Selling Your Prompts: The Honest Guide, With No Income Promises

Let's compare two prompts. The first: "Write an engaging LinkedIn post about [TOPIC]." The second was refined by an accountant over several weeks: it turns an anonymized trial balance into a summary note for a non-finance business owner, spots the line items that moved, and writes the questions to ask before the meeting.

The first is worth nothing on a marketplace, because anyone can write it in twenty seconds. The second contains something the buyer doesn't have: a professional method, tested on real cases. If you want to sell your prompts, everything comes down to that gap.

Let's say it right away: this guide promises you no income. Nobody can tell you how much you'll sell. Selling prompts is a recent activity, and the easy-income claims you see floating around aren't a basis for a decision. What it will do is help you judge whether it's worth your time, and do things properly if you decide to go ahead.

What the buyer really pays for

A prompt is still just text, and the buyer could write one themselves. So they're paying for something else:

• Tuning time. The failed attempts, the variants discarded, the edge cases you've already run into.
• Knowledge of the trade. Knowing what a good result contains, what a client expects, the mistakes the model makes in this field.
• Proof. A real result you can see, on a realistic case, showing that the prompt delivers what it promises.

A simple test: try to write a decent version of your prompt in five minutes, without your experience. If you can, a buyer can too, or will simply ask the AI to do it for them.

Keep in mind that AI is improving fast. A recent model understands vague requests better, so prompts that merely word a simple request well lose their appeal. Those that encode a method, an analysis framework, or professional requirements hold up better. This reasoning isn't a forecast, just a reason to aim for depth.

The prompts that have a chance, and those that don't

We have no data to show you on what sells best, and you should be wary of anyone who claims to have reliable data. Here are common-sense criteria instead.

The best starting point is almost always a prompt you already use. A restaurant owner who turns each week's deliveries into the daily menu and a social media post has already done most of the work. They know which ingredients cause problems, which tone their clientele likes, how long a line fits on a chalkboard. That knowledge is worth more than sophisticated wording.

Look at your work week: what task are you already handing to AI, with a prompt you've corrected several times? Start with that one.

Make a prompt sellable, then test it

Turn it into a tool

A prompt that works for you isn't necessarily ready for a stranger. It relies on things you know that the buyer doesn't. You have to turn it into a tool, following the principle of reusable templates with variables: everything that changes from one use to the next becomes a clearly named [VARIABLE], and everything that stays true stays in the text.

Take a tiler who writes: "Write the job report for Ms. Perrin's bathroom, porcelain stoneware tile installation, two-day delay because of the delivery driver." Generalized, the same prompt becomes: "Write the job report for the [TYPE OF JOB] for [CLIENT], from these notes: [DAY'S NOTES]. Mention any delay with its cause, without blaming anyone, and end with the next visit." The instruction "without blaming anyone" comes from his experience. That's the part with value, and it stays in the text.

You can get help with this step.

Here's a prompt I use for my own work. I want to make it usable by someone who doesn't know me and works in the same trade.

Prompt:
"""
[YOUR PROMPT]
"""

Do things in this order:
1. Spot everything specific to my case (names, numbers, place, sector) and replace it with a variable in [CAPITALS IN BRACKETS], with an explicit name.
2. Spot what the prompt implicitly assumes and the user won't know, then add the missing instruction.
3. Add an instruction telling the model to ask questions if a variable isn't filled in, instead of guessing.
4. Keep my method and my professional requirements without weakening or simplifying them.

Give the final prompt, then the list of variables with an example value for each.

Read the result line by line. The model tends to smooth over the most specific requirements, yet those are exactly what gives the prompt its value. Also check that no real client data is lingering in the text or in the example.

Test it as if the buyer were in front of you

A paid prompt has to work for someone else, on another case, perhaps with another tool. Our method for testing a prompt details this step. The essentials come down to a few points:

1. Run it on several different real cases, including a difficult or incomplete one.
2. Try it in a fresh conversation, without your custom instructions or your memory, which can hide its weaknesses.
3. If you state that it works on a tool, test it on that tool, and note which one.
4. Have someone else try it, without explanations. Watch where that person gets stuck.
5. Note what it can't do. That will go in your listing.

