You’ve tried it. You type something like “write a persuasive email about our new policy” and ChatGPT hands back something technically correct, reasonably structured, and completely forgettable. The output is fine. Nobody reads it twice. Nothing changes.
The problem isn’t the tool. The problem is the prompt. And more precisely: the problem is that most prompts don’t contain any behavioural thinking. They describe the task but leave out the psychology. The result is text that informs rather than activates, that looks like communication but doesn’t function as influence.
I have spent years studying why communication fails to change behaviour. The answer is almost always the same: it treats people as rational information-processors who just need the right facts delivered clearly. They aren’t. And ChatGPT, by default, reflects that same assumption back at you. Feed it better mental models and you get fundamentally better output. Below are the prompts that do that.
Why standard prompts fall short
When you ask ChatGPT to “write a persuasive text,” the model draws on patterns from persuasive writing: clear structure, strong opening, call to action. That’s reasonable. But persuasion in the behavioural science sense is something different. It means understanding which specific psychological forces are working for or against the behaviour you want to trigger, and designing your message to work with those forces rather than against them.
Without that specification, AI optimises for readability and completeness. These are not the same as behavioural effectiveness. A well-structured message can still fail entirely if it activates the wrong frame, asks for too much too soon, or neglects the anxieties that are quietly blocking action.
The prompts that follow give ChatGPT the missing ingredient: a behavioural lens. Each one is built around a specific mechanism from psychology and designed for a scenario that communication professionals face regularly.
Prompts for writing persuasive messages
1. The loss-framing prompt
Loss aversion is one of the most consistent findings in behavioural science. Losing something hurts about twice as much as gaining the equivalent feels good.[1] Most communication leads with gain: here is what you get, here is the benefit. The loss-framing prompt turns this around.
Prompt:
Write a [type of message] about [topic] using a loss frame. Show the reader concretely what they stand to lose, miss out on, or have taken away if they don’t act. Do not lead with what they gain. Use specific, vivid language. Tone: [insert tone]. Target audience: [insert audience].
A practical example: a communication team at a municipality needs residents to register for a neighbourhood improvement scheme before the deadline. Instead of “register and receive support,” the message becomes: “After the 15th, your street loses its place in the scheme. The funding goes elsewhere. The new pavement, the extra green, the lighting on the corner: gone for at least another three years.” Same information. Different psychological weight.
2. The social proof prompt
People look to others when they are uncertain. It is one of the most automatic and reliable tendencies in human behaviour. You can instruct ChatGPT to build social proof into the architecture of a message, not just tack it on at the end as a quote.[2]
Prompt:
Write a [type of message] about [topic] in which social proof is woven throughout the text as evidence. Show that others in comparable situations have already made this choice or taken this action. Use concrete numbers where possible. Do not place the social proof only at the end. Audience: [insert audience]. Tone: [insert tone].
The difference is structural. Most texts put a testimonial quote at the bottom. This prompt asks for social proof as the backbone of the message: “Eighty-three of the hundred teams in your sector have already completed this training. The majority did so in the first quarter. The ones who waited longest reported the most friction afterward.” That lands differently from a quote in italics below a headline.
3. The friction-reduction prompt
Most calls to action fail not because people don’t want to act, but because the path to action feels too effortful. This is what we call friction: every extra step, every moment of uncertainty, every unclear instruction that stands between intention and behaviour.[3] This prompt asks ChatGPT to design the CTA around removing that friction.
Prompt:
Write a call to action for [topic or product] that minimises perceived effort for the reader. The next step should feel small, specific and immediately doable. Avoid vague calls to action like “click here” or “learn more.” Tell the reader exactly what they do, how long it takes, and what happens next. Target audience: [insert audience].
The result shifts from “Register here” to “Fill in two fields. You’ll receive a confirmation within two minutes. Your seat is secured.” That specificity removes the ambiguity that causes hesitation. It is the difference between telling someone to “get in shape” and telling them to “walk for twenty minutes after lunch on Tuesday.” The second one actually happens.
Prompts for audience analysis
4. The pain-gain analysis prompt
Before you write anything, you need to understand what actually drives the people you are writing for. Not their demographics. Their psychology. The SUE | Influence Framework© maps this as pains and gains: the fears, frustrations and desires that determine whether someone moves toward or away from a behaviour. You can use ChatGPT to sketch this map before you write a word of actual copy.
