Seeing Isn’t Believing Anymore: Deepfakes and the Collapse of Media Trust

Just 56 percent of American adults now say they trust national news organizations at least somewhat — down 20 points since 2016. Meanwhile, the number of deepfake videos circulating online is projected to hit roughly 8 million by the end of this year, up from about 500,000 two years ago. Those two lines on the graph are not a coincidence. They’re the same story told from opposite ends.

The Floor Keeps Dropping

Pew Research has been asking Americans about their trust in national news for years, and the trend line has only gone one direction. Trust sat at 56 percent this year, and it’s not just a partisan story anymore — while only 44 percent of Republicans and Republican-leaning independents say they trust national news organizations, Democratic trust also slid sharply, dropping from 81 percent to 69 percent in a matter of months. When both sides of the aisle are losing faith at the same time, even if from different starting points, the problem isn’t red or blue. It’s structural.

Local news still fares better, with 70 percent of Americans expressing at least some trust — but even that figure has slipped from 82 percent in 2016. There is almost nowhere left in the information ecosystem that hasn’t lost ground. Adults under 30 now trust information from social media (52 percent) at nearly the same rate they trust national news organizations (56 percent), which tells you less about social media getting more credible and more about journalism losing the benefit of the doubt it used to carry by default.

“We used to argue about what the news meant. Now we argue about whether the video is real. That’s not a smaller disagreement — it’s a much deeper one.”

Then the Video Itself Stopped Being Proof

Layer the deepfake numbers on top of that eroding trust and the picture gets genuinely unsettling. Verified deepfake incidents climbed from just 22 total between 2017 and 2022 to 42 in 2023, then 150 in 2024 — and 179 in the first quarter of 2025 alone, already surpassing the entire prior year. Fraud researchers at Resemble AI logged over 2,000 verified incidents in a single quarter of 2025. Deepfake-enabled fraud losses in the U.S. reportedly tripled year over year, from roughly $360 million to $1.1 billion.

For most of modern history, a photograph or a video clip functioned as a kind of trump card in an argument: seeing was believing, full stop. That card doesn’t work anymore. A convincing fake of a CEO announcing a merger, a politician saying something they never said, or a phone call from a “family member” asking for emergency cash is now cheap and fast to produce. The technology got good at almost exactly the moment institutional trust was already circling the drain, and the combination is worse than either problem alone. It’s not just that people believe fewer true things. It’s that they’ve lost the tool they used to use to sort true from false in the first place.

The Liar’s Dividend

Researchers have a name for the darkest part of this: the “liar’s dividend.” Once fakes are common enough, real footage of real wrongdoing becomes easy to dismiss as fabricated. A genuine recording of misconduct can be waved away with a shrug and a “that’s probably AI.” The technology doesn’t just create false belief — it provides cover for denying true ones. That is a much harder problem to solve with a fact-check, because the doubt itself becomes the weapon, and doubt spreads faster than any correction ever could.

The Index tracks media trust and synthetic content together because they’re feeding each other in a loop that’s hard to break from either end. A public that already doubts the news is more vulnerable to convincing fakes; a flood of convincing fakes gives that same public one more reason to doubt everything, real footage included. A society can survive disagreeing about what things mean. It has a much harder time surviving when it can’t agree on what actually happened.