Synthetic Faces Just Got Good Enough to Stop Being a Joke

A writer at The Verge has staked out a position that feels overdue: AI influencers are no longer obviously fake, and that matters. The piece is framed as a detection problem. I'd argue it's something older and stranger — a credibility problem that social media never actually solved, now wearing a new face.

The early wave was easy to dismiss. The Verge points to figures like Lil Miquela, Imma, and Shudu Gram — names that arrived with enough visual artifice to signal their own constructedness. The uncanny valley was the whole aesthetic. You weren't supposed to mistake them for people; you were supposed to find it interesting that you almost could. Brands played along, audiences played along, and the discourse stayed contained inside a kind of knowing irony. Look at this weird thing. Isn't the weird thing interesting.

That era, apparently, is closing.

The Gap Closes Faster Than the Rules Do

The Verge cites Aitana Lopez — an AI avatar created by a Spanish agency called The Clueless — as an example of where this is heading. The implication is clear: the production quality has crossed some threshold. The tells are getting harder to find. And once detection becomes genuinely difficult, the entire framework of "virtual influencer as disclosed novelty" starts to buckle.

This is the part worth sitting with. The original virtual influencers were transparent about their nature almost by necessity — the technology announced itself. What the piece is describing is a transition from synthetic identity as spectacle to synthetic identity as ambient background noise. That's not an incremental development. That's a category shift.

And the platforms? They're not ready. They were barely ready for the simpler version. Disclosure norms for human influencers are still being argued about in regulatory bodies across multiple continents. Adding "also, some of these people don't exist" to that conversation doesn't simplify anything.

What One Piece Can't Carry

Here's my friction with where the coverage lands: a single newsletter installment noting that AI influencers are getting harder to spot is an observation, not a reckoning. The Verge piece is useful — it names names, it traces the arc from obvious to ambiguous — but the meta-question it opens is bigger than what one source can close.

Who's responsible for disclosure when the creator is a software pipeline owned by an agency? What does "authenticity" mean in a space that was already performing authenticity before the AI arrived? Does it matter to a 19-year-old watching a skincare tutorial whether the person recommending the product has a nervous system?

Those questions don't have answers in this piece. They barely have answers anywhere. Which is precisely why the detection problem is the wrong frame. Detection implies a solution — look harder, build a classifier, add a label. But the deeper issue is that social media influence was always partially synthetic. Lighting rigs, editing bays, PR agencies writing captions. The AI version just removes the human in the middle and makes the whole operation more efficient.

The difficulty of spotting a fake face is a symptom. The condition is that we built an attention economy where the difference between real and constructed was never really the point.

Somewhere, an algorithm is optimizing for engagement, and it genuinely does not care which side of existence the content creator is on.