Straight answer: AI is brilliant for pricing research, terrible for pricing decisions, and if you let ChatGPT or Claude set your final number you will almost always underprice yourself, because the data they're trained on is stuffed with freelancers and small businesses who are already underpricing. Use AI to gather the market picture, then add your own confidence multiplier on top, because no model on earth knows what your reputation, your waiting list, or your results are worth.
Why I started testing this
Last spring a client of mine, I'll call her Sarah because she'd rather I didn't use her real name, runs a virtual assistant business from her kitchen table in Kent. She'd asked ChatGPT to help her price a new "done for you inbox management" package. It came back with £450 a month for 20 hours of support. She used that number. Three months later she was burnt out, working evenings to keep up, and still not covering her own overheads once you counted software, insurance and the VAT threshold creeping up on her.
When we sat down together I asked her to run the same package past three competitors she rated, the ones with waiting lists. Their prices ranged from £850 to £1,200 for near identical scope. The AI hadn't lied to her. It had just averaged her against the entire internet, including the £8-an-hour VAs on Fiverr and Upwork who are competing on price because they have no other lever to pull. That's the part nobody tells you when they write "use ChatGPT to price your services" as a headline.
The uncomfortable bit about how these tools work
Large language models are trained on enormous scrapes of the public internet. A huge chunk of publicly available pricing information for services like copywriting, VA work, social media management, bookkeeping and web design sits on marketplaces built around undercutting, plus forums where people ask "am I charging too much?" and get told to lower their rates. There's very little public data from the agencies quietly charging £3,000 a month for the same work through word of mouth, because that pricing never gets typed into a forum post.
So when you ask an AI model "what should I charge for X", you're not getting the market rate. You're getting the median of everyone who was willing to publish a number, weighted heavily toward the cheapest end because that's where most of the internet chatter about pricing lives. It's the same reason Duolingo's freemium model works so well for them and terribly as a pricing template for a one-person consultancy: free-to-cheap only makes sense at massive scale with near-zero marginal cost. You are not Duolingo. Your hour is finite.
What AI is good for in pricing
I'm not saying skip it. I use AI for pricing research every single week in my own consultancy. It's good at:
- Summarising publicly listed rates for a service category across regions (ask it to separate UK, US and remote/global rates, they're wildly different)
- Drafting three or four pricing tiers based on scope you give it, so you're not starting from a blank page
- Reframing your pricing language, turning "I charge £80 an hour" into value language a client responds to
- Modelling scenarios fast, "what happens to my monthly revenue if I raise this package by 15% and lose two clients but keep eight"
- Building a quick interactive pricing calculator for your website so prospects can self-select into the right tier before they ever email you
That last one matters more than people realise. A calculator on your site does two jobs at once: it filters out tyre-kickers, and it primes the visitor with a number before you ever get on a call, which makes the eventual conversation about scope rather than sticker shock.
The six-step process I run with clients
1. Get the AI baseline first
Ask ChatGPT, Claude or Gemini something specific: "What's the typical monthly rate for a UK-based social media management service covering three platforms, weekly content, and monthly reporting, for a small business with under 10 employees?" Push it to give you a range with sources or reasoning, not a single number. Write the range down and ignore it for now.
2. Find five real competitors, not scraped ones
Go and find five businesses in your niche who you'd want to be compared to. Not the cheapest ones that show up on page one of Google. The ones with case studies, testimonials, and a waiting list mentioned somewhere. Check their pricing pages if they publish one, or ask a colleague or a mutual client what they pay. This takes maybe 45 minutes and is worth more than three hours of AI prompting.
3. Compare the two numbers
In Sarah's case the AI baseline was £450, the real competitor range was £850 to £1,200. That gap, roughly 90 to 165 percent, is not unusual. I see it in copywriting, in web design, in AI consulting itself. The AI number tends to sit near the bottom third of what real, respected providers charge.
4. Add your confidence multiplier
This is the step almost nobody talks about because it sounds unscientific, and it is a bit unscientific, but it's also the whole game. If you have testimonials, a waiting list, a specific result you can point to (grew a client's email list by 3,000 subscribers in four months, cut someone's admin time by 12 hours a week), you are allowed to price toward the top of the real competitor range, not the middle of the AI range. Confidence isn't arrogance here, it's pattern matching against your own evidence.
