AI search & GEO
EU AI Act: What It Means for Your Marketing and How to Comply
Only what matters if you do marketing, read in the official text: when your chatbot must say it's AI, what needs a label and what doesn't, what's banned, which dates count, whether it reaches US companies, and where AI chatbots got it wrong.
The EU AI Act has required, since August 2, 2026, that your chatbot tell people it's an AI and that any deepfake you publish carries a clear notice. Marketing copy written with AI and reviewed by a person needs no label, and using ChatGPT in your business isn't banned.
In the US, "eu ai act" gets about 8,100 Google searches a month, according to DataForSEO, and most results explain the entire law: risk tiers, foundation models, conformity assessments. This guide covers only what touches marketing, read in the official text and in the European Commission's own FAQ.
This isn't legal advice: it explains what the rules say as of October 2026. If your case is sensitive, talk to a lawyer. For the rest of our AI guides, head to AI search.
On this page
What the EU AI Act is (and what it isn't)
The EU AI Act is Regulation (EU) 2024/1689, which regulates artificial intelligence according to the risk it poses to people.
The EU AI Act, also called the European AI Act or the EU Artificial Intelligence Act, is the law that sets the rules for building, selling and using AI systems in the European Union. It entered into force on August 1, 2024 and applies in phases. Here's the EU AI Act summary in one line: the riskier the use, the stricter the rules.
What it isn't: a ban on ChatGPT, Gemini or Claude. It isn't a rule to stamp "made with AI" on everything either. Most marketing falls into the minimal or transparency tiers, whatever types of AI you use. These are the EU AI Act risk categories:
| Risk tier | What it covers | What it means for your marketing |
|---|---|---|
| Prohibited | Harmful manipulation, exploiting vulnerabilities, social scoring | You can't do it |
| High risk | Hiring, credit, education, biometrics | Rarely touches marketing |
| Transparency | Chatbots, deepfakes, synthetic content | Disclose and label |
| Minimal | Spam filters, recommendations, draft copy | No new obligations |
EU AI Act high risk rules are the heavy part of the law, and the AI Omnibus pushed them back to December 2027 and August 2028.
Does the EU AI Act apply to US companies?
Yes, the EU AI Act applies to US companies when their AI systems are offered in the EU or their AI output is used there.
Article 2 covers providers that put AI systems on the EU market wherever they're based, and providers and deployers outside the EU when the output of the system is used in the EU. So an Austin coffee roaster that ships to Germany and runs an AI chatbot that German customers chat with is in scope for that chatbot. A US shop with no EU customers isn't.
Provider or deployer: which one are you?
You're usually a deployer: a business using an AI system under its own authority. The provider is whoever develops the system and puts it on the market under its name, like the company behind your chatbot tool. If you build your own bot on top of a model's API and launch it under your brand, you become the provider of that bot. The role decides who owes which part of EU AI Act compliance.
AI regulation in the US: which rules reach your marketing
The US has no single federal AI law for marketing, so the rules come from existing consumer law plus a patchwork of state laws.
| Rule | What it asks | Who it hits |
|---|---|---|
| FTC fake reviews rule | No fake reviews or testimonials, AI-generated ones included | Anyone who writes, buys or spreads them |
| California bot law | A bot that sells to Californians can't pose as a person; a clear disclosure is the safe harbor | Any business with a sales chatbot |
| California AI Transparency Act | Detection tools and embedded disclosures, operative from August 2, 2026 | Generative AI providers with over 1 million monthly users |
The FTC also treats a deceptive AI claim like any other deceptive claim, so "100% AI-powered" needs proof, just as "clinically proven" does.
Notice the pattern. The chatbot disclosure you'd add for the EU already covers California's bot rule. The AI Transparency Act lands on the big tool makers, not on you. State laws and bills in this area move fast, so check the current text of a state law before you rely on it, and ask a lawyer if you sell in several states.
EU AI Act Article 50: what changes in your marketing
Under EU AI Act Article 50, what changes in marketing is transparency: chatbot disclosure and deepfake labels from August 2, 2026.
