DisclosureGuardRegulatory dossier

Blog · ← All posts

Freelancer AI-Disclosure Checklist: Keep Clients Compliant

Last updated: October 13, 2026 · 8 min read

By the Ingenious AI team

Most disclosure advice is written for companies with legal departments. This one is for the solo video editor, designer, or copywriter who uses AI tools every day and doesn't have a compliance team, just a reputation to protect.

Here's the uncomfortable truth: you probably use AI more than your clients think. Say you're a designer who delivers a product mockup series. The client's CMO spots the same generated background across three "different" photoshoots. The project ends with an awkward email chain and no repeat business. Nothing illegal happened. It was a trust problem, and trust problems are the ones that cost freelancers their best clients.

This checklist exists so that never happens to you.

Part 0: Know what counts as "AI use"

Before you disclose anything, you need an honest inventory of what you're actually doing, because the gray area is wider than most freelancers admit.

Obviously AI: generating the hero image in Midjourney, writing the draft in Claude, producing the voiceover with ElevenLabs. Clearly human with AI assistance: using AI to upscale footage, clean up audio, remove a background, auto-caption. The trap is the middle. Generative Fill to remove a speck of dust? Meta's detectors have flagged exactly that as AI-modified content, minor AI edits can trigger the "AI info" label on real photos, as photographers discovered when Meta rolled out its labeling in 2024.

The rule we recommend: if an AI tool touched pixels, audio waveforms, or sentences in the deliverable, it's in the disclosure inventory. Not every touch needs a visible label on every platform, YouTube explicitly exempts things like AI-assisted outlines, thumbnails, captions, and cloning your own voice. But the client conversation and your own log should reflect the full picture. You can always explain nuance. You can't explain why you hid something.

Part 1: Disclose to the client before the work starts

The disclosure conversation belongs in the onboarding, not in the delivery email. Clients don't mind AI-assisted work nearly as much as they mind discovering it.

What if the client says "no AI at all"? That happens, and it's a legitimate brief, some brands want a provable human-only chain, especially for testimonial-style content. Treat it as a scope decision, not an insult. Quote it higher, because human-only production genuinely costs more time, and keep your production records clean enough to prove it. A client willing to pay for verified human work is a good client.

Part 2: Label the deliverables

What's due on the files themselves depends on where they end up, and that's often out of your hands. So plan for the strictest destination the work could reach. Here's the per-platform reality as of late 2026:

Two things to add on the files themselves:

Don't rely on the platform checkbox alone. TikTok's AI-content toggle is a disclosure to TikTok. It doesn't put a label in front of the viewer, and it doesn't satisfy the EU AI Act's requirement that the disclosure reaches the person seeing the content. The checkbox and the visible label are two different jobs.

Part 3: Keep a disclosure log

For every project, keep one row per deliverable: file name, AI tools used, what was generated vs. edited vs. untouched, disclosure applied (visible + embedded), and destination platform. This takes two minutes per project.

Why bother? Three reasons, in increasing order of seriousness. First: six months later, when a client asks "wait, was this one AI?", you have an answer instead of a guess. Second: when a platform's detector mislabels your honest work, which happened to real photographers under Meta's labeling rollout, that log is how you contest it. Third: if a regulator or platform ever questions a piece of work, the log is your evidence that disclosure was intentional and documented. We built DisclosureGuard's audit trail around exactly this instinct, a tamper-evident record of what was declared, when.

A spreadsheet works. So does a text file. The format matters less than the habit. What matters is that the log exists before anyone asks for it.

Part 4: The conversation scripts

The hardest part isn't the labels. It's telling the client. Two scripts that actually work:

One more script, for the awkward middle of a project: when the plan changes. If you quoted human-shot photography and end up generating the backgrounds because the shoot fell through, tell the client before delivery, not in the delivery note. Mid-project surprises read as confessions. Advance notice reads as professionalism. Same facts, different framing, completely different outcome.

Part 5: Charge for the disclosure work

Here's the part freelancers skip: disclosure is billable work. Exporting with provenance intact, writing platform-specific disclosure copy, keeping the log, checking each destination's requirements, that's 30 to 60 minutes per project that used to be zero. Don't absorb it silently.

Add a line item: "AI disclosure & provenance handling, €75." Or bake it into the project fee and name it in the quote. Clients who've been burned by undisclosed AI work will pay for it gladly; the rest won't object to a line item they understand. The freelancers who treat compliance as a cost center compete on price. The ones who name it compete on trust, and trust is the thing clients actually buy.

One more number to keep in your back pocket: the EU's Art. 50 fines run to €15M or 3% of global turnover. You're not scaring anyone with that, your client's exposure is what it is, but when a client asks why the disclosure line item exists, it answers the question. Frame it as insurance language: "this is the documentation that shows we did it right."

Part 6: Re-check at delivery

Disclosure decays between production and publish. The file you exported with a C2PA manifest gets recompressed by the client's CMS and the manifest is gone. The caption you wrote with the AI label gets rewritten by the client's social manager. The YouTube toggle gets skipped because the client's VA uploads the video and nobody told her.

So the final checklist item is a delivery-day verification: open the actual published post or ad, check the visible label is there, and confirm the platform control was set. Two minutes. This is the step that catches the embarrassing misses, the ones where you did everything right and the pipeline undid it. If you want to see how fast embedded disclosure dies in a real upload pipeline, run a file through our free drift check before you hand it to the client.

What this won't do

Honest caveat: no checklist covers every jurisdiction. China requires both visible and machine-readable labels; the EU's Art. 50 applies to content your audience sees; US state laws are a patchwork, New York, for instance, requires conspicuous disclosure inside ads containing synthetic performers, with fines starting at $1,000 for a first violation. If you work with clients in regulated industries (pharma, finance, politics), this checklist is the starting point, not the whole answer, get proper advice for those.

But for the working freelancer, this checklist closes the gap that actually costs money: the trust gap. Enforcement rarely starts with a regulator knocking. It starts with a client who feels misled. Don't give them that feeling.

Scan your first piece of content free →

Related reading