How to Spot a Deepfake or Fake Image

The photograph that never happened

You are scrolling and there it is. A public figure doing something appalling. A celebrity recommending a supplement. A photo of a fire in a city you know, posted by an account you have never heard of. It looks real, it feels urgent, and your thumb is already moving toward share.

Generated and altered media is now cheap, fast and good. That is the honest starting point for this course. The useful response is neither panic nor blanket cynicism, because both of those leave you worse off. What actually works is a small set of habits that take about a minute.

The old tells are fading

You have probably seen the advice: count the fingers, check the teeth, look for garbled text on signs in the background, watch how often the person blinks. That advice was genuinely useful a few years ago. It is much weaker now, and it weakens further every few months. Hands got fixed. Text got fixed. Skin and hair got fixed.

So look for glitches by all means, because sloppy fakes are still common. But change what you conclude from them. Finding an obvious error is decent evidence that something is fake. Finding no errors is not evidence that it is real. A clean image is unproven, not proven.

The durable skill is context, not pixels

Four questions survive every improvement in the technology, because they are about the world rather than about the file.

Where did this come from? Not who sent it to you, but the original account or outlet. Scroll back to the source. An account created three weeks ago with a generic name is a very different proposition to a newsroom with a named photographer.

Who benefits if I believe this? Almost every fake that spreads widely is selling something: a product, a political feeling, a share price, or simply your attention. If you can name the beneficiary, slow down.

Is it reported anywhere credible? Genuinely dramatic events get covered by several independent outlets within hours. If an enormous event exists only in one post, that absence is the story.

Is there an older original? A reverse image search is the fastest check available. Very often the picture is real but seven years old, or from another country entirely.

Do not ask a chatbot whether an image is fake

This is worth stating plainly, because it is the mistake almost everyone makes first.

Search the web for reporting on this claim: [paste the claim in one sentence].
Give me:

Every credible outlet covering it, with links and publication dates.
The earliest date you can find for this claim anywhere.
Whether the accounts spreading it look independent of each other.
If you cannot find supporting reporting, say so plainly. Do not fill the gap with a guess.

An image is being shared as [what people say it shows, and when].
Help me check it. List the search terms I should try in a reverse image search,
the specific details I should try to date (clothing, signage, vehicles, weather,
language on shopfronts), and three ways this photo could be genuine but mislabelled.

Asking the sender, without a fight

Most fakes reach you from someone who believes them. Accusing that person rarely works. Curiosity works better.

Write a short, friendly message to [a relative / a colleague] who shared an image
with me. I want to ask where they first saw it, without implying they are gullible
and without starting a political argument. Two sentences. Warm, not lecturing.

When you are still not sure

You will often finish these checks and remain uncertain. That is a normal outcome rather than a failure. In that case, do not forward it. Not sharing costs you nothing. And if you have already shared something that turns out to be fake, post the correction in the same place you posted the original. Everyone gets caught eventually, and quietly deleting helps nobody.