Turkish advertising woke up to a new rulebook on 1 August. The amendment the Ministry of Trade published in the Official Gazette on 1 July is now in force, and for the first time it speaks directly to AI-generated advertising and targeted advertising. If your team generates creative with a model, puts digital characters in front of consumers or serves ads based on user data, there are a few new articles you now need to read.
What changed
The package is wide. AI and targeting draw the attention, but they are not the whole story:
| Area | New rule |
|---|---|
| AI in ads (Art. 18/8) | AI or software use that materially affects consumer economic behaviour, and digital characters indistinguishable from humans, must be disclosed clearly, comprehensibly and distinguishably |
| Digital replica ban (Art. 27/12) | Ads in which an AI-generated replica of a real person falsely appears to use, experience or endorse a product are prohibited |
| Targeted advertising (Art. 25/A) | Consumers must get directly and easily accessible information on the targeting criteria and how to change them; profiling-based targeting of children is banned |
| Influencer advertising (Art. 23/A) | The 2021 Guideline codified: “Advertisement/Promotion” labels, advertiser identity, per-post marking |
| Environmental claims | “Sustainable”, “natural” and similar claims now require documentary substantiation and explanation |
| Discounts and reviews | Pre-discount price is the lowest of the last 10 days; response window for negative consumer reviews cut to 48 hours |
Which uses of AI need a disclosure?
The new paragraph in Article 18 covers two situations: using AI or other software in a way that materially affects the consumer’s economic behaviour, and using AI-generated digital characters that cannot be told apart from a human. Where either is present, the ad has to say so clearly, comprehensibly and in a way the consumer can actually notice.
The difficulty is that nobody yet knows where “materially affects” begins. The text offers no definition, and the line will be drawn over time by Advertising Board decisions. A reasonable reading is still available. Optimisation tools running in the background of media buying, which never change what the consumer sees, are unlikely to need a disclosure. Product visuals, campaign copy or voiceover produced with a generative model sit in more debatable territory once they shape how authentic the content feels. And for a presenter who cannot be told apart from a human, a virtual influencer or a customer testimonial written by a model, there is no real debate; those are exactly what the article was written for. Until the Board draws its line, our suggestion is a practical one: if whether the thing is real or generated could move the purchase decision, label it.
For digital replicas, disclosure is not enough
The new paragraph in Article 27 works differently. It is not a duty you can satisfy with transparency; it is simply a ban. Ads in which the AI-generated replica of a real person is shown using, trying or recommending a product that person never touched can no longer be run at all. The provision is aimed at the wave of unauthorised celebrity deepfake ads of the past couple of years. We covered the personality rights, copyright and criminal law sides of that picture in our deepfakes piece; the Regulation has now added the consumer law side.
The new standard for targeted advertising
Targeting was not banned; it was defined for the first time and made conditional. The definition is broad: a seller, supplier or intermediary service provider analysing online behaviour, preference history, location, demographics or similar personal data to serve ads to particular people or groups. The condition is that the consumer can easily find out which criteria put the ad in front of them and how to change those criteria. A line saying “we personalise your ads” does not meet it.
One distinction is worth keeping clear: this rule does not replace data protection law. Whether you may process the personal data at all is still a question for Law No. 6698; the Regulation adds a separate transparency duty on the advertising side. However good your privacy notice is, an ad interface with no criteria information leaves you short, and the reverse is just as true. On children the drafters left no room for argument: where the advertiser knows, or reasonably should know, that the consumer is a child, targeted advertising based on profiling cannot be done. Gaming, education and entertainment apps whose audiences plausibly include children would do well to reread their targeting settings with that ban in mind.
Reading it next to Article 50 of the AI Act
What makes this August interesting is that two similar rules arrived in two legal systems almost at once; the AI Act’s transparency article has applied on the EU side since 2 August. At first glance the two texts do the same job, but they trigger differently, and a team advertising into both markets will want one design that satisfies both.
| Issue | Turkish Regulation | AI Act Art. 50 |
|---|---|---|
| Trigger | Material effect on economic behaviour, or human-indistinguishable characters | No effect threshold: AI interaction and synthetic content are marked as a rule |
| Form | Clear, comprehensible, distinguishable statement | Machine-readable marking plus user-facing disclosure |
| Deepfakes | Replica ads creating a false impression of use or endorsement are banned | Deepfakes must be disclosed; national laws remain applicable |
| Enforcement | Advertising Board: suspension, correction, administrative fines | Market surveillance authorities; turnover-based fines |
On the ground, if you build one layer in your content pipeline that marks synthetic content in a single place, the same layer can feed both the Turkish disclosure language and the AI Act’s machine-readable format. That is cheaper than chasing the two regimes separately, and it leaves less room for something to slip through. We walk through that architecture at the AI Compliance Hub, and our AI customer agents piece covers the chatbot-side notification duties.
The opposite risk: claiming AI you do not have
The disclosure duty also has a mirror image. Calling a product “AI-powered” in advertising when it has no meaningful AI component is misleading advertising, and it causes separate trouble in investor communications; the practice has a name, AI washing. As the new rules make AI use visible, AI claims in marketing copy become easier to check as well. Honest labelling is needed in both directions.
Do we need a disclosure if generative AI only drafted the ad copy?
Having a model draft the text is not what the rule is aimed at, as long as it creates no illusion about the authenticity of what the consumer sees; the threshold is a material effect on economic behaviour. Once AI-written experience narratives presented as human, or indistinguishable characters, come into the frame, disclosure is the safest route. Until the Board settles the boundary, we suggest labelling every scenario you are not sure about.
Are consented digital twins banned too?
The ban attaches to a false impression of use or endorsement. A digital twin created with the person’s consent and reflecting a genuine experience is not its direct target, but the Article 18 disclosure duty, personality rights and the licence terms all remain in place. Do not launch a digital twin campaign before the consent paperwork and the scope of use are settled.
Where to start
Start with the campaigns already running. If the creative contains generative AI or a digital character, is it labelled? If you target, can the consumer see the criteria and change them in the interface? If your audience might include children, is profiling-based targeting switched off? Turn the three answers into a one-page note, push it into your agency agreements, and the first enforcement wave will hold no surprises.
Author
If this is on your desk
Templates and checklists are free in the Founder Academy; for a specific situation, book a 30-minute intro call.
Founder AcademyBook an intro callMore from Vircon Insights
Türkiye’s Crypto Advertising and Sweepstakes Rules: The Banned Pages of the Growth Playbook
July 16, 2026Türkiye Writes Gaming into Law No. 5651: A Guide to the New Platform Regime
August 4, 2026From the ICTA to the Cybersecurity Presidency: Türkiye's New Digital Authority Map and the Two-Hour Rule
August 4, 2026The CASP Licence in Türkiye: An Application Guide in Build Order (Law 7518 / CMB III-35/B.1)
August 4, 2026Tax Due Diligence in Türkiye: What Buyers Check and How Sellers Prepare
August 3, 2026AI Washing: When "AI-Powered" Becomes a Legal Claim in Advertising, Competition and the Fundraising Deck
August 3, 2026Related Practice Areas
Corporate Law
Share transfers, capital increases, board structuring, governance.
View service →Mergers & Acquisitions
End-to-end M&A: due diligence, structuring, documentation, negotiation.
View service →ICOs, Crypto & Blockchain
Crypto-asset regulation, token offerings, exchange and custody licensing.
View service →