The short version: yes, you can legally sell software, templates, chatbots or apps that you built using AI, and thousands of people are doing it right now on Gumroad, AppSumo and the GPT Store. The catch isn’t legality, it’s durability: most AI-built tools have no defensible edge because anyone with the same prompt can rebuild yours in an afternoon, and the platform you’re built on top of can quietly kill your business with a single terms-of-service update.
What “selling a tool built with AI” means in 2026
This isn’t one thing. It covers a lot of ground, and the answer changes slightly depending on which of these you mean:
- A custom GPT or Claude project you built and sell access to (or list on the GPT Store for a cut of usage revenue)
- A no-code AI wrapper: you’ve bolted a nice interface onto the OpenAI or Anthropic API and charge a monthly fee
- A Notion template, spreadsheet system or Zapier/Make automation that uses AI as one step in a workflow
- A full SaaS product where AI does the heavy lifting behind the scenes (transcription, image generation, copywriting)
- Prompt packs or prompt libraries sold as a digital product
I’ve built or tested versions of all five for my own business and for clients. The legal answer is the same across all of them: yes, you can sell it. The commercial answer is very different depending on which one you picked.
The same thinking applies city by city; for a Scottish worked example see AI consultant in Glasgow.
The real story: the automation I built and sold, and why it stopped working
In early 2025 I built a small AI-powered content repurposing tool for my own newsletter, then for two clients. It took a long-form blog post, ran it through GPT-4 with a custom prompt, and spat out five LinkedIn posts, three tweets and a short video script. I charged one client £150 a month for access to it as a “done for you” service, wrapped around a simple interface I’d built in Make.com with an OpenAI API call.
For the UK businesses the sector examples on the AI consultants for UK small businesses page show where the first month of savings usually comes from.
If you run a business in Sydney, the local page walks through what firms there tend to automate first: AI consultant in Sydney.
It worked well for four months. Then OpenAI updated pricing on the model I was using, my API cost per run tripled, and my £150 monthly fee went from a 70 percent margin to barely breaking even. Around the same time, LinkedIn’s own AI writing assistant started doing a rougher version of the same job for free, built directly into the compose box. I hadn’t lost my client to a competitor. I’d lost margin to a price change and relevance to a feature the platform gave away.
That’s the bit nobody tells you when they show you the screenshot of their £3,000 launch month: you’re renting your product’s core engine from a company that can change the rules whenever it wants, and you have zero say in it.
Yes, it’s legal, here’s the actual licensing detail
OpenAI’s usage policies and Anthropic’s commercial terms both allow you to build and sell products on top of their APIs, including charging customers for outputs. Same with Google’s Gemini API and Midjourney’s commercial licence (paid tiers only). What you can’t do, under any of their terms, is:
- Resell raw API access at a markup with no product wrapped around it (that’s against most providers’ terms and gets accounts suspended fast)
- Claim outputs are entirely human-made when a client specifically pays for “no AI” work
- Use a free-tier or personal ChatGPT Plus account to power a commercial product at scale (you need the actual API, which is billed separately and has its own commercial terms)
I know three small business owners who got caught out on the last one. They built something useful using ChatGPT Plus in the browser, started selling it as a service, then hit rate limits or account flags once volume grew, because personal subscription terms explicitly exclude reselling the output as a paid product at scale. Read the actual commercial API terms before you build your pricing model on top of a consumer subscription. It’s a fifteen minute read that saves you a rebuild later.
Where people are selling these things, and what they charge
Real numbers from real marketplaces, checked as of late 2025 and early 2026:
- GPT Store: custom GPTs with real usage can earn a share of OpenAI’s builder revenue programme, but most creators report earnings in the tens to low hundreds of dollars a month, not the thousands the early hype suggested
- Gumroad: AI-built Notion templates, prompt packs and automation kits typically sell for £9 to £49 as one-off purchases, with successful sellers moving 20 to 200 units a month
- AppSumo: lifetime deals for small AI SaaS tools commonly launch between $39 and $79, with founders reporting five-figure launch weeks but heavy refund rates once support demand kicks in
- Direct client work: consultants (myself included) charge £500 to £2,500 to build a bespoke AI-powered tool or automation for a single business, plus a monthly retainer for upkeep
The direct client route is, honestly, where the money is steadiest right now. Marketplace products compete on price and get copied within weeks. A tool built specifically for one business’s exact workflow, with your name attached to keeping it running, is much harder to walk away from and much harder to undercut.
The uncomfortable bit: most of these tools have no moat
Here’s the truth that launch-day screenshots never show you. If you built your tool by wrapping a prompt around GPT-4 or Claude, someone else can copy your exact output by pasting your marketing copy into ChatGPT and asking it to reverse-engineer the prompt. I’ve watched this happen to a colleague’s AI writing tool almost exactly: a competitor product launched six weeks later doing the identical job, at half the price, built by someone who’d simply tested the original tool and worked backwards.
