AI Disclosure Decisions: The Complete Guide to When AI Content Requires Disclosure

On this page

Start here: the four questions that matter
AI images
AI video
AI text and chatbots
By AI tool
By industry
By scenario
Principles that apply across everything
Next: turn decisions into a process
FAQ

Quick Summary

The hard part of AI content is no longer making it — it is knowing when it has to be disclosed. Under Article 50 of the EU AI Act, the answer is rarely "always" or "never." It depends on what was created, how AI was used, whether the result could mislead someone, and which exemptions apply. This guide gives you the decision framework, then links to detailed guidance for every major content type, AI tool, and real-world scenario.

Start here: the four questions that matter

Almost every disclosure decision comes down to four questions. Work through them in order and you will usually reach an answer before you ever open the legal text.

1
Was AI involved? If no AI touched the content, Article 50 does not apply. If it did, keep going.
2
What type of content is it? Images, video, audio, text and chatbots each carry different risks and different disclosure expectations.
3
Is it synthetic, or merely edited? Generating an entirely artificial person is very different from sharpening an image or removing a stray telephone wire. Article 50 targets content that creates or substantially alters reality — not routine editing.
4
Could someone reasonably mistake it for authentic? This is the biggest question of all — and the one most of the guides below exist to help you answer.

A cartoon illustration rarely pretends to be real. A photorealistic AI-generated person often does. Replacing a cloudy sky usually does not recreate reality; generating an entire fake event probably does. That line — could a reasonable person be misled — is the heart of almost every disclosure decision.

AI images

Images raise more disclosure questions than any other content type — product photography, marketing graphics, banners, ads, social posts, stock imagery, illustrations, icons and logos each behave differently. The legal question is usually not "was AI used?" but "what exactly was generated?"

AI product photo disclosure AI infographics disclosure AI artwork & illustration disclosure AI enhancement & retouching disclosure

Coming soon: AI website images, AI marketing images, AI logos & icons, and a general AI-generated images primer.

AI video

Video adds complexity because several elements can be created or changed independently — an AI presenter, AI voice, AI subtitles, lip sync, background replacement or inserted objects. Some trigger disclosure; some do not.

AI-generated video disclosure rules AI-edited video disclosure rules AI translation disclosure

Coming soon: AI product demo videos, AI avatars, AI lip-sync, and AI podcasts.

AI text and chatbots

The EU AI Act generally treats AI-written text differently from synthetic media, but there are still situations where organisations must — or choose to — disclose AI-generated writing or make clear that a user is talking to a machine.

AI-written blog disclosure How to label AI chatbots

By AI tool

Many teams do not start with the law — they start with "I used Midjourney," "I edited this in Photoshop," or "Canva made this." So we have mapped the guidance to the tools you actually use.

Midjourney ChatGPT Photoshop Generative Fill Canva Magic Media Apple Image Playground Microsoft Designer

As new tools emerge, this section will keep growing.

By industry

The same AI-generated image can carry very different compliance risk depending on where it appears. Industry-specific guidance is on the way for ecommerce, retail, fashion, travel, hospitality, real estate, recruitment, healthcare, financial services and manufacturing — each with its own customer expectations and typical AI uses.

Industry decision guides are coming soon.

By scenario

Most people do not think in terms of "image" or "video" — they ask practical questions: Does this homepage need disclosure? What about our hero banner? What if only the background changed, or AI removed an object, or AI generated only the model? Our scenario guides mirror those real workflows.

When you do not have to disclose AI: exemptions by jurisdiction

More scenario walk-throughs are coming soon.

Principles that apply across everything

A few themes recur no matter the content type:

Not every use of AI needs disclosure. Everyday editing — noise reduction, colour correction, sharpening, cropping, resizing — raises very different questions from generating synthetic people or events.
Context matters. The same image might need a label on a product page but not in an obvious meme. Purpose is often as important as method.
Partial AI generation is the hard part. Modern images mix real photography with AI backgrounds, products, people and lighting. What matters is which elements are synthetic — not simply whether AI software was opened.
Transparency is becoming the norm. Even where disclosure is not strictly required, many organisations now label AI use voluntarily to build trust and stay ahead of regulation.

Next: turn decisions into a process

Once you have decided an asset needs disclosure, the next question is how to label it consistently. Run a free scan to find AI-generated images, video and text across your site, then apply clear disclosures and keep a record of each decision.

Further reading:

The complete Article 50 compliance checklist Who does Article 50 apply to?Understanding Article 50 in the context of the EU AI Act

FAQ

Does every AI-generated image need disclosure?

Not necessarily. Article 50 contains exemptions, and the answer depends on the type of content, how realistic it is, how it is used, and whether an exemption applies.

Do AI-edited photos need disclosure?

Sometimes. Simple edits generally raise different considerations from edits that substantially alter reality or create synthetic content. Our AI enhancement and AI-edited video guides walk through where the line falls.

Does using ChatGPT automatically require disclosure?

No. The answer depends on what was created, how it is presented, and whether Article 50 applies to that particular use. See our dedicated ChatGPT guide for common business cases.

What about Photoshop Generative Fill?

Generative Fill is one of the most common sources of uncertainty because it can make both minor edits and substantial synthetic changes. Our dedicated guide explains the difference and when disclosure applies.

What if multiple AI tools were used?

The legal assessment focuses on the resulting content rather than the number of AI tools involved. Work through the four questions above for the finished asset, not each tool in the chain.

See where your site stands

Run a free scan to find AI-generated images, video, and text on your site that may need an Article 50 disclosure.

Run a free scan

This article is general information, not legal advice.