Which tool to choose
They are not equivalent for this task.
Choose Claude (claude.ai) if the text is long or you need nuance. On the free plan you get its most capable model and a very large context window: you can paste an entire sales page, a long email, or a chapter, and Claude keeps it all in mind without losing the thread. For emotional writing it returns a more natural tone.
Choose ChatGPT (chatgpt.com) if you want more quick versions and to iterate rapidly. Here too the best model is available for free, but with a cap: after a certain number of messages within a few hours you get rerouted to a reduced version or invited to switch to Plus. It's fine for short texts (posts, slogans, email subjects) where you generate and discard quickly.
Already have a Microsoft 365 subscription or use Edge? Copilot serves as a fallback when you've used up the other quotas: same kind of engine, zero added cost.
For the rest of the guide I use ChatGPT as a reference, but the prompts are identical on Claude and Copilot: only the box where you paste them changes.
How to do it
Persuasion doesn't come from asking "make it nicer." It comes from giving the machine three things: the target emotion, the audience, and a rhetorical framework. Without these, the AI inflates the adjectives and nothing else.
- Open ChatGPT or Claude. From a browser or from the app the path doesn't change: you need an empty chat.
- Paste the prompt below and, in place of the marked lines, put your text and your data.
- Read the generated versions. Don't copy them wholesale. Fish out the sentences that make something click for you and ignore the rest.
- Ask for a targeted refinement on a single point (e.g., "make only the first sentence stronger").
- Recompose the final version by hand, mixing the best pieces.
The operational syntax with the PAS framework (Problem-Agitation-Solution, the structure that makes you feel the problem before offering the cure):
You are a copywriter expert in persuasive writing.
Rewrite the text I give you following the PAS framework:
1. Open by naming the reader's concrete problem.
2. Make the weight of that problem felt with a vivid detail.
3. Present the solution as a natural relief.
Target emotion: [e.g., relief after frustration]
Audience: [e.g., small business owners who handle their accounting on their own]
Tone: warm and direct, short sentences, no inflated superlatives.
Constraints: keep all facts, numbers, and names unchanged.
Do not invent benefits not present in the original text.
Give me 2 versions that differ from each other.
Text to rewrite:
"""
[paste your text here]
"""
Check: if the two versions seem like twins, the AI didn't understand the audience. Reply in the same chat: "the two versions are too similar, make them really different in tone: the first sober, the second warmer."
For texts where pure emotion matters (a speech, a letter, a fundraiser), replace PAS with this block:
Rewrite this text to move people, not to sell.
Use the "show, don't tell" method: instead of writing "it was hard,"
describe a concrete scene that makes the difficulty felt.
A single dominant emotion: [e.g., gratitude].
Keep the length similar to the original.
Avoid clichés like "an incredible journey" or "there are no words."
Text:
"""
[paste here]
"""
A concrete example
Marta runs a small Pilates studio and needed to rewrite the sentence on the site that was converting very poorly. Original: "We offer Pilates classes for all levels in a professional and welcoming environment."
She pasted the first prompt setting: target emotion "the confidence of someone afraid of not being up to it," audience "women over 40 who haven't done sports in years."
ChatGPT returned two versions. The first left her cold. In the second there was a sentence that struck her: the opening didn't talk about the classes but about the fear of "being the only one out of shape in the room." Marta discarded the rest, kept that opening, and stitched the ending back together by hand with her phone number.
Result on the site: "Are you afraid of being the only one out of shape in the room? Here there are no right or wrong levels: you start from where you are today." The sentence is hers in rhythm, but the AI found the emotional angle for her. Total time: nine minutes.
When it does NOT work (and how to fix it)
If the AI inflates everything with adjectives
You get a text full of "extraordinary," "unique," "revolutionary" that sounds fake. Fix: add to the prompt "no superlatives or enthusiastic adjectives allowed; persuade with concrete facts and strong verbs, not with adjectives." Real persuasion lies in specific details.
If it changes the facts or invents benefits
It happens that the AI adds guarantees, discounts, or features that never existed just to make the text more convincing. Fix: always keep the line "do not invent benefits not present in the original" in the prompt and reread, comparing with the starting text. On a commercial text this check is not optional.
If it stops you mid-job
You're iterating and ChatGPT stops or reroutes you to the reduced version of the model, which writes more flatly. Fix: switch to Claude to finish the session (it has a separate reset window) or pick it up later. Don't open double accounts: creating extra profiles to bypass the limits violates OpenAI's terms of service and risks permanent suspension.
If the tone stays the same as it was
You asked for "more emotional" but it comes out identical. Almost always the audience is missing. "Make it emotional" is an empty order; "make a parent enrolling their child in their first summer camp tear up" is an executable order. Rewrite the prompt putting a real person at the center.
A tip from someone who really uses it
Always generate two or three versions and never use a whole one. The point isn't to have the AI write the text: it's to have it find for you the angle you couldn't see on your own. Ninety percent of the output you throw away, but that one sentence that makes you say "there, that's what I wanted to say" is worth the whole round. Then you rewrite it in your own words, so the text stays yours in rhythm and no one senses the plastic of the AI.
Frequently asked questions
Which prompt works better, PAS or "show don't tell"?
It depends on the purpose. PAS is for when you need to push toward an action (buy, sign up, click): it puts the problem in front and the solution behind. "Show don't tell" is for when you need to make someone feel an emotion without selling anything (a speech, a thank-you, a story). Using PAS on a personal letter makes it sound like an advertisement.
Can I do everything with the free plan?
Yes, for most texts. Both ChatGPT and Claude offer their good models for free, with the limit of the number of messages within a window of a few hours. When you reach it, you get stopped or moved to a weaker model, then the quota recharges on its own. If you rewrite texts every day for work, the $20-a-month subscription removes the mid-session interruption.
Can the AI make even a long text persuasive, like a sales page?
Yes, and here Claude has a concrete advantage: even on the free plan you can upload files and use web search, with generous limits on the number and size of documents. You upload the entire page as a file and have it rewritten section by section without losing the context. If you hit an upload limit, split the document in two and work on them in separate chats.
Can a text written by AI be recognized and penalize whoever reads it?
This is the serious objection. A text left raw by the AI shows: uniform rhythm, predictable transitions, zero edges. But the problem isn't the AI, it's using it as a printer instead of as a consultant. If you take its output, cut the excess, add a true detail only you know, and break up a few overly smooth sentences, no one can tell anything, because at that point the text is yours. The machine only saved you from the blank page.