TL;DR
On this page
There is no law that forces you to tell a client you used AI, and the data says disclosure is usually the lower-risk move, not the higher one. The honest answer to "should I tell them" is: disclose when your contract, your platform, or the content rules require it, stay quiet about tool-level AI the way you would about spell-check, and when you do disclose, frame it around the task and your accountability rather than the tool.
This walkthrough is part of the complete guide to freelancing in the AI era.
The reason this question feels heavy is that it is really two questions wearing one coat: a legal one (do I have to?) and a relationship one (will it cost me the client?). The data answers both, and neither answer is the scary one you are bracing for.
What the data actually says
Start with how common the silence is. In a survey of 157 writers by coach Ed Gandia, 49.7% had not told their clients they use AI. So if you have been quietly using it, you are in the majority, not the margins. Adoption itself is near-total: 93% of the writers surveyed use AI to some degree, and about 74% are regular or fully integrated users. Disclosure, in other words, means admitting to a standard tool, not confessing to a secret weapon.
Now the part that should change your calculus. Among the freelancers who did disclose, reactions skewed positive or neutral over negative by roughly four to one. The thing most people are afraid of, the client recoiling, is the rare outcome, not the common one. The fear is overpriced.
Why clients react better than you expect
Clients are not running a purity test on your tools. They are protecting against specific outcomes, and AI is only a proxy for those. Consultant J. Kelly Byram frames what they actually care about:
Most clients aren't anti-AI. They're anti-confusion, anti-risk, and anti-lower-quality output.
Source: J. Kelly Byram, Evergreen Biocomm
Read that closely, because it tells you exactly how to disclose without losing the room. A client hears "I use AI" and silently asks three things: will this be confusing to manage, will it create risk for me, and will the quality drop. If your disclosure answers those, that you direct the tool, you review everything, and you stand behind the result, you have removed the objection in the same breath you raised the topic. Framed that way, the disclosure describes process control rather than a shortcut, which is the distinction clients are actually weighing.
When you genuinely must disclose
Staying quiet by default is fine in many cases, but three triggers turn disclosure from optional into required. Miss one and a normal workflow choice becomes a breach.
| Trigger | What it requires |
|---|---|
| Your contract has an AI clause | Whatever the clause says, disclose to that standard |
| The platform requires it | Fiverr: disclose when a buyer asks. Upwork: policy leans to disclosure plus human-oversight expectations |
| The content falls under AI labelling rules | EU AI Act: AI-generated content may need labelling depending on context |
Your contract. If you or the client wrote an AI-use or disclosure clause into the agreement, that clause governs, full stop. This is the cleanest case because the answer is already in writing. The mechanics of what such a clause should contain are in AI clauses in freelance contracts.
The platform. If you work through a marketplace, its rules stack on top of your client relationship. Fiverr's seller guidelines require disclosure when a buyer asks, and Upwork's policies lean toward disclosing AI use and demonstrating human oversight. Check your platform's current policy rather than assuming, because these rules have been tightening.
The content rules. Under the EU AI Act, AI-generated content may need to be labelled as such depending on the context, with general-purpose AI transparency rules in effect since August 2025 and full enforcement arriving August 2026. If you produce content for EU clients or audiences, treat labelling as a compliance item, not a courtesy.
When disclosure is optional, and when it is just noise
Outside those triggers, the useful line is tool versus author. Nobody discloses spell-check, autocomplete, a thesaurus, or a search engine, and AI used at that level belongs in the same drawer: brainstorming, reorganizing an outline, tightening a sentence you already wrote, speeding up research you then verify. Mentioning it would be noise, the professional equivalent of announcing you used a calculator.
The line moves when AI shifts from tool to author. Generating the bulk of a deliverable from a prompt and lightly editing it is the kind of thing a client who hired you for original human work would reasonably expect to know. Translator and educator Corinne McKay puts the underlying principle plainly:
your freelance clients need to know if, how, and when you use (or don't use) artificial intelligence tools
Source: Corinne McKay, Training for Translators
The spirit of that is not "disclose every keystroke." It is "do not let the client form a false picture of how the work gets made." When you are unsure which side of the line you are on, a short proactive mention costs little, and the 4 to 1 reaction data suggests the expected downside is small.
How to actually say it
When you do disclose, the whole job fits in one sentence built from two parts: the task the AI does, and the accountability you keep.
I use AI to speed up research and first drafts, and I review and take full responsibility for every word I deliver.
That line works because it pre-answers Byram's three client worries. It is specific (no confusion), it names your oversight (no risk), and it asserts the standard you hold (no quality drop). Avoid the two failure modes on either side of it: over-explaining, which inflates a normal workflow into a confession, and vagueness, which leaves the client to imagine the worst version.
pro tip
Before you disclose anything, make sure you are not creating a different problem. Pasting a client's confidential brief into a consumer AI tool can breach confidentiality regardless of disclosure. The data-hygiene side is covered in is it safe to paste client info into ChatGPT.
Have the conversation once, in writing
If a client relationship is ongoing, the highest-leverage move is to stop having the disclosure conversation per project and settle it once. Put your disclosure language into the contract as an AI-use clause. That turns a recurring judgment call into a standing term both sides agreed to, and it protects you if the question ever resurfaces in a dispute.
Your AI-disclosure decision, made
FreelanceDesk builds contracts you can drop an AI-use clause straight into, so the disclosure standard is settled once in writing rather than renegotiated per project. It is free, and the document never leaves your browser. For the legal-footing question behind all this, see whether an AI-generated contract is even binding, and for the broader client-relationship side, setting freelance client boundaries covers the conversations that surround it.
References
- AI and Writers: a 157-Writer Survey : Ed Gandia / B2B Launcher
- What Clients Want From Freelancers Using AI : J. Kelly Byram, Evergreen Biocomm
- Does Your Freelance Business Need an AI Usage Policy? : Corinne McKay, Training for Translators
- The EU AI Act, Explained : FreelancerMap
- Using AI on Fiverr: Guidelines for Freelancers and Clients : Fiverr Help Center