YouTube deepfake detection: what changes for brands and SMEs

What changed: likeness detection is now open to everyone

As of May 2026, YouTube has extended its AI likeness detection program to all adult users on the platform. The Verge, which has followed the project's entire evolution, reports the news. Previously, the feature was available only to creators, politicians, journalists and a few specific categories.

The mechanism is relatively simple. Users upload a scan of their face, similar to a selfie, and YouTube's AI system searches the platform's content for matches. If it finds one, the platform sends an alert. From there, the user can request that the content be removed.

YouTube also stated that the number of removal requests generated so far has been "very low." However, the global rollout could change that ratio significantly.

The context: why deepfakes have become a business problem

The spread of generative tools has drastically lowered the technical barrier to producing believable fake videos. You no longer need an advanced post-production team. A few minutes of original footage and an AI model that even non-technical people can use are enough.

According to an analysis published by McKinsey, the adoption of generative AI tools in organizations has grown exponentially over the past two years. As a result, misuse of these tools has multiplied too. Deepfakes are no longer just about celebrities or politicians: they increasingly target professionals, entrepreneurs and mid-profile company figures.

Italian SMEs, in particular, face a specific risk. Many build their communication around recognizable faces: the founder, the sales manager, the professional who appears in tutorial videos or LinkedIn campaigns. Cloning these people can cause reputational damage that is hard to quantify and even harder to contain.

The immediate impact on video content strategies

For SMEs that use YouTube as a communication channel, whether for digital marketing or internal training, this update has direct operational implications. First of all, if you haven't yet registered your face or those of your ambassadors, you should consider doing so.

However, the current feature has a structural limit: it's designed to protect individuals, not the brand as a legal entity. A company can't register "the face of the brand" in the abstract. It has to do so through the real people who represent it.

This creates an interesting asymmetry. Large companies with structured legal teams can manage the process systematically. SMEs, which often have no dedicated oversight, risk reacting only after the damage is done. That's why it makes sense to build deepfake risk management into a broader SEO and content strategy that also includes online reputation monitoring.

How YouTube's system works under the hood

The system is built on a computer vision model that compares the biometric features of the face uploaded by the user with frames from videos on the platform. It's not simple static facial recognition. The model is trained to spot similarities even in altered conditions: different lighting, unusual angles, partial changes to the face.

Like other AI content moderation systems, this tool isn't infallible. YouTube hasn't made the rate of false positives and false negatives public. The platform also hasn't clarified whether the system can detect synchronized audio-video deepfakes, which are the most sophisticated and dangerous type.

To dig deeper into the technical architecture of AI detection systems applied to media, MIT Technology Review has published detailed analyses of the state of the art in this field. Finally, it's worth remembering that YouTube isn't the only platform moving in this direction: Meta and TikTok are developing similar approaches, albeit with different architectures.

What to do now: a practical approach for SMEs

At SHM Studio, we suggest SMEs tackle this issue on three distinct levels, which can run in parallel without requiring extraordinary investment.

• Early registration: company figures who regularly appear in video content should enable the likeness detection feature on YouTube as soon as it becomes available for their account. It costs nothing and offers high protective value.
• Audit of existing content: it's worth checking whether unauthorized content using your company's faces already exists. A systematic analysis of search results, combined with SEO monitoring tools, can reveal unusual amounts of cloned content.
• Internal risk management policy: define who has the authority to request content removal, within what timeframe and through which channels. This process should be documented and built into your company's digital marketing plan.

On top of that, we recommend updating contracts with any ambassadors or external collaborators who appear in your brand's video content. There's a legal dimension to this issue that shouldn't be underestimated.

Ad campaigns: an exposed front

One often overlooked area is paid campaigns. Google Ads campaigns and LinkedIn campaigns that use videos with real faces are particularly vulnerable. A convincing deepfake of a company founder promoting a fake product can cause direct commercial damage on top of reputational harm.

In this scenario, response speed is decisive. Having YouTube's monitoring system already active shortens detection time. That in turn narrows the window in which fake content can circulate and do harm.

For SMEs investing in video advertising, this topic should be part of every campaign brief. Optimizing targeting and copy is no longer enough. You also need to plan how to respond if the content you produce is abused.

What nobody is saying yet: the risk of normalization

One aspect deserves a broader reflection. As tools like YouTube's become standard, there's a risk that the problem starts to feel normal. In other words, companies might hand all responsibility over to the platform and lower their own active vigilance.

But automated systems have inherent limits. They can't assess context, they don't know your brand strategy, and they can't tell satire from an attempt at fraud. Human oversight remains essential.

At SHM Studio, we believe adopting these tools should go hand in hand with a company culture that understands AI risk. This applies to your AI strategy as a whole, not just to protecting video content. In short, technology offers a first layer of defense. But strategy remains a human responsibility.

Outlook: where the market is heading

Between 2027 and 2028, it's reasonable to expect likeness detection to extend to content generated by company accounts, not just personal profiles. It's also likely that platforms like LinkedIn, increasingly central to B2B strategies, will introduce similar tools.

According to Gartner forecasts, by 2028 most major digital platforms will have built synthetic content detection into their standard features. As a result, SMEs that start getting organized today will gain a significant operational edge over those that only adapt once the problem has already become critical.

Finally, it's worth keeping an eye on European regulation. The AI Act already sets transparency obligations for synthetic content. As a consequence, the regulatory framework could impose stricter standards even on platforms that act voluntarily today. To explore the practical implications of these developments, you can contact the SHM Studio team or read the blog for regular updates on AI and digital strategy.