This is what Article 50 says, checked against the European Commission's FAQ on transparency obligations, updated July 24, 2026. These are the EU AI Act requirements that actually reach a marketing team:
| What you do | What the law requires | Who's responsible | Since |
|---|---|---|---|
| Chatbot on your site or in DMs | Tell people it's AI, unless that's obvious | The provider (you, if you built it on an API) | Aug 2, 2026 |
| Realistic image, video or voice in an ad (deepfake) | Clear, visible notice that it's artificial | You, as the deployer | Aug 2, 2026 |
| Generating images or video with a tool | Machine-readable marking in the file | The tool's company | Aug 2, 2026 (Dec 2, 2026 if already on the market) |
| Marketing copy reviewed by a person | Nothing | No one | Doesn't apply |
| Your team using AI | Support their AI literacy | You | Feb 2, 2025 |
Do you have to label all AI-generated content?
No. The mandatory label applies to deepfakes and to AI-generated text published to inform the public on matters of public interest without human review. A post, email or product page that you review and sign off on doesn't need one. Meta and TikTok, however, have their own labeling rules.
AI disclaimer examples: wording for chatbots, images and ads
An AI disclaimer works when it's short, sits where people look and appears the first time they meet the content. The Commission's FAQ says a deepfake notice must come at first exposure at the latest, visible or audible without any special tool.
| Where | Example wording | Where it goes |
|---|---|---|
| Chatbot | "I'm an AI assistant, not a person. Ask for a human any time." | First message |
| Realistic AI image in an ad | "Image generated with AI" | On the image |
| AI voice in a video or audio ad | "Voice generated with AI" | Spoken or on screen at the start |
| Video with an AI-generated person | "This person doesn't exist: generated with AI" | On screen from the first second |
| Text you reviewed and signed | Nothing required | No label |
Two traps. A tool's invisible file mark doesn't replace your visible notice, and a notice buried behind a link isn't visible on first exposure. Artistic, satirical or fictional works get a lighter notice that doesn't spoil the piece. An ordinary ad is a different case, so ask a lawyer before you lean on that exception.
EU AI Act prohibited practices: the manipulation you can't use
The EU AI Act prohibited practices include using AI to manipulate people or exploit their vulnerabilities when it causes significant harm.
Since February 2, 2025, Article 5 bans AI systems that use subliminal, manipulative or deceptive techniques to materially distort what someone decides. It also bans systems that exploit age, disability or a specific social or economic situation. In both cases there has to be significant harm, or a reasonable likelihood of it.
In plain terms: personalizing offers or recommending products is fine. A bot that spots older people in financial trouble and pressures them with fake urgency into taking a loan is not. And anything the AI Act doesn't ban can still be false advertising, which the FTC already polices in AI claims.
From December 2, 2026, another ban kicks in: AI systems that generate child sexual abuse material or intimate images of people without their consent.
EU AI Act timeline: what applies now and what's next
The EU AI Act timeline runs in phases since 2024, and the part that matters for marketing, transparency, applies from August 2, 2026.
These dates come from the European Commission, including the changes made by the AI Omnibus, checked in October 2026.
| Date | What applies |
|---|---|
| Aug 1, 2024 | Entry into force |
| Feb 2, 2025 | Prohibited practices and AI literacy |
| Aug 2, 2025 | General-purpose AI models (the ones behind ChatGPT, Gemini or Claude) and governance |
| Aug 2, 2026 | Transparency (Article 50) and general application |
| Dec 2, 2026 | Marking for tools already on the market, plus the new sexual content ban |
| Dec 2, 2027 | High risk in areas like employment or education |
| Aug 2, 2028 | High risk in regulated products, like toys or machinery |
What did the AI Omnibus change?
The AI Omnibus, Regulation (EU) 2026/1744, in force since July 27, 2026, delayed the high-risk deadlines, added the December 2026 ban and gave existing generative tools until December 2 to add marking. It also softened AI literacy: you now have to support your team's AI knowledge, not guarantee a set level. Article 50 for chatbots and deepfakes stayed on August 2, 2026.
