest changes, offers possible reasons without overstating them and proposes next steps. Flag any figure that looks inconsistent.

Draft the report, then add the context only your team knows.

What to avoid

  • Inventing proof. Never publish statistics, reviews or quotes that the model wrote. Use your own data.
  • Claims you can't back up. Ads and product pages fall under advertising rules, so check health, money and comparison claims carefully.
  • Asking for market news. The chat cannot see the internet or today's prices. Give it the facts you have.
  • Pasting customer data. Remove names, contact details and order numbers from feedback before you send it.

Got a question? Ask it now

No account and no download. Open the chat, type what is on your mind, and the answer starts appearing in a second or two.

Start chatting