Picture a car ad. The agency shoots one version with a generic actor pointing at the vehicle. But when you see it, it isn't an actor: it's you. Or a friend of yours. Or it's your own voice saying the name of your city.
This isn't science fiction: it's what a patent Meta filed in October 2025 describes, in some detail. And what happened next is almost more interesting than the patent.
What exactly does the patent say?
Application US20260105492A1, "Personalized advertising generation using user profile data", from Meta Platforms Technologies with Prateek Yadav as inventor, was filed on 14/10/2025 and published on 16/04/2026.
It describes a system that replaces part of an ad — a face, a voice — with content from your own profile.
In plain terms: the advertiser supplies an "advertisement package" containing an image and a portion marked as replaceable. The system picks a replacement image from the user's profile — their own photo, that of a direct contact, or content they've shown an affinity for. For audio it can use their voice, extracted from recordings they had already saved or sent, or insert text such as their name or their city. A generative machine learning model composes the final ad.
The patent text itself is firm on one point: all use of images as replacements is "on an opt-in basis", with prior consent.
The ad isn't chosen: it's generated
This is the leap, and it's worth not skimming past it.
Today, when Meta "personalises" an ad, what it decides is who it shows it to. The ad is the same for everyone sharing that interest: one asset, thousands of people. What the patent describes is different: generating a different asset for each person, inserting content from their profile into the ad.
| Advertising today | What the patent describes |
|---|---|
| You choose who sees it | It's generated for each person |
| One asset, many users | One asset per user |
| Interest-based targeting | Insertion of personal content |
It isn't choosing the ad. It's building it to measure.
So is this going to happen?
Not with this particular patent. The legal events record on the application itself (Google Patents, with data from IFI CLAIMS) shows an entry dated 17 April 2026, one day after publication: EXPRESSLY ABANDONED -- DURING EXAMINATION. Meta expressly abandoned it during examination.
It's worth being honest about what that does and doesn't mean:
| What we know | What we don't know |
|---|---|
| Meta drafted and filed this idea | Why it abandoned it |
| It expressly abandoned it in examination | Whether it's being reworked into another filing |
| It never became an announced product | Whether the technology arrives by another route |
One nuance worth flagging: the sidebar on Google Patents still shows the status as "Pending". That's a label Google infers, and the page itself warns it isn't a legal conclusion. The legal events table, by contrast, mirrors the patent office's own codes.
Abandoning an application is routine and doesn't mean the idea is dead: it happens for strategic reasons, for cost, or to reformulate it into a different filing. But an abandoned patent is not a roadmap, and anyone telling you Meta is "about to put your face in ads" on the strength of this document is making up the ending.
It isn't an isolated idea
The patent doesn't come out of nowhere. It fits with two moves Meta has announced officially and that are genuinely under way:
- Since 16 December 2025, Meta has used conversations with Meta AI as one more signal for personalising content and ads. It explained this in its own release, which also clarifies that it doesn't use sensitive topics — health, sexual orientation, religion, ethnic origin, political views — to show advertising.
- On 7 July 2026 it launched Muse Image, its first in-house image generation model. The announcement indicates that in the following weeks advertisers and agencies would be able to use it through Advantage+ creative.
Profile data feeding ads on one side; in-house image generation heading towards advertisers on the other. Both pieces already exist. The patent merely described the third step: joining them and inserting the person into the ad.
Where this points
Whatever becomes of this particular application, the direction of the industry is the same: advertising is moving from "choosing who I show it to" towards "generating something different for each person". And that changes where the value of a campaign sits.
- Consent and image management stop being separate legal paperwork and become part of the advertising strategy.
- Having real photos, videos and content of your own — not templates or stock — starts to be worth as much as the budget.
- Campaigns differentiate less on the generic message and more on the first-party data and assets each brand brings.
A business with fifty good photos of its own is better prepared for any of these shifts than one relying on stock. And that's already true today, without waiting for any patent to be granted.
Would you like to see yourself in an ad?
Your face, or a friend's, selling the next product you buy.
That's the question this patent — abandoned or not — has already put on the table. And the answer, according to the text of the application itself, came down to a consent checkbox.