Should Freelancers Tell Clients They Use AI? A Practical Disclosure Script for 2026
Should freelancers tell clients they use AI? Separate disclosure from confidentiality, check five sources of obligation, and use this proposal script.
Affiliate disclosure: Some links below are affiliate links. We may earn a commission if you sign up through them — at no extra cost to you. See our affiliate disclosure for details.
The Short Answer
Usually yes — briefly, in your proposal, as part of how you work rather than as a confession. But first check whether you are allowed to put the client's material into an AI tool at all: that confidentiality question matters more than disclosure, and it is answered by your contract, the client's policy, the platform you work through, sector rules and data-protection law. The script below covers both in four sentences.
Two Different Questions, Constantly Confused
"Should I tell clients I use AI?" is actually two questions wearing one coat, and they have different answers, different stakes and different people to ask.
Question one is about transparency. Does the client need to know how the work was produced in order to be getting what they think they are buying?
Question two is about confidentiality. Are you allowed to put their material into a third-party tool at all?
The second one is the serious one. Disclosure failures cost you trust. Confidentiality failures can cost you a contract, and in regulated work they can cost the client rather more than that. Most articles on this subject answer question one at length and never reach question two, which is the wrong way round.
Where the Obligation Actually Comes From
There is no single rule, and anyone who tells you there is has not read a contract recently. What binds you comes from a stack of overlapping sources, and you have to check yours rather than assume the general case:
- Your contract or statement of work. The first place to look, and the one people
- The client's own internal policy. Increasingly common, frequently unstated, and
- The platform's rules, if you work through one. Marketplaces set their own
- Sector rules. Legal, medical, financial, academic and public-sector work carry
- Data-protection law where the recipient's material originates, which may not
Read all five before you decide anything. The answer for a marketing blog post and the answer for a clinical document are not remotely the same, and the difference is not about your ethics — it is about which of those five actually applies to you.
The Script
The reason disclosure goes badly is almost always framing. Announced as a confession, it invites a discount conversation. Stated as method, it is unremarkable — nobody apologises for using a spell checker, a camera or a CAD package.
What works is short, specific, and placed in the proposal rather than raised later. Something close to this, in your own voice:
"Here is how I work. I use AI tooling for research passes, structural drafts and routine transformation."
"Everything that reaches you is reviewed, edited and fact-checked by me, and I am accountable for it."
"I do not put your confidential material into third-party tools without asking you first."
"If you have a policy on this, send it and I will work to it."
Four things are doing the work there. It says where the tooling is used rather than whether. It keeps accountability with you, which is what they are actually buying. It raises confidentiality before they have to. And it invites their policy, which converts a potential dispute into a requirement you can meet.
What to Say for Which Kind of Work
| Work type | Sensible default | Why |
|---|---|---|
| Research and summarising | Mention in your method, no per-job flag | Reversible, low stakes, output is checked |
| Structural first drafts | Mention in your method | The client is buying the finished thing |
| Anything published under the client's name | Confirm their policy explicitly | Their reputation, their call |
| Work with a stated originality warranty | Read the warranty first | You may be promising something specific |
| Regulated, clinical or legal content | Ask before, in writing | Sector rules override preference |
| Anything under an NDA | Ask before, in writing | This is the confidentiality question, not disclosure |
The pattern down that table: the more consequence sits with the client rather than with you, the earlier and more explicitly you ask.
What You May Put Into a Tool
This is the half that gets skipped, and it is answered by facts about the tool rather than by your judgement. Before pasting anything belonging to a client, know:
- Whether inputs are used for training, and whether that differs between the free
- Where the data is processed and stored, and for how long.
- Whether the vendor publishes a data-processing agreement you could actually
- Whether your specific contract permits third-party subprocessors at all, and
- Whether you can turn retention off, and whether you have.
A practical habit that removes most of the risk: keep a clean line between work you do on client material and work you do on your own artefacts — your notes, your outline, your structure. A great deal of useful tooling can be applied to the second without the first ever leaving your machine.
The Counter-Argument: You Can Over-Disclose
Disclosure is not a virtue in itself, and there is a real failure mode on this side too. A proposal that spends three paragraphs on your tooling has made the tooling the product. Clients hire for outcomes; a supplier who narrates their methods at length reads as either insecure or as someone about to explain why something is not their fault. It also invites a negotiation you did not need to have, on the least favourable possible ground.
There is a subtler version. Disclosing at the wrong moment — after delivery, or when challenged — converts a neutral fact into an admission. The same sentence lands completely differently depending on whether you volunteered it in week one or produced it in week six.
The line worth holding: be completely transparent about anything that affects what the client receives, their confidentiality, or their risk. Be brief about everything else. Method belongs in the proposal, once, in plain language.
The Limits
Nothing here is legal advice, and it is not a substitute for reading your own contract. Rules differ by jurisdiction, by sector and by platform, and they are moving. The one durable principle is that the client should never learn something about how their work was made that they would have wanted to decide on — and if you are unsure whether a given fact clears that bar, it does.
→ AI tools for freelancers | Browse all agents
Related: How to Price Your AI Freelance Services | How to Find Your First 3 Freelance Clients | AI Agents and the Reality of Job Replacement
Frequently Asked Questions
Do freelancers have to disclose AI use to clients?
Sometimes. The obligation can come from your contract, the client's policy, a marketplace's rules, sector regulation or data-protection law. Check those first; where none applies, brief disclosure in your proposal is still the sensible default.
How should I tell a client I use AI?
State it as method in the proposal: where you use AI tools, that you review and are accountable for everything delivered, that you do not put confidential material into third-party tools without asking, and that you will follow their policy.
Can I put client documents into ChatGPT or other AI tools?
Only if your contract and the client allow third-party processing. Know whether the tool trains on inputs, where data is stored and for how long, and whether retention can be turned off.
Can you over-disclose AI use?
Yes. Paragraphs about your tooling make the tooling the product and invite a discount conversation. Be fully transparent about anything that affects what the client receives or their risk, and brief about the rest.