A critical test can also complement these runs.

You're a demanding buyer, [BUYER'S PROFESSION], who has just paid for this prompt to [TASK].

Prompt:
"""
[YOUR PROMPT]
"""

Response obtained with this prompt on a real case:
"""
[RESULT]
"""

Evaluate without mercy:
1. What could you have gotten on your own with a simple request? Be frank.
2. Where does the result remain generic or inaccurate for someone in this trade?
3. Which instructions are ambiguous or likely to give a different result from one time to the next?
4. In which cases would this prompt fail?
5. What are the three changes that would add the most value?

Don't invent qualities. If the prompt doesn't add much, say so.

The opinion produced is still an AI's: it gives you leads to work on, not a verdict. Use it before publishing, and not to write a sales pitch.

The platform's concrete rules

If you choose to sell on YouPrompt, here are the rules, with no sugarcoating.

Account. Publishing requires an account with a verified email address.

Price. You set your price, between €1.99 and €99.99, and you receive all of it. A service fee is added on the buyer's side: 20%, with a minimum of €0.99. The price shown to the buyer includes everything. Our article on the price of a prompt shows what that looks like at different amounts.

Purchase. The buyer sees an excerpt and a real result before buying. They pay once, with no subscription, through Stripe, and the full text appears after payment.

Final sale. Because the excerpt and the result are visible in advance, there are no refunds, apart from legal guarantees. This rule creates a responsibility for you: your listing must honestly describe what the prompt does.

Payout. Authors are paid on the 1st of the month, starting at €20 in earnings.

The catalog is still young. It would be dishonest to let you believe a buyer is waiting for your prompt as soon as it goes live.

What you can reasonably expect

The most cautious scenario is also the healthiest: a published prompt may earn nothing for a long time, or even never. Make your decision starting from that assumption.

Under these conditions, the useful question is: is the work worth it even if sales don't follow? Often the answer is yes if you start from a prompt you already use, because making it more solid serves you first. You end up with a better-tested tool for your work, and a small showcase of your expertise. On the other hand, spending weekends making dozens of prompts for trades you don't know, hoping for extra income, is a bet nothing justifies.

Give yourself a limit too: for example a single prompt, finalized over a few evenings, before preparing a second one. If nothing sells, you'll have lost little time and you'll keep a sturdier tool.

Don't quit a job, borrow money, or commit to expenses counting on these sales.

Rights and taxes: questions to ask yourself

These topics depend on your situation, and what follows isn't legal or tax advice.

Publish only what comes from you. Reusing a prompt found elsewhere, or the text from a course you paid for, can cause you problems, and above all it's not honest toward the buyer. Copyright protection for a prompt is far from obvious and depends on the case: if the subject matters to you, consult a professional.

Look at the examples you display. A result containing a client's data, or a brand name, can create difficulties. Use realistic fictional cases, and label them as such. For images, also check the commercial terms of the tool used.

Read the tool's terms. Each provider sets its own rules on permitted uses and on generated outputs.

Think about reporting. Sales income, even small, may need to be declared, and the regime depends on your situation. Consult impots.gouv.fr or an accountant before taking this on seriously.

Your first listing, step by step

1. Choose a prompt you've been using for a while in your work.
2. Convert it to a version with variables using the first prompt in this guide.
3. Test it on three different cases, in a fresh conversation.
4. Have a trusted person try it and fix whatever stumped them.
5. Prepare an example of a real result, with no confidential data, and note the limits.
6. Set a price and write your listing.

So start with step 1: open your conversation history and find the prompt you've tweaked most often. If, after rereading it, you think it brings something your peers wouldn't have in five minutes, you have a candidate. If not, you've at least learned what it's missing.