AI disclosure in ads is becoming a basic part of brand trust. Audiences in 2026 are mostly fine with AI imagery, and they react badly when they feel tricked, as the wave of AI ads at Super Bowl LX showed. Platforms already label AI content, and the EU AI Act's transparency rules for deepfakes apply from 2 August 2026, so the simplest rule is to label anything realistic that a viewer could mistake for real.
Super Bowl LX showed how quickly audiences tire of AI.
Super Bowl LX on 8 February was meant to be a showcase for AI. iSpot counted 15 of the game's 66 national ads featuring AI in some way, either selling AI products or made with it. Social listening by Meltwater found close to half of the conversation about those ads was negative, with viewers calling many of them uninspired or low quality, and fans joked within the first quarter that every break was another AI pitch. Many of the spots felt alike and slightly hollow. People are sensitive to being sold to by something that does not seem to care.
Disclosure is about trust, and trust is about not being fooled.
An AI-generated mountain backdrop behind a car is unlikely to bother anyone. A realistic AI person telling you how a product changed their life is a different matter. The more realistic the content, and the closer it gets to a claim or a real person, the more a viewer feels cheated when they find out later.
That gives brands a useful test. Ask: if this viewer later learned the shot was AI, would they feel misled? If yes, label it. If the shot is obviously stylised, or the AI was only used to clean up a background, a label is usually optional.
The platforms already label AI content.
YouTube asks creators to disclose realistic altered content.
Since March 2024, YouTube has required creators to disclose realistic content that is made or altered with AI, such as a real person saying something they did not say, or a realistic event that never happened. The label appears in the description or, for sensitive topics, on the player. YouTube can also add the label itself. Clearly unrealistic content and simple production help such as colour correction are exempt.
Meta shows an "AI info" label.
Meta labels AI content on Facebook, Instagram and Threads when its systems detect industry-standard signals or when users disclose it. The original "Made with AI" tag was renamed "AI info" in July 2024 after photographers complained that lightly edited photos were being flagged.
TikTok requires creators to label AI-generated content.
TikTok introduced rules for AI-generated content in 2023 and asks creators to mark realistic synthetic content with its "AI-generated" label.
For brands, the practical point is that platforms may label your ad whether you do or not. It looks much better if you got there first.
The EU AI Act makes deepfake disclosure a legal duty from August 2026.
Under Article 50 of the EU AI Act, anyone deploying AI to create image, audio or video content that counts as a deepfake, meaning realistic content resembling real people, places or events that could falsely appear authentic, must disclose that it was artificially generated or manipulated. These transparency duties apply from 2 August 2026. There are lighter rules for work that is evidently artistic, creative or satirical, though disclosure is still expected in a way that does not spoil the work. Brands running campaigns in Europe should plan for it now, and brands in the Gulf and Egypt with European audiences would do well to adopt the same habit.
Here is how we decide when and how to label AI.
Joe G. runs AI production at iklipse, and his team sorts every AI asset into three buckets before it leaves the building.
- Assistive. Clean-ups, extended backgrounds, colour, upscaling. No label needed, the same as normal retouching.
- Stylised or obviously generated. Fantasy worlds, illustrated looks, impossible camera moves. A light "made with AI" credit is good manners, and on most platforms optional.
- Realistic people, places or events. Synthetic humans, realistic scenes that never happened, cloned voices. Always labelled, on screen or in the caption, plus the platform's own AI tag.
We also apply the rule from part one of this series: AI never shows a product result that is not real, labelled or not. A label does not make a misleading claim acceptable.
Our own AI models were built openly as AI.
Eve and Nova, the AI influencer models we created in 2024, were introduced from day one as AI-generated. That was the point of them. Nova is sun and energy, Eve is night and mystery, and the craft is in how real they look. Nobody needs to be fooled for that to be impressive. See them at our Eve & Nova page.
Labelling AI well is a craft skill.
A clumsy disclaimer can ruin a film. A good one is short, honest and part of the design: a small line in the end card, a mention in the caption, a behind-the-scenes cut that shows how the work was made. Audiences often enjoy that last one most. Our Social Media Management team plans these into the content calendar, and our AI-Infused Production team builds the work with labelling in mind from the brief.
iklipse is a creative and AI production agency working across the US, the GCC and Egypt. If you are unsure whether a campaign needs an AI label, ask us.
Frequently asked questions.
Do I have to label AI-generated ads?
Platform rules on YouTube, Meta and TikTok already ask for labels on realistic AI content, and in the EU, deepfake disclosure becomes a legal duty from 2 August 2026. Elsewhere, labelling realistic AI is the safest habit.
Does using AI to retouch a photo need a label?
Usually not. Minor edits such as colour correction and clean-ups are generally treated like normal retouching, provided they do not change what the product does.
Will an AI label hurt my ad's performance?
There is no solid public evidence either way yet. What is clear is that being caught without one damages trust.
Sources: iSpot via Adweek, Meltwater, Rolling Stone, YouTube (Google blog), Meta, TikTok, EU AI Act Regulation (EU) 2024/1689, Greenberg Traurig, McCann FitzGerald.