Your moat, if you have one, isn’t the AI part. It’s the data you’ve collected, the specific workflow you understand better than anyone else, the relationships with the clients who trust you to keep it working, or the interface and support wrapped around the AI. The AI itself is a commodity that OpenAI, Google, Anthropic and Meta are all racing to make cheaper and more available. Betting your whole business on “I use AI to do X” without a second layer of defensibility is betting on something that gets easier to copy every single month, not harder.
What holds up over time
From what I’ve seen work for clients and for my own experiments, the tools that keep making money share these traits:
- They solve a narrow, specific, annoying problem for one type of business (not “AI marketing tool” but “AI tool that turns estate agent listings into Instagram carousels”)
- They own proprietary data or a workflow the AI alone can’t replicate, like a client’s historic sales data feeding the AI’s recommendations
- The founder keeps improving support and the interface faster than competitors can copy the output
- Pricing has room to absorb API cost increases without wiping out the margin, so build in at least a 60 percent gross margin buffer, not the 20 percent margin that looks fine until OpenAI changes its pricing tier
- They’re not entirely dependent on one AI provider; a tool built to swap between OpenAI, Anthropic or an open-source model if pricing or access changes survives shocks that kill single-provider products
A step-by-step for building and selling one
- Pick a problem you’ve personally solved for a client or your own business at least three times, not a general idea you think sounds useful
- Build a rough version using no-code tools (Make, Zapier, or a simple custom GPT) before you write a line of proper code
- Test it on five real people who aren’t friends or family and watch where they get confused, not just whether they like it
- Price it so a 40 percent API cost increase still leaves you profitable
- Read the actual commercial terms of whichever AI provider powers it, not the summary someone wrote on Reddit
- Decide whether you’re selling a product (marketplace, self-serve) or a service (bespoke, retainer), because the pricing, support load and margin expectations are completely different
- Launch small, to an audience that already trusts you, before spending money on ads to strangers
If step five feels like more than you want to take on alone, it’s the exact reason business owners bring in outside help rather than guessing; getting the licensing and margin structure right before launch is cheaper than fixing it after fifty customers have signed up, and it’s the kind of detail I walk clients through when they hire me as an AI implementation coach rather than trying to build it themselves from a YouTube tutorial.
Where this fits alongside other tools you already use
Most people selling AI-built tools aren’t starting from scratch. They’re bolting AI onto workflows they already run: repurposing content for Facebook ad campaigns, generating short clips using the same platforms covered in my rundown of where to make AI videos as a small business owner, or automating parts of the reporting they’d normally pull from Google Analytics tools. If you’re not clear yet on what counts as an “AI tool” versus just software with a chatbot bolted on, my plain-English explainer on what AI tools mean for non-technical business owners is worth ten minutes before you build anything to sell. And if the tool you’re building is really just a smarter version of something you already do daily, it’s worth checking my list of productivity tools tested over five years to see if you’re solving a problem that’s already been solved well enough that nobody will pay you for a new version of it.
The bit I’d tell a friend over coffee
Selling a tool built with AI is easier to start and harder to sustain than almost anything else in business right now. Starting is a weekend project. Sustaining it means watching API pricing like a hawk, accepting your “unique” tool will get copied, and building the parts around the AI, support, trust, specific data, a client relationship, that a competitor can’t paste into a chatbot and recreate by Friday. Sell the tool. Just don’t kid yourself that the AI part is the valuable part.
Frequently asked questions
Do I need permission from OpenAI or Anthropic to sell a tool built with their API?
No separate permission is required beyond agreeing to their standard commercial API terms, which explicitly allow selling products built on top of the API, provided you’re not simply reselling raw API access at a markup with no product wrapped around it.
Can I sell a tool built using a free ChatGPT or Claude account?
You can build a prototype that way, but consumer subscription terms generally exclude reselling outputs at commercial scale; once you have paying customers, move to the actual billed API, which has separate commercial terms designed for exactly this use.
How much can you realistically earn selling an AI-built tool?
Marketplace products (Gumroad, AppSumo, the GPT Store) typically earn tens to low hundreds of dollars a month for most sellers, with a small number of standout launches hitting five figures in week one; bespoke tools built for one client and billed as a service, at £500 to £2,500 plus a monthly retainer, tend to produce steadier income than marketplace sales.
What’s the biggest risk in selling tools built with AI?
Two risks matter more than legality: your margin can collapse if the underlying AI provider raises API pricing, and your product can be copied within weeks because the core output is often reproducible by anyone who tests it and reverse-engineers the prompt, so the AI itself is rarely the part worth protecting.
