The age of synthetic content

Telling truth from AI fakes · Lesson 1 / 20

What actually changed

Fakes have always existed: retouching, montage, quotations torn from context. What changed isn't that faking became possible but what it costs. A convincing fake once required skill, time and money; now it takes minutes and no qualifications at all.

Everything else follows from that. Fakes used to be made for large stakes, because they were expensive. Now they're made for small ones: deceiving one person out of a modest sum, inflating reviews, boosting reach. The volume has grown by orders of magnitude, and the quality of any individual specimen has risen enough that the old ways of spotting them no longer work.

Three things that stopped being evidence of truth

Literacy. Typos and clumsy language no longer give away a fraudster — a model writes the text.
The existence of a photograph. "There's a photo" has proved nothing for several years.
A familiar voice. A voice is cloned from a short recording, and over a phone line the difference is inaudible.

People use each of these automatically, without thinking. That's exactly why they're so hard to update: consciously you accept that a photo proves nothing, and you still react the old way.

What to trust now

Verification has shifted from content to provenance. Not "does this look real" but "where did it come from, who published it first, does anyone independent confirm it". That's the only approach that doesn't expire with the next model release.

What not to do

Don't try to become an expert in generation artefacts. The signs that reveal a fake today will be gone within a year — while the habit of trusting your own eye will remain and become dangerous. Invest in the habit of checking the source: that always works.

Insight. What changed is the cost of faking, not its possibility. That's why fakes now serve small ends — and volume has grown faster than quality.

Common mistake. Learning to spot fakes by eye. The signs of generation expire with every model, while the confidence you built in your own eye stays and misleads you.

Pro tip. Move your checking from content to provenance. "Where did this come from" doesn't expire; "does this look real" expires every year.

Cheat sheet

Faking got cheaper — volume and small-scale targets grew.
Literacy, photos and voices stopped being proof.
Check provenance, not appearance.
The skill of "spotting fakes" expires; checking doesn't.

1. What is the main thing that changed?

2. Why is learning to spot fakes by eye risky?

3. Where has verification shifted to?

Task — checked by AI

Recall three occasions in the past year when you believed something online that turned out to be untrue (or the reverse — doubted something genuine). For each, work out: what specifically made you believe or doubt it, whether one of the three expired signals was involved (literacy, the existence of a photo, a familiar voice), and what provenance question you could have asked at the time.

🔒 Answer the question correctly to move on to the next lesson.