Will Clients Know You Used AI? I Found Out

A client once emailed me a single line about a draft I’d sent: “Did you use AI for this?” My stomach dropped the way it does when you’ve been caught, even though I wasn’t sure I’d done anything wrong.

I had used AI. Not to write the thing for me, but to get past a blank page on a section I was stuck on. I’d edited it after, or thought I had. Clearly not enough.

What followed was a few hours of the shame spiral every freelancer who uses these tools knows. Was I a fraud? Was I supposed to disclose it? Had I just torched the relationship? I drafted three different defensive replies and deleted all of them. Then I actually read the draft again, slowly, trying to see what they saw.

And there it was. The section I’d leaned on AI for didn’t sound like the rest. It was smoother and emptier. It used their industry’s words without their company’s specifics. It could have been about any business in their sector, which meant it wasn’t really about theirs. The rest of the piece had their actual situation in it, the details only I knew because I’d talked to them. That one section had none of that. It floated.

So I wrote back honestly. Yes, I used a tool to draft one section, the same way I might use a calculator or a thesaurus, but I clearly didn’t finish the job of making it yours, and that’s on me. Give me a day to rewrite it properly.

Here’s the part that reframed the whole thing for me. They didn’t care that I’d used AI. They cared that a section of their work sounded generic. When I sent back the rewrite, grounded in their actual details, the question never came up again. The problem was never the tool. The problem was that I’d let the tool’s output leave the building before I’d made it specific.

That’s the real answer to whether clients can tell you used AI. Mostly, they can’t tell by some forensic signature. What they can feel is the absence of you. AI writing has a default texture: competent, fluent, and weightless. It says true things that could apply to anyone. Your clients hired a specific human who knows their specific situation, and when a passage suddenly stops sounding like that human and starts sounding like the internet’s average, something registers, even if they can’t name it. The tell isn’t “robotic.” The tell is generic.

I didn’t quit using AI after that. That would have been the wrong lesson. What changed is the line I now hold, the same one I keep coming back to in the client conversation AI can’t have for you: the tool can help me draft, but the specifics, the judgment, and the voice have to be mine before anything reaches a client. AI is allowed to give me a worse first version. It is not allowed to give them my final one. I went deeper on what these tools genuinely do and don’t do well in AI writing tools: what they’re actually good at.

Do I disclose it now? Honestly, I’m still working that out. I don’t announce that I used spellcheck either. But I’ve landed somewhere like this: I don’t volunteer it, I never lie about it if asked, and I make sure the work is specific enough that the question rarely comes up. That last part does more than any disclosure policy.

The freelancers who’ll get hurt by AI aren’t the ones who use it. They’re the ones who let it speak for them and forget to put themselves back in. I learned that from one short email and a very long afternoon, and I’m not fully past the flinch yet. The deeper reassurance, the one that actually lets me sleep, is in AI won’t replace you, but someone using AI might.