The Pixels Came With a Question Mark: Handling Client-Supplied AI Images

A client sends over a brief, a deadline and a folder of images. Somewhere between the moodboard and the logo files, you notice the tell-tale sheen: a six-fingered hand, a suspiciously smooth sky, or a file name like ChatGPT_Image_10Nov2026.png. The images are AI-generated. Now what?

The temptation is to either ignore it and get on with the layout, or panic and start drafting a legal memo. Neither is especially useful. The better move is to treat client-supplied AI images like any other third-party asset: something that needs a source, a permission check and a paper trail before it becomes part of your studio's output.

Put the brakes on gently

Don't make it a moral panic. A lot of clients are using AI tools to explore ideas, and some of those images will be perfectly fine to use. But 'fine to use' is a conclusion, not an assumption. Before you drop anything into a layout, ask a short, friendly set of provenance questions:

• Which tool generated the image?
• Whose account was used — a personal one, a work one, a free tier or a paid plan?
• When was it generated, and was it edited afterwards?
• Can they share the prompt or a screenshot of the generation?
• Are there any real people, brands or logos in the image?

You don't need to interrogate anyone. A line like 'Before I place these, can you confirm where they came from and whether you've got commercial rights to use them?' is usually enough. If the client can't answer, that's useful information too.

'We own it' is a starting point, not a receipt

Clients often say 'it's ours' because they paid for a subscription or because the image appeared in their account. That may be true, but it isn't the whole picture. AI tools have different terms for free and paid users, different rules about commercial use, and different positions on whether outputs can be protected or licensed. Some tools also have restrictions around people, trademarks or sensitive content.

You don't need to become a lawyer. You need a record. Ask for written confirmation that the client has the right to use the images in the way the project requires — including in paid advertising, on a website, in print, or on social media. If the images came from another agency or a previous supplier, ask for the chain of permission. If the client generated the images using a photo of a real person as an input, stop and check that too.

This isn't legal advice, and it isn't about being difficult. It's about making sure that when the work goes live, the question of who owns what has an answer that isn't 'I think it was fine'.

Disclosure is a design decision, not an afterthought

If the image is going in front of an audience, ask whether anyone could be misled by it. UK advertising rules expect marketing communications to be transparent, and a photorealistic AI image can blur the line between illustration and documentary evidence. A generated image of a product in use, a fake customer testimonial, a before-and-after shot or a scene that looks like a real event all deserve a second look.

In many cases, a simple caption or credit is enough: 'AI-generated image' in the alt text, a note in the image credits, or a line in the page footer. If the client doesn't want to disclose it, put your recommendation in writing and ask them to confirm the decision. You don't have to win the argument; you do have to keep the paper trail.

Metadata: don't trust it, but don't ignore it

Embedded metadata is patchy. Some platforms strip it on export, some add it, and some change it when an image is resized or converted. So don't rely on EXIF or a watermark to tell the whole story. Keep a sidecar record instead: a plain text file or a note in your project management tool that says where the image came from, when it arrived, what the client told you, and which version you used.

File naming matters too. A folder full of image1.png, image2.png and final_final.jpg is a provenance nightmare. Try something like:

clientname_2026-11-10_ai-source_tool_v01.png

It's not glamorous, but it means that six months later, when someone asks where a picture came from, you can answer without opening seventeen folders. Keep the original file, the working file and the exported version separate. If the client shared prompts, save them alongside the assets — with a note about confidentiality if they contain sensitive information.

Make the scope do the heavy lifting

The easiest way to stop AI images becoming your problem is to decide, before the project starts, how they'll be handled. Add a short clause to your terms or project proposal:

• The client is responsible for confirming rights and permissions for all client-supplied assets, including AI-generated images.
• You'll carry out a basic provenance check on up to a set number of images as part of the project.
• Additional clean-up, retouching, re-generation or licensing checks are billed separately or handled as a change request.

That last point matters. If a brief arrives with thirty AI images and an expectation that you'll 'just tidy them up', that isn't a moodboard — it's asset production. Price it accordingly, or push back before the deadline gets tight.

Red flags that warrant a pause

Some images need more than a note in the file. Pause if you see a real person's likeness, a celebrity, a brand logo, a copyrighted character, a sensitive news event, a medical or financial claim, or a style that clearly mimics a living artist. Also pause if the client can't say where the image came from at all.

You don't have to police every pixel. But you do have to avoid quietly turning a questionable asset into a polished final file. Ask for a replacement, suggest a licensed stock image, or get written instruction if the client wants to proceed. Then keep that instruction with the project files.

A short intake checklist

Before any client-supplied AI image goes into a layout, check:

• Source tool and account type
• Written confirmation of commercial use
• Any real people, brands or sensitive content
• Disclosure requirements for the end audience
• File name, folder and sidecar note
• Scope and billing for clean-up work

The bottom line

AI images aren't the enemy. Unclear ownership is. Treat them like any other third-party asset — with a source, a permission check and a record — and you'll spend less time worrying about the picture and more time doing the work you're actually paid for. The prettiest image in the deck shouldn't be the thing that keeps you up at night.