How to comply with the EU AI Act in your marketing, step by step
To comply with the EU AI Act in marketing, map where you use AI, add the disclosures that apply and keep a record of what you did.
-
Map where you use AI
Chatbot, copy, images, video, voices, ad targeting. Note which ones reach EU customers.
-
Tag each use with its obligation
Use the table above: disclose, label or nothing. Most of yours will be "nothing" or "disclose."
-
Put the disclosure in the chatbot's first message
It should say it's an AI assistant in the greeting. If you're unsure what kind of bot you have, start with what is a chatbot and how it's set up.
-
Label deepfakes inside the ad itself
On the image or the video, not hidden behind a link. A cloned voice or a person who doesn't exist needs a clear notice.
-
Review what you publish by hand
Human review keeps your copy out of the labeling rule and protects you from invented facts. Before you hit publish, learn how to humanize AI text so it also sounds like you.
-
Train your team and write it down
A short session on which tools you use, what data never gets pasted and what gets reviewed. Keep the date and who attended.
Here's the list of AI tools we use in marketing: [list, with what each one is used for]. For each one, tell me whether the EU AI Act requires us to disclose, label or nothing, and why, citing the article of Regulation (EU) 2024/1689. If you're not sure about something, say so instead of guessing. Flag which cases I should check with a lawyer.
Review this campaign before it goes live: [copy, plus a description of images and videos]. Flag which pieces could count as a deepfake under the EU AI Act definition, which claims I can't prove and which lines could pressure vulnerable people. Don't rewrite the campaign: give me a checklist of what to review.
Frequently asked questions about the EU AI Act
Is there an EU AI Act?
Yes. The EU AI Act is Regulation (EU) 2024/1689, published in the Official Journal of the EU on July 12, 2024 and in force since August 1, 2024. It's the first comprehensive AI law adopted by a major regulator, and it applies directly in all 27 member states. It was amended in 2026 by Regulation (EU) 2026/1744, the so-called AI Omnibus, which pushed back the high-risk deadlines but kept the transparency rules on schedule.
Where can I download the EU AI Act PDF?
On EUR-Lex, the EU's official law portal. Search for Regulation (EU) 2024/1689 and you'll get the text as published in the Official Journal, in English and every other official EU language, with a PDF download. Because it was amended in July 2026, check the consolidated version for current dates. Unofficial sites that break the law into articles are handy for reading, but EUR-Lex is the version that counts.
What are the 5 principles of AI ethics?
There's no single official list, but the most cited one comes from the OECD AI Principles, adopted in 2019 and updated in 2024. Its five values-based principles are: inclusive growth and well-being; respect for human rights, democratic values, fairness and privacy; transparency and explainability; robustness, security and safety; and accountability. The EU AI Act turns several of those ideas, like transparency and human oversight, into legal obligations.
What are the EU AI Act penalties?
Fines go up to 35 million euros or 7% of worldwide annual turnover for prohibited practices, up to 15 million or 3% for breaking other obligations, including transparency, and up to 7.5 million or 1% for giving authorities false or misleading information. For each tier, large companies face whichever amount is higher. For small and medium-sized businesses and startups, it's whichever is lower, which makes the realistic exposure much smaller.
Who enforces the EU AI Act?
Each EU country appoints national market surveillance authorities to enforce the AI Act in its territory, and they handle most cases involving businesses. Spain, for example, created AESIA, a dedicated AI supervision agency. At EU level, the European Commission's AI Office oversees general-purpose AI models like the ones behind ChatGPT or Gemini, and the European AI Board coordinates the national authorities so the rules are applied consistently.
What counts as high-risk AI under the EU AI Act?
High-risk AI covers systems used in areas where errors can seriously affect people's lives: hiring and worker management, education and exams, credit scoring, access to essential services, biometrics, critical infrastructure, law enforcement, migration and justice. AI built into regulated products, such as toys, machinery or medical devices, can be high-risk too. Typical marketing uses, like drafting copy, generating images or running a support chatbot, are not.