Prompt:
I am a [role] working on [topic or project]. My target audience is [description]. Analyse the likely pains (fears, frustrations, barriers) and gains (desires, ambitions, motivations) of this audience in relation to [specific behaviour or decision]. List at least five pains and five gains. Then identify the two pains and two gains that are most likely to determine whether they act or not.
This exercise alone often reveals that what the organisation wants to communicate and what the audience actually needs to hear are two entirely different things. Most internal campaigns assume the audience shares the organisation’s enthusiasm. The pain-gain analysis shows whether that assumption holds.
5. The anxiety-mapping prompt
People often do not act not because they lack motivation but because they have unresolved anxieties about the step ahead. These are rarely stated explicitly. The product seems fine. The information is clear. And still, people hesitate. This prompt surfaces those hidden barriers.
Prompt:
My target audience is [description]. I want them to [specific behaviour]. List the ten most likely unspoken anxieties or objections that would prevent someone in this group from taking that step. Include both rational concerns (cost, time, risk) and emotional or social ones (embarrassment, fear of judgment, uncertainty about competence). Rank them from most to least common.
Once you see the anxiety list, you can address them directly in your communication. This is far more effective than repeating the benefits. A message that says “no experience needed” removes a specific anxiety. A message that says “join thousands of professionals who have already done this” removes a different one. Without the list, you are guessing which anxieties your message needs to resolve. With it, you can target them systematically.
Prompts for campaigns and policy communication
6. The reframing prompt
The framing effect shows that identical information produces different decisions depending on how it is presented. This is not manipulation; it is the reality of how human perception works. Communication professionals have always known this intuitively. The reframing prompt makes it systematic.
Prompt:
I have a core message: [your message]. Rewrite this message in five different frames: (1) a loss frame, (2) a gain frame, (3) a social norm frame, (4) an identity frame (who the reader wants to be), and (5) a concrete consequence frame. Keep each version to a maximum of three sentences. Target audience: [insert audience].
Run this prompt on any piece of copy you are unsure about. The five versions will immediately show you which angle carries the most weight for your specific audience. For a sustainability campaign aimed at employees, the identity frame might dominate: “People who care about the city they work in already do this.” For a health communication campaign, the loss frame might be stronger: “Every week you wait, the option becomes harder to reverse.” You test; you choose; you adapt. That is how effective communication is designed, not guessed at.
7. The behaviour change sequence prompt
Most campaigns try to do too much with a single message. They want to inform, persuade and activate all at once. Behaviour change rarely works that way. It moves through stages: from unawareness to awareness to consideration to action to habit. Each stage needs a different type of message.
Prompt:
I am designing a communication campaign to get [target audience] to [desired behaviour]. Create a four-stage message sequence: (1) a message for people who are not yet aware of the issue, (2) a message for people who are aware but not yet motivated, (3) a message for people who are motivated but not yet acting, and (4) a message for people who have acted once but need to repeat the behaviour. Each message should have a different psychological hook. Tone: [insert tone].
This prompt forces you to stop treating your audience as a single, undifferentiated group. It is one of the most common errors in campaign planning, and it is expensive. Writing four messages that meet people where they actually are is not more work in the end. It is less, because each message does its specific job instead of trying to do everything and failing at most of it.
How to make these prompts sharper
Every prompt above becomes more effective when you add three things.
First: a specific audience description. Not “employees,” but “mid-level managers in financial services who are sceptical of top-down change initiatives and have seen multiple failed transformations in the past five years.” The more concrete, the better the output.
Second: a constraint on length. ChatGPT defaults to comprehensiveness. Real communication benefits from brevity. Add “maximum 150 words” or “three sentences maximum” and watch the result sharpen considerably.
Third: an explicit instruction on what to avoid. If your organisation’s communication has a tendency toward jargon, passive voice or hedging language, say so. “Avoid corporate language, passive constructions and any sentence that could appear in a policy document” works well for most internal communication contexts.
The output of any prompt is a first draft, not a finished product. The value is speed and breadth: ChatGPT gives you ten versions in the time it would take you to write one. Your job is then to bring the behavioural judgement that selects, refines and applies the best version to the actual audience in the actual context. That judgement is what AI cannot replace. It is also what makes the difference between communication that informs and communication that changes behaviour.
If you want to go deeper into the behavioural mechanisms behind these prompts, the framing effect explained and the article on why informing doesn’t work are good starting points. And if you are working on broader change communication, the piece on change communication that actually works applies these principles to organisational contexts directly.
The quality of your AI output is determined by the quality of your mental model. Behavioural science gives you that model. These prompts put it to work.