5. Build tiers, not a single price
Ask AI to help you draft three tiers based on the scope you've settled on: a lighter entry option, your core offer, and a premium version with faster turnaround or more strategic input. This is the same logic Calm uses with its subscription tiers, giving people a comfortable middle choice while the top tier makes the middle one look reasonable.
6. Test it on one real prospect before rolling it out
Quote your new number to the next genuine enquiry before you update your website. If they say yes without flinching, you've probably still priced too low. A small, quick intake of breath followed by a yes is the sweet spot. That flinch is useful data, not something to avoid.
A number worth sitting with
When Sarah raised her package from £450 to £795 a month (she split the difference between the AI number and the top of the real range, which is a sensible first move if you're nervous), she lost one client out of six. Her monthly revenue went from £2,700 to £3,975 across five clients, for less total hours because the remaining clients respected the higher price enough to stop asking for extras "just this once." That's a 47% revenue increase from a pricing conversation that took about two hours of actual work.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Where AI pricing advice quietly damages trust
There's a specific failure mode I see a lot with AI-generated pricing pages: they read as generic because the underlying logic is generic. A prospect who's shopped around three competitors in your space can smell a copy-pasted structure. Revolut's pricing pages work because every tier maps to a specific, named customer problem, not a vague "starter, growth, premium" template. If you're using AI to draft your pricing page copy, make sure the final version has your specific client language in it, the actual phrases your clients use when they describe their problem to you on discovery calls. That's the one thing no model can generate for you because it lives in your inbox, not the training data.
Negotiation is still a human skill, not a prompt
Once your number is set, the conversation where a prospect pushes back is where the real money is won or lost, and it's worth remembering that Dale Carnegie's core principle, that people buy on emotion and justify with logic, still applies whether the quote came from a spreadsheet or an AI tool. When someone says "can you do it for less," AI can help you draft a calm, professional response in seconds, but it can't read the tone of their voice or know that this particular prospect always pushes and always pays anyway. That judgment call stays yours.
A simple monthly habit
I now run a 20-minute pricing check once a month, same as a bookkeeping review. I ask an AI tool for a fresh rate range in my niche (rates shift, especially in anything touching AI consulting right now because demand keeps climbing), I glance at what three competitors are charging, and I check whether my own confidence level has moved, up or down, based on recent client results. It's the same discipline as checking analytics on Hotjar to see how visitors behave on your site rather than guessing. Pricing without a regular check-in drifts, usually downward, because it's easier to leave a number alone than to have the slightly awkward conversation about raising it.
If you're working through a bigger AI adoption project in your business and pricing is only one piece of a wider strategy conversation, that's usually the point where working with someone who does this daily, rather than piecing it together from prompts, pays for itself. An AI consultant for small business will have seen the same underpricing pattern across dozens of clients and can shortcut the two hours of confidence-building I described above into a single working session.
The short version to act on today
- Never let an AI tool's number be your final price, treat it as the floor, not the answer
- Spend 45 minutes finding five real competitors with waiting lists, not the cheapest five on Google
- Price toward the top of that real range if you have testimonials and results to back it up
- Build three tiers, not one flat price
- Test the new number on one real prospect before you publish it anywhere
- Recheck monthly, because rates in AI-adjacent work especially are moving fast in 2026
Frequently asked questions
Can AI tell me what to charge for my service?
It can give you a useful starting range, but that range tends to sit low because it's trained heavily on marketplace and forum data where people undercut each other. Use it as a floor, then adjust upward based on real competitors and your own results.
What's a reasonable amount to raise my prices by if I've been undercharging?
Somewhere between 30% and 75% is common for small businesses correcting a low starting price, applied gradually to new clients first rather than existing ones all at once, unless you renegotiate existing contracts separately with notice.
Should I tell clients I used AI to help price my services?
No need to lead with it, and there's little upside in mentioning it. Clients care about the outcome and the value, not the tool that helped you draft the number.
How often should I revisit my pricing?
Monthly is a good habit for a fast-moving service like AI consulting or social media management, quarterly is fine for more stable services like bookkeeping or design retainers. The key is having a fixed check-in date so it doesn't just get forgotten.
Related reading: Red Bull Marketing Strategy: How They Built a Brand That Wins and Apple Marketing Strategy: How They Built a Brand That Wins.
If you would rather not build this yourself, here is write for us about finance.
