7,600 Deepfake Nudify App Ads on Facebook: The Real Cost

👁 113 reads

The story reported by Niebezpiecznik sounds like one advertiser’s mistake, but the numbers tell a different story. An account that WIRED tracked on Facebook and Instagram ran 32 ads for an app that lets you upload a photo of a real person and insert their face into a generated sexual scene. All 32 creatives were targeted exclusively at men. None exceeded ten impressions, which sounds reassuring until you multiply it out: 32 ads with a maximum of 10 views each means at most 320 contacts with the message - a microscopic campaign that still passed Meta’s verification. The advertiser’s page was created on August 3 and had zero followers. That didn’t prevent paid promotion from launching.

The cost of one fake nude image

We checked the pricing of apps in this category that can still be found on the open web. The business model is the same everywhere: monthly subscription in the $10-30 range, or a credit package where one generation costs 2 to 10 credits, and a pack of 100 credits goes for about $10. If a $19.99 monthly subscription allows for one hundred generations, the cost of one fake nude photo is $0.20. In the credit model, at $10 for one hundred credits and five credits per image, you get twenty images for $10, or $0.50 each.

This is the most important number in this entire matter, and we didn’t find it in any coverage. Producing material that can destroy someone’s reputation at school, at work, or in family court costs less than a bus ticket in any major city. For comparison: a single bus fare in most US cities ranges from $2 to $3, meaning the cost of one ride could generate between four and fifteen such images.

The other side of this equation concerns advertising scale. Tech Transparency Project established that accounts linked to one Chinese company ran 7,600 ads for this type of app targeted at Europe. Of 210 pages examined, 179 had ads removed for policy violations - that’s 85.2 percent. We compared these two numbers: 7,600 ads spread across 210 pages averages 36.2 creatives per page, exactly the same order of magnitude as the 32 ads from the account WIRED described. In other words, the described case is not an exception - it’s a statistically typical element of an industrially conducted operation.

Moderation effectiveness is 85 percent, but the other 15 matters

Meta declares that it checks every ad before publication. In practice, this is done by automated systems, and advertisers bypass them with a trivial trick: the first seconds of the material are innocent, the actual content appears later. TTP data shows that 31 of 210 pages never had an ad removed, and some returned to promotion after previous blocks. With 7,600 creatives, even 85 percent effectiveness leaves about 1,125 ads that got through. At an average of ten impressions per creative, that’s 11,250 contacts with an offer to generate someone else’s nudity - and that’s just within one identified operation.

Apple removed one such app from its store, where it pretended to be a photo editing tool and, after payment, opened sexual scenarios. The company explains that “prohibited functions may have been added after verification.” This explanation reveals the weakness of the entire app store model: code is verified at the time of submission, not what the server returns to the user three weeks later.

What you can do if your image is already circulating

The order of actions matters, and most people start with the wrong step - writing to the perpetrator. Start with evidence: screenshot with visible URL, date and time, preferably along with the entire publishing profile. If the content concerns a person under 18, report it to organizations like the National Center for Missing & Exploited Children’s Take It Down service, which accepts anonymous reports through a form and has procedures for cooperation with foreign hosting providers through networks like INHOPE, which private individuals don’t have access to. In parallel, file a report with local law enforcement.

The third step is the platform. Meta has a separate form for intimate content published without consent, and for adults StopNCII.org works by creating a cryptographic hash of the photo locally on your device and passing only that hash to platforms, not the image itself. For minors, the equivalent is the Take It Down service run by NCMEC. It’s worth doing this even if the material is generated - the hash blocks further uploads of that specific file.

Synthetic content labeling takes effect, but doesn’t cover these apps

As of August 2, 2026, provisions of regulation 2024/1689 (AI Act) concerning transparency are in effect, including Article 50, which mandates marking deepfake content in machine-readable format. Penalties for violating transparency obligations reach 15 million euros or 3 percent of global turnover. The problem is that the operation with 7,600 ads was conducted by a company outside the Union, promoting an app accessible through a browser, with no representative in Europe. A sanction imposed on an entity that cannot be located is a number in a document, not a consequence.

The real point of pressure therefore remains the intermediary. Meta as a very large platform is subject to the Digital Services Act, where the maximum penalty is 6 percent of global turnover. With Meta revenues on the order of $164 billion for 2024, this gives a theoretical ceiling of $9.84 billion. That’s a number that has a chance of changing the moderation department’s calculation, as opposed to pursuing an anonymous advertiser with an account created on August 3.

There remains one more effect that no regulation will fix. Since generating a fake recording costs $0.20, every real recording can be challenged with the statement “it’s a deepfake,” and vice versa. Niebezpiecznik cites the example of a Netanyahu recording that someone tried to invalidate with exactly this argument. As long as source marking is not built into recording devices, and only required from generators, both sides of this dispute have $0.20 worth of arguments.

Facts verified September 20, 2026. Sources: Niebezpiecznik, Tech Transparency Project, WIRED, regulation 2024/1689, StopNCII.org.

// read moreHow to Report a Deepfake on Facebook and InstagramGuidesAI Act Deepfake Rules: Who Must Label AI Content From Aug 2026GuidesDeepfake: What It Is, How to Spot It, and Legal ConsequencesGuides

Kinetyra