Editorial correction — September 20, 2026. Dated the trust survey, removed unsupported deepfake totals and loss estimates, and withdrew the claim that parallel trends prove causation.
A convincing recording can deserve investigation without deserving immediate belief. The alternative to credulity is verification, not the assumption that nothing can be known.
Two separate pieces of evidence
Pew’s October 2025 analysis reported that 56% of U.S. adults had at least some trust in information from national news organizations, 20 percentage points below 2016. This survey measures trust; it does not establish that deepfakes caused the change. Read the survey and question wording.
The FBI’s December 2024 warning describes criminals using synthetic audio and video for impersonation and financial fraud. That establishes a documented risk, not a total count of all deepfakes online. Read the warning and protective steps.
The second danger: dismissing genuine evidence
Once fabrication is possible, a person accused of misconduct may suggest that genuine material is fake. That possibility should not grant automatic immunity from scrutiny. An accusation that a recording was fabricated needs examination, just as the recording itself does.
Our editorial standard should work in both directions. We should not accept an explosive clip because it confirms our view, or dismiss an inconvenient one merely because synthetic media exist.
A practical verification sequence
Find the original publication and its context. Look for an identifiable source that can explain how the material was obtained. Seek independent corroboration of the underlying event. Keep the claims separate: whether the file is authentic, whether the caption is accurate, and whether the interpretation follows.
For an unexpected request for money made through a familiar voice, the FBI recommends verifying identity through an independently obtained contact route. A plausible voice or image should not replace that check.
Visible glitches can raise suspicion, but the absence of glitches is not proof of authenticity. A responsible publisher explains what has been verified and what remains uncertain. The moral failure is not admitting that verification takes work. It is using uncertainty selectively to protect the story one already wants to tell.