A year ago, using AI in marketing was mostly a question of speed and cost. Could it write ad copy faster? Could it produce product images without a photoshoot? Could it power a chatbot that answered questions at midnight? Today, there’s a new question every marketing team needs to answer: when do we have to tell people AI was involved?
It’s a question that affects almost everyone in the industry now. 83% of ad executives report using AI in the creative process, from first drafts and image concepts to voiceovers and video edits. AI has quietly become part of how campaigns are made, often in ways customers never see.
That question now carries legal weight. Since 2 August 2026, the EU AI Act’s transparency rules have applied to businesses using AI to create content or interact with people in the EU, with fines of up to €15 million or 3% of worldwide annual turnover for non-compliance. And Europe isn’t alone: California’s own AI transparency law took effect on the same day, which means a single campaign running across markets can face several sets of rules at once. At the same time, the advertising industry has published its own guidance on when AI should be disclosed, and when labelling might do more harm than good.
For most brands, the answer is more reassuring than the headlines suggest. Not everything touched by AI needs a label. But the things that do need one are exactly the assets customers judge you on most.
In brief
- The EU AI Act’s Article 50 transparency rules have applied since 2 August 2026
- Deepfakes, AI chatbots and certain AI-generated content must be disclosed
- Routine AI use, such as drafting copy or editing images, generally doesn’t need a label
- Over-labelling can hurt performance, so the goal is targeted, honest disclosure
What Changed on 2 August
Article 50 of the AI Act requires transparency in four areas: direct interaction with individuals, AI-generated content, emotion recognition and biometric categorisation, and deepfakes and AI-generated text on public-interest matters. These obligations apply to all AI systems in scope from day one, although content generated and published before 2 August 2026 doesn’t need to be labelled retroactively.
The rules aren’t limited to large AI companies. A business that runs a chatbot on its website, produces marketing creative with generative AI or publishes model-written text has transparency obligations, even when the tools belong to third parties. They also reach businesses wherever they’re based, so a company outside the EU serving EU users is still caught.
For marketers, the most important definition is the deepfake: AI-generated or manipulated image, audio or video that resembles real people, objects, places or events and would falsely appear authentic. Clearly fantastical content, like dragons or people flying unaided, falls outside that definition.
To help businesses comply, the EU has published a Code of Practice and created a set of icons deployers may use to label AI-generated content. Around 190 organisations had signed the code by the end of July.
The Industry’s Answer: Label What Matters
On 18 August, the IAB released version two of its AI Transparency and Disclosure Framework, which calls for disclosure when AI materially affects authenticity, identity or representation, rather than labelling every use of AI in marketing. The framework is voluntary, but it gives brands a practical decision process that sits comfortably alongside the EU rules.
According to the framework, disclosure is recommended for images and videos generated from prompts, some synthetic voices, synthetic avatars, digital twins of deceased people, digital twins of living people in fabricated situations, and AI chatbots or assistants consumers could mistake for humans.
Just as useful is the list of what generally doesn’t need a label: routine post-production, internal workflows, clearly fantastical imagery, authorised synthetic voices of real people, generic synthetic voices, background music, standard audio enhancement, digital twins in normal brand endorsements, obvious cartoon or stylised avatars, and text or copy.
“The AI you have to disclose is usually the AI you give a face or a voice.”
Why Labelling Everything Isn’t the Answer
It might seem safest to put an AI label on everything. The IAB warns against it. As its vice president of AI put it, labelling everything teaches consumers to ignore labels and could negatively impact advertisers. The IAB framework also cites an NYU study reporting that AI labels on ads cut click-through rates by 31.5%, which shows why unnecessary labels carry a real cost.
At the same time, hiding AI where it matters is riskier still. More than half of consumers surveyed said they wanted brands to disclose when an ad was fully AI-generated or used AI imagery or video, and 73% of Gen Z and Millennials said clear disclosure would increase or have no effect on their likelihood to buy. Honest disclosure, used in the right places, doesn’t seem to put customers off.
What does put them off is feeling deceived. Research from Raptive found that content readers suspected was AI-generated cut their trust by nearly half and reduced purchase consideration by 14%. In other words, the damage comes less from using AI and more from customers discovering it on their own.
What It Means for B2B Brands
For many B2B companies, this is mostly good news. As one analysis pointed out, most B2B AI work is text, such as emails, ad copy, landing pages and sales decks, and the IAB says that kind of work doesn’t need a label. The disclosure question lands on a narrower set of assets: the synthetic spokesperson in a product video, the AI voice in a webinar recap, or the chatbot on your pricing page.
If your website uses an AI chatbot, that’s probably the first thing to review. Anyone talking to it should know clearly, from the first message, that they’re speaking with AI rather than a person.
Ad platforms are also shifting responsibility onto brands. Google’s latest documentation places the disclosure duty on advertisers rather than publishers, giving them an AI label setting across five products, so it’s worth checking those settings before your next campaign goes live.
A Simple Test Before You Publish
Before any AI-assisted asset goes live, a few questions go a long way. Does it show a realistic person, voice, place or event that didn’t actually exist or happen? Could a customer reasonably believe they’re talking to a human when they aren’t? Would someone feel misled if they later learned AI created it? If the answer to any of these is yes, add a clear disclosure. If AI simply helped you write, edit or polish something, a label usually isn’t needed.
It’s also worth keeping a simple record of where AI is used across your campaigns and which assets were labelled. If a regulator, platform or customer ever asks, you’ll have a clear answer.
Your Questions Answered
Do I need to label AI-generated text in marketing? Generally not. The IAB framework lists text and copy among uses that don’t need a label, and the EU AI Act’s text rule focuses on AI-generated content published to inform the public on matters of public interest, rather than everyday marketing copy.
Do AI-generated product images need a label? If they’re realistic images generated from prompts that could be mistaken for real photography, the IAB recommends disclosure. Routine editing, such as colour correction, doesn’t need a label.
Does my website chatbot need to disclose that it’s AI? Yes. Transparency about direct interaction with AI is one of the four areas covered by Article 50, and the IAB also recommends disclosure for chatbots consumers could mistake for humans.
What are the fines for breaking the EU AI Act labelling rules? Up to €15 million or 3% of total worldwide annual turnover, whichever is higher, with the lower figure applying to SMEs and start-ups.
Do old campaigns need to be relabelled? No. Content generated and published before 2 August 2026 doesn’t need to be retroactively marked or labelled.
This article is general guidance, not legal advice. For specific compliance decisions, it’s worth speaking with a qualified legal adviser.
Transparency as a Brand Asset
It’s easy to see these rules as another layer of compliance. But at their core, they reflect something marketers have always had to protect: trust. Being clear about where AI plays a role, while avoiding unnecessary labels or explanations, can help brands communicate with greater confidence and credibility.
For most teams, that starts with a simple understanding of where AI is being used, where transparency matters, and how those decisions are being documented. The goal isn’t to make every piece of content about AI. It’s to use it thoughtfully, with enough clarity for audiences to understand what they’re seeing and hearing.
As AI becomes a more familiar part of marketing, that balance between transparency and practicality will become increasingly important. For more perspectives on how AI is shaping digital marketing, explore our articles on LinkedIn and AI slop and ChatGPT ads.
