How to Price Your Product with ChatGPT (Without Letting It Guess)

Ask ChatGPT "what should I charge for my product?" and you'll usually get a confident three-tier table: Starter, Pro, Business, at prices that look like every other SaaS page. It sounds reasonable, but the model doesn't know your buyer, your costs, or what your buyer pays today to solve the problem. So it guesses from the average of the internet.

This guide is for solo founders setting a first price, or rethinking one, for a SaaS, a digital product, or a productized service. It includes three copy-paste prompts: build a price-range brief from facts you actually have, turn a classic pricing survey into short customer conversations, and stress-test your pricing page before you ship it. There are no invented conversion stats or "optimal price" benchmarks. Where we lean on research, there's a link.

Why ChatGPT is bad at picking your price (and good at the work around it)

A language model has no access to your buyers' wallets. Any specific number it gives you without real inputs is a plausible-sounding average. That matters, because the first number people see tends to stick. Psychologists call this anchoring: Tversky and Kahneman showed in 1974 that estimates are pulled toward an initial value, even an arbitrary one (Wikipedia: Anchoring effect). If you anchor yourself on a number the model made up, every later decision drifts around it.

Where ChatGPT is useful:

• Structuring the inputs you already have (costs, alternatives, the outcome you deliver) into a clear range with a floor and a ceiling.
• Writing neutral research questions so you don't lead buyers toward the answer you want.
• Playing a skeptical buyer who reads your pricing page and points out what's confusing or unjustified.

So the job is simple: you bring the facts, the model brings the structure.

What to gather before you open ChatGPT

Twenty minutes of prep is enough:

• Your cost floor. What it costs you to deliver one more sale or one more month: hosting, AI API calls, support time, and payment processing. Check your processor's current fees on its own pricing page (for example, Stripe pricing) instead of typing a remembered number.
• The buyer's current alternative. What they do today: a spreadsheet, a freelancer, a competitor, or nothing. Write down what that costs them in money or hours, from conversations, not from assumptions.
• Two to five real competitor prices. Copy them from their pricing pages today, with the URL and the date.
• The outcome you deliver. One sentence, in your buyer's words if you have them.
• Your pricing model question. One-time, subscription, or usage-based. If you aren't sure, say so and let the model lay out the trade-offs.

Prompt 1: Price-range brief from your own facts

This prompt asks for a range and its reasoning, not a magic number. It also forbids invented benchmarks.

You are helping me set a first price for my product. Do not use industry averages or benchmarks I didn't give you.

Product: [WHAT IT IS, ONE SENTENCE]
Buyer: [WHO PAYS, e.g. solo founders who already sell online]
Outcome: [WHAT CHANGES FOR THE BUYER]
Pricing model I'm considering: [ONE-TIME / MONTHLY / USAGE / NOT SURE]
My cost per sale or per customer-month: [NUMBER + WHAT'S INCLUDED]
Buyer's current alternative and what it costs them: [FACTS FROM CONVERSATIONS]
Competitor prices (copied today, with URLs): [LIST]

Output:
1) Cost floor: the lowest price that doesn't lose money, with the math shown.
2) Alternative ceiling: the most a rational buyer would pay before sticking with their current alternative. Show the reasoning; mark every assumption as ASSUMPTION.
3) A suggested launch range (low / middle / high) and which end fits a product with zero reviews yet.
4) One-price vs. tiers: recommend one, and say what evidence would change your mind.
5) A list of the 3 facts that would most change this range if I learned them.

What good output looks like: visible math, assumptions labeled, and a short list of things to go and find out. If the answer contains a statistic you didn't supply, ask where it came from, and delete it if there's no source.

Prompt 2: Van Westendorp questions for five customer conversations

The Van Westendorp Price Sensitivity Meter, developed in 1976, asks buyers four questions: at what price a product would feel too cheap to trust, a bargain, getting expensive, and too expensive to consider (Wikipedia: Van Westendorp's Price Sensitivity Meter; SurveyMonkey's guide). The full method plots answers from a larger survey. With five to ten conversations you won't get a statistically reliable "optimal price," but you will learn the range buyers find credible, and the words they use about value.

I'm going to talk to [5-10] potential buyers of [PRODUCT] this week. Help me prepare.

Product description I'll show them (rewrite it to be neutral, no hype, no price): [YOUR DESCRIPTION]

Write:
1) A 3-sentence neutral product description I can read out loud.
2) Two warm-up questions about how they solve this problem today and what that costs them. No leading questions.
3) The four Van Westendorp questions (too cheap / bargain / getting expensive / too expensive), worded plainly for a 1:1 call.
4) One follow-up after the price questions: "What would have to be true for the 'getting expensive' price to feel worth it?"
5) A simple table template to record each person's four answers and one quote.

Rules: don't mention any price before the four questions. Don't ask "would you pay X?"

After the calls, paste your table back in and ask: "Summarize the range where most answers overlap, quote the strongest reasons people gave, and tell me which answers look like outliers. Don't claim statistical significance."

Prompt 3: Pricing-page stress test

A good price can still fail on a confusing page. Before you publish, have the model read it the way a cautious buyer would.

Act as a cautious solo founder deciding whether to buy, with 30 seconds and a tight budget.

Pricing section draft: [PASTE PRICE, WHAT'S INCLUDED, CTA, FAQ, REFUND/SUPPORT POLICY]

Do this in order:
1) In one sentence, what do I get and what does it cost in total? If you can't tell, say why.
2) List every line that raises a doubt: hidden costs, vague inclusions, unclear billing, unearned claims. Quote the phrase.
3) Which objection is most likely to stop me from buying, and does the page answer it?
4) If there are tiers: can I tell which one is for me within 10 seconds? If not, suggest clearer tier names based on who each tier is for.
5) Rewrite the "what you get" list and the CTA. Keep my facts and my price. Add no new claims, guarantees, or numbers.

Run this once per change. If the model's one-sentence summary doesn't match what you meant, the page is the problem, not the buyer.

Launch, then let real buyers move the number

Your first price is a hypothesis. A few habits keep it honest:

• Change one thing at a time. Don't change the price, the page, and the traffic source in the same week, or you won't know what moved.
• Watch behavior, not opinions. Clicks to checkout and completed purchases tell you more than "looks reasonable" in a DM.
• Test increases on new buyers first. Keep existing customers on the price they agreed to unless you've told them clearly in advance.
• Write down why you set the price. Next month you'll want to know whether the facts behind it changed.

Common traps

• Letting the model pick the number. It will. It shouldn't, unless every input came from you.
• Copying a competitor's tier table. Their tiers fit their buyers and their costs, not yours.
• Asking "would you pay $X?" People are polite, so you'll hear yes. Use the four neutral questions instead.
• Forgetting payment and tax costs. Your floor includes processing fees and any tax you have to collect. Check your own situation, and ask an accountant if you're unsure.
• Treating five calls as a survey. Use them for direction and language, not for a precise optimal price.

The short version

1. Gather your cost floor, the buyer's current alternative, and real competitor prices.
2. Use Prompt 1 for a range with visible math and labeled assumptions.
3. Use Prompt 2 to run the four Van Westendorp questions in five to ten conversations.
4. Use Prompt 3 to stress-test the pricing page before you publish.
5. Launch inside your range, change one thing at a time, and let purchases move the number.

ChatGPT can do the structuring in minutes. The price itself still has to come from your buyers.