By Zeynep Tufekci / The New York Times

Current AI technologies have become surprisingly good at creating realistic images and video, unleashing fears that fake images can be used for political and election manipulation.

Well, yes and no.

Fake AI imagery is a challenging problem, and not simply because it looks realistic. The key issue is that these images muddy the waters of credibility for everyone, while providing a handy excuse for political operatives willing to lie to their supporters already eager to believe the lie.

Take Donald Trump’s recent social media post, in which he accused the Kamala Harris campaign of manipulating an image to make her crowd seem bigger at a Detroit airplane hangar last week.

“Has anyone noticed that Kamala CHEATED at the airport?” he wrote. “There was nobody at the plane, and she ‘A.I.‘d’ it, and showed a massive ‘crowd’ of so-called followers, BUT THEY DIDN’T EXIST!”

How do we know an image is real in this day and age? An average person can no longer be certain of the authenticity of images — or increasingly even videos — through individual sleuthing. The AI is that good, and getting better. (That’s why the classic media literacy advice — do your own research — doesn’t work anymore).

This makes it difficult to know what to believe, except through a key mechanism: Trusting sources, and trusting that they have either taken the image or video themselves, or carefully vetted it as authentic.

That’s how we do know that the crowd waiting for Harris is real, because there are pictures from photo agencies like Getty, as well as images and reports from multiple other news organizations that were on the tarmac, that match the circulating social media photos that caught Trump’s ire. We know that credible news organizations and photo agencies have very strict rules about images and videos. But that, in turn, requires trusting the photo agency — or the media source — furnishing the image or video.

It’s no accident that Trump has made it a habit to portray credible news organizations as untrustworthy liars, and many of his supporters seem to have internalized that message they were open to in the first place.

Once trust is lost and all credibility is questioned, the lie doesn’t have to be high quality. It doesn’t have to be supported by highly realistic fake AI. It doesn’t have to be so easily disprovable. To work, the lie just needs a willing purveyor and an eager audience. The AI, then, is but a fig leaf.

This article originally appeared in The New York Times, c.2024.