Straight answer: ask ChatGPT or Claude what you should charge and it will nearly always land somewhere lower than what you're already worth, because it's pulling from Fiverr listings, Reddit freelancer threads and Upwork bids where everyone is racing to the bottom. Use AI to do the maths on your costs and targets instead, never to set the number itself, and expect the honest figure to be higher than the one it gives you.
The £900 quote that made me laugh out loud
Last spring I typed a fairly ordinary prompt into ChatGPT: "What should I charge for a half day AI marketing workshop for a 50 person UK company?" I was rebuilding my consulting business after five rough years and I wanted a sanity check on my own number, which at the time was £1,750 for the day.
It came back with £600 to £1,200, and it cited "typical freelance workshop rates" as its reasoning. I laughed, then I got annoyed, then I got curious about why a tool trained on most of the internet had landed so low.
The answer was simple once I looked at it: most of the pricing data floating around online comes from people who are underpricing themselves. Upwork profiles, Fiverr gigs, freelance subreddit threads full of people asking "is £300 too much to charge for a logo." That's the training data. AI isn't giving you market rate. It's giving you the average of a market that is, by its own admission, chronically underpaid.
I didn't lower my price. I raised it to £2,450, a 40% increase on what I'd been charging, and told three existing clients the new rate was coming in for anything booked after September. One pushed back and asked for a loyalty discount, which I gave once, as a one-off. The other two paid it without a single question. That told me everything about where my old number sat.
Why the chatbot keeps finding reasons you're too expensive
Here's the uncomfortable bit nobody selling AI courses wants to say out loud: if you ask a large language model "am I charging too much," it will almost always find a way to agree with the anxious tone of the question. These tools are built to be agreeable and to avoid conflict with the person typing. Frame the question as "help me justify a lower price" and it will oblige. Frame it as "help me justify a higher price and here's the value I create," and it will oblige that too, just as convincingly.
The model has no idea what you're worth. It only knows what you told it to look for. That's not a flaw you can patch, it's just how the tool works, and I think it's the single most missed point in every "use AI to price your services" article I've read.
I wrote about this same trap from a different angle when I covered what happened when I let AI churn out 569 blog posts in 45 days: the tool will happily produce volume and a plausible-sounding number, but plausible is not the same as correct, and you find out the gap the expensive way if you don't check its working.
What I changed in how I use it
I still use AI for pricing decisions, most weeks in fact. I just stopped asking it "what should I charge" and started asking it to do arithmetic on numbers I give it. That's a completely different job for the tool and it's much better at it.
- I tell it my target annual income (say £85,000), my realistic billable days per year after holidays, admin and no-shows (I use 130, not 220), and ask it to work backward to a day rate. That gets you to a floor, not a ceiling.
- I feed it my real overheads, not guessed ones: software subscriptions, a part time VA, travel, my accountant's fee, the tax I pay. Most people underestimate this by 20 to 30% because they forget the boring recurring bills.
- I ask it to price against the outcome, not the hours. A workshop that helps a company fix its lead generation and adds £40,000 in pipeline is not a "half day," it's a return on investment conversation, and I get the model to draft that framing for me rather than a day rate.
- I ask for three tiers, then I deliberately add 15 to 20% to whatever it suggests, every time, because I now know its default setting is cautious.
- I test the new number on one real prospect before I change it everywhere. If they say yes without flinching, I know I'm still under the ceiling.
None of that is complicated and you don't need to be technical to do it. I've written before about how I use these tools day to day without any coding background at all in this piece on being non-technical and using AI every day, and pricing is one of the plainest, most immediately profitable uses of the lot.
The websites AI is quietly copying its numbers from
If you want to see where the low numbers come from, go and look yourself. Upwork's public job listings, Fiverr's gig pages, and freelance forums on Reddit are full of people bidding against each other with no floor. I keep a small list of places I trust for market research, some of which most business owners have never heard of, in this roundup of 50 useful sites, and a couple of them are better sources for benchmarking your rate than anything a chatbot will summarise for you, because they're built on real transaction data rather than scraped listings.
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.
Put the maths on your website, not just in your head
One change that made the biggest difference to how prospects react to my pricing wasn't a chatbot conversation at all, it was building a simple interactive calculator on my site so people could work out their own rough investment before they ever spoke to me. It shifts the whole first call from "how much do you charge" to "here's what I already know it'll cost, tell me if that's realistic." I go into exactly how to build one and why it converts so well in this piece on interactive calculators, and it's still one of the highest-converting things on my whole site.
Where hiring proper help beats doing it yourself
There's a point where DIY prompting stops being enough, usually when you're trying to price a new service, like an AI-driven marketing package, and you have no history to work backward from at all. That's a different problem to fixing a rate you already know is too low, and it's worth being honest about the difference. If you're in that spot, I'd rather you spend a couple of hours with someone who's priced these things before than another week feeding prompts into a chatbot hoping it'll tell you something new. I've laid out realistic numbers for what that kind of help costs, and what you should expect for the money, on this page on AI consultant costs.
The clients who push back are not the ones you lose sleep over
When I raised my rate 40%, I braced for a wave of rejections. It didn't happen. One long-standing client negotiated, most didn't blink, and the one prospect who did walk away was, on reflection, the kind of client who'd have haggled over every invoice regardless of the number I'd started with. Raising a price doesn't lose you good clients nearly as often as fear tells you it will. It mostly just filters out the ones who were always going to be a headache.
My business is still being rebuilt in public, slowly, after five difficult years, and the pricing fix was one of the quieter wins in that process, the kind that doesn't make a dramatic story but shows up on the bank statement every single month. I wrote about the bigger version of that recovery, the traffic and revenue side of it, in the month my website started paying me again, and the pricing lesson sits right alongside it: small, unglamorous corrections compound faster than any single big idea does.
Frequently asked questions
Can I trust AI to tell me what to charge?
Not directly. Use it to calculate a rate based on your real costs, tax, admin time and income target, then check that number against your own gut and one real client conversation, rather than accepting its first suggestion as market rate.
Why does ChatGPT always suggest a lower price than I expect?
It's summarising freelance marketplace data from places like Upwork and Fiverr, where underpricing is common, and it also tends to mirror the anxious framing of whatever question you ask it. Ask it to justify a higher number with the same confidence and it usually can.
How much should I raise my prices if I think I'm underpricing?
Somewhere between 15% and 40% is a realistic first jump for most small business owners, applied to new bookings first rather than existing contracts, with a short heads-up period so current clients aren't blindsided by a sudden invoice change.
Is it worth building a pricing calculator instead of just listing a rate?
Yes, if your service varies by scope. A simple calculator sets expectations before the sales call, filters out people who were never going to afford you, and consistently converts better than a flat price list on its own.
Related reading: The AI Subscription Stack: What I Pay For (and What I Cancelled) in 2026 and AI Phone Receptionists for Small Business: What They Cost and Where They Quietly Lose You Money.
