As AI-generated text, images, audio and video become harder to identify, the next layer of online trust may focus less on detecting every fake and more on verifying where authentic content came from.
WHAT’S HAPPENING
Generative AI is making it increasingly difficult to determine whether something online was created by a person, altered by AI or generated entirely by a machine.
Digital-forensics researchers including Dartmouth professor Hany Farid are working on tools designed to identify manipulated and synthetic media. Companies such as GetReal and Pangram are developing systems for detecting AI-generated images, audio, video and text.
But detection has an inherent problem: as generative models improve, detection systems must continually adapt.
That is pushing another approach into the spotlight — content provenance.
Instead of only asking whether something appears fake, provenance technology attempts to establish where content came from, who created it and whether it was modified.
WHY IT MATTERS
The internet was built around distributing information, not proving its origin.
AI changes that equation.
A convincing photograph, voice recording, video or written statement can now potentially be generated without the person or event it appears to represent.
Detection tools can help, but no detector is perfect.
That means digital trust may increasingly depend on something more fundamental:
Can the origin of the content be verified?
WHO BENEFITS
News organizations, businesses, governments, creators and consumers could gain stronger ways to authenticate legitimate digital material.
Technology such as C2PA Content Credentials can attach information about a file’s origin and editing history, while companies including OpenAI are already incorporating provenance signals into some AI-generated media.
For users, that could eventually provide another signal beyond simply deciding whether something “looks real.”
WHO LOSES
Scammers, impersonators and coordinated misinformation operations could face more friction if authenticated content becomes easier to distinguish from material with unknown origins.
But provenance creates its own limitations.
Metadata can disappear. Not every platform supports the same standards. Older content may contain no authentication information at all.
And the absence of a provenance marker does not automatically mean something is fake.
WHAT HAPPENS NEXT
The battle over AI-generated content may gradually move beyond building better fake detectors.
Cameras, AI systems, publishers and online platforms could increasingly attach verifiable credentials to legitimate content from the moment it is created.
That would represent a major change in how the internet establishes trust.
Instead of constantly asking:
“Can we detect the fake?”
The more important question may become:
“Can we prove the real?”
