A person seen from behind at a sunny window, holding a phone that shows a bright, dreamlike edited photo of a seaside town
AI-generated image, not a photograph of a real person or place.
aiphotofun

AI PHOTOS, LABELLED

Have fun,
then check the label.

LOOK AGAIN before you share

  • Eyes are not enough
  • Labels travel with files
  • No label is not proof
  • Enjoy the tools, share with care

AI pictures can now pass for real photos.

In one study, people judged AI-made faces correctly less than half the time. So the industry is turning to labels instead: signed records, hidden watermarks and plain tags that say how a picture was made.

Updated

Read a label

THE SHORT ANSWER

Your eyes can be fooled.
Labels help.

Looking hard is not a reliable test any more.
Three kinds of label do the work instead.

YOUR EYES

Close to a coin toss

In a study of 315 people, AI-made faces were classed correctly 48.2% of the time, about chance. Training lifted it only to 59%.

See the numbers

CONTENT CREDENTIALS

A signed history

The open C2PA standard attaches a signed record of how a file was made and edited. Change the picture and the record no longer matches.

How to check

THE LAW

Deepfakes must say so

Under the EU AI Act, AI-made content must be identifiable and deepfakes clearly labelled. The transparency rules apply from August 2026.

What the rules say

READ A LABEL

What does the
source tag say?

Photo files can carry a standard tag for how the picture was made, used inside Content Credentials. Pick one to see what it means. The bar shows how much of the picture came from AI.

Source tag reader

Choose a tag
Digital captureCamera original

Captured from real life with a digital camera or recording device. This is the plain photo.

Computational captureReal scene, merged frames

Several frames of a real scene merged automatically into one, using signal processing or non-generative AI.

Algorithmically alteredTidied, not changed

Sharpened or de-noised by an algorithm without changing the main content of the picture.

Edited using generative AIPartly AI

A real picture augmented or extended with a generative AI model, for example by filling in or extending parts of the scene.

Created using generative AIMade by AI

The whole picture was created by an AI model trained on captured content. No camera saw this scene.

Source: IPTC NewsCodes, Digital Source Type controlled vocabulary (names and definitions as published by IPTC).

Based on the IPTC Digital Source Type vocabulary. Not every app writes these tags, and some sites remove them.

TWO NUMBERS WORTH KNOWING

Fakes are common,
spotting them is hard.

One figure on how often people meet deepfakes,
one on how well we can tell.

43%

Saw at least one deepfake in six months

Ofcom found 43% of people aged 16+ said they had seen at least one deepfake online in the last six months, rising to 50% of children aged 8 to 15.

UK ADULTS · OFCOM 2024

9%

Confident they could spot one

Fewer than one in ten people aged 16+ said they were confident in their ability to identify a deepfake. Children aged 8 to 15 were more likely to say so, at 20%.

CONFIDENCE · OFCOM 2024

In a 2022 PNAS study, people also rated AI-made faces as slightly more trustworthy than real ones: 4.82 against 4.48 on a seven-point scale.

FOUR HABITS TO TRY

A quick check
before you share.

Hands holding a smartphone over a laptop keyboard while comparing a landscape photo on both screens
AI-generated image, not a photograph of a real person or place.

None of these needs special skills. They are about using the labels that already exist, and knowing what they can and cannot tell you.

  1. Look for the pin or label

    A Content Credentials pin, or a platform's AI label, means the file carries information about how it was made. Open it to see the history.

  2. Upload it to a checker

    Content Credentials verify tools read a file's signed record and show whether it was created or edited with AI, and whether it has been tampered with.

  3. Treat no label as unknown

    Metadata can be removed, so a missing label does not mean a picture is real. It just means you have no record either way.

  4. Label your own AI fun

    If you share an AI-made or AI-edited picture, say so in the caption, especially if it shows a real place or a person.

WHAT LABELS CAN AND CANNOT DO

A record of history,
not a verdict on truth.

Two printed portrait photographs of different anonymous adults side by side on a wooden desk next to a magnifying glass
AI-generated image, not a photograph of a real person or place.

The C2PA explainer is clear that Content Credentials do not judge whether a picture is true. They show whether the record is well formed, untampered and signed by a known signer. Because metadata can be stripped, the standard also allows invisible watermarks and fingerprints so the record can be found again.

Ofcom describes the same toolkit for tech firms: invisible watermarks, metadata attached at creation and visible AI labels on upload, alongside detection and clear platform rules. It says there is no silver bullet, and that tackling harmful deepfakes needs several approaches together.

Questions

Questions people ask

Can I tell an AI photo just by looking?

Not reliably. In a 2022 study published in PNAS, 315 people classed real and AI-made faces correctly 48.2% of the time, close to chance. With training and feedback, a new group reached 59%.

What are Content Credentials?

An open standard from the Coalition for Content Provenance and Authenticity (C2PA). A signed record travels with the file, describing its origin, edits and any use of AI. Any change to the file or the record breaks the match.

If a photo has no AI label, is it real?

Not necessarily. The C2PA explainer confirms that provenance metadata can be removed, so a missing label only means there is no record. That is why the standard also supports invisible watermarks and fingerprint lookups.

Do AI images have to be labelled by law?

In the EU, the AI Act says providers of generative AI must make AI-made content identifiable, and deepfakes must be clearly and visibly labelled. The European Commission says these transparency rules apply from August 2026.

What is a deepfake?

Ofcom defines deepfakes as videos, pictures or audio clips made with artificial intelligence to look real. Many are harmless fun or satire, but some are used to demean people, run scams or spread false information.

Sources

Where the facts come from

  1. Ofcom: A deep dive into deepfakes that demean, defraud and disinform (July 2024)
  2. Nightingale and Farid (2022), AI-synthesized faces are indistinguishable from real faces and more trustworthy, PNAS
  3. C2PA and Content Credentials Explainer (specification 2.2)
  4. IPTC NewsCodes: Digital Source Type vocabulary
  5. European Commission: AI Act, regulatory framework for AI

ABOUT THIS NAME

Interested in leasing aiphotofun.com?

aiphotofun.com is part of the DoodleSale portfolio. It isn't in an auction right now, but you can ask about leasing it.

Ask about this name