Free quiz: takes about 2 minutes
Underwater photos: real or AI?Can you spot the AI reef, jellyfish or deep-sea creature?
TimeBinder adds a layer of protection against AI and human fraud and disputes. Six side-by-side underwater photographs, from shallow reefs to the deep sea. In each pair, one is a real photograph and one is AI.
Your score is worked out in your browser. Real photos are CC0 or public-domain photographs. The AI images were made for this quiz and show no real person.
Pick the AI photo in each pair
Tap the photo you think is AI-generated. You will see the answer and a tip straight away.
Takes about 2 minutes. Your score is worked out in your browser and is not tied to your name.
Six rounds. Two underwater photos each. One is AI.
Each round shows two underwater photographs: reefs, kelp, jellyfish or deep-sea animals. One is a real photograph. The other was generated by AI. Pick the one you think is AI, then see what gave it away.
- The real photos are CC0 or public-domain photographs from Wikimedia Commons, including deep-sea images from NOAA, listed with photographer and licence at the bottom of this page. No real people are shown, and no logos or brands are used.
- The AI images were made for this quiz from text descriptions, not as copies of the real photographs. Both sides were cropped to the same shape. Sides are shuffled every time you play.
Which photo is AI-generated?
What gave it away here
So can you trust your eyes on an underwater photo?
Not on their own. Real underwater photos are often blurry, blue-green and murky, and generated ones can be sharp and dramatic or just as murky. Clues help on a still image, then fade once a photo is cropped, filtered or re-shared.
TimeBinder is not a tool for identifying generated media, and it cannot tell you whether AI made a file. A TimeBind shows only that an exact file existed no later than a recorded time and that a copy you hold still matches it.
First Pooled TimeBind free, no card, then from US$5.
Can you tell AI underwater photos from real ones by looking? The short answer
Often not. Underwater footage is a common test of what viewers believe. A news report (The Guardian, 24 October 2026) said a wildlife documentary series would keep murkier, less polished underwater shots to reassure viewers that its footage is real. This quiz uses no footage from any series, and no real person is shown. This quiz uses no news photo, no real person and no named brand. It is a practice test only.
Spotting by eye is unreliable. TimeBinder is not a tool for identifying generated media. A TimeBind shows only that an exact file existed no later than a recorded time and that a copy you hold still matches it. This quiz cannot check a photo you found online or a file you hold.
Five clues in underwater photos, and why none is proof
These are the habits that showed up in the quiz. They are worth knowing, but each one fails often enough that you should never settle a “is this photo real?” question on them alone.
- Fish and corals spread too evenly. Many fish of the same size and neat coral shapes. Real reefs can look orderly in clear water.
- A perfectly centred, glassy subject. A jellyfish or animal placed in the middle with a smooth background. Real photographers centre subjects too.
- Kelp or plants that frame the subject. Blades curling neatly around an animal. Real kelp tangles, but it can also fall that way.
- A light that fades in a smooth circle. A spotlit pool of light with a smooth dark edge. Real deep-sea cameras have a hard light too.
- Crisp, colourful, crowded frames. Every part of the frame in focus and full of colour. Real deep-sea coral gardens can be colourful, and video stills can look soft or sharp.
All five get weaker once an image is cropped, compressed and re-shared, which is how most people see photos online. If a photograph matters later, the stronger record is the original file and when it existed, not a guess from the way it looks.
If a photo matters, keep a record of the original
When a photo is questioned, an argument about how it looks goes nowhere. A better question is whether you can show the exact original file and when it existed.
- Keep the original file untouched. Do not edit or re-save it. Work on copies.
- Make a TimeBind when you take or finish it. Your browser makes the SHA-256 fingerprint, and the file stays on your device. TimeBinder records the fingerprint with a time on the Bitcoin blockchain.
- Keep the file and the Proof of Time Certificate together.
- Check later with the file comparison checker. A copy that still matches is the same file. One that does not match is a different file.
First Pooled TimeBind free, no card, then from US$5. An ASAP TimeBind is US$15.
What a TimeBind shows, and what it does not. It shows that an exact file existed no later than a recorded time and that a copy shown later still matches its fingerprint. It does not show who made the file, who owns it, that its contents are true, or whether AI was used to make or edit it. This is general information, not legal advice.
Image credits and licences
Each AI image was generated by TimeBinder for this quiz from a text description of an invented scene, not as a copy of the real photograph. No real people, divers, brands or logos are shown. The deep-sea photographs are from NOAA, a United States government agency, and NOAA video overlays were cropped off. Real photos are CC0 dedications or public-domain works, listed with their licence in the table below, and both permit reuse and modification. Real photos were cropped to a 4:3 frame, and the AI images were cropped to the same shape, so image size and shape do not give the answer away. All image metadata (EXIF) was removed from both sides.
| Subject | Real photo source | Licence | Photographer / uploader |
|---|---|---|---|
| A shallow coral reef | Wikimedia Commons | CC0 | 74papa |
| A jellyfish in open water | Wikimedia Commons | CC0 | Jeremy Bishop tidesinourveins |
| An anemone among kelp | Wikimedia Commons | Public domain | National Marine Sanctuaries |
| An octopus on the seafloor | Wikimedia Commons | Public domain | National Marine Sanctuaries |
| Deep-sea corals on a seamount | Wikimedia Commons | Public domain | NOAA Office of Ocean Exploration and Research (Deep-Sea Symphony: Exploring the Musicians Seamounts) |
| A deep-sea star on the seabed | Wikimedia Commons | Public domain | NOAA Okeanos Explorer |
For something on top of the water, the free Swimming pools: real or AI? quiz is six rounds. It cannot tell you whether a file you hold is AI-made.
Underwater photos: real or AI? quiz: FAQ
How does the Underwater photos: real or AI? quiz work?
You get 6 rounds. Each shows two photographs side by side. One is a real photograph and one was generated by AI. Tap the one you think is AI, see which was which, read what gave it away, and get a score you can share. There are 6 rounds because the quiz has 6 photo pairs, and each is used once per game.
Where do the real photos come from?
They are CC0 or public-domain photographs from Wikimedia Commons, including deep-sea images from NOAA, listed with the photographer and licence in the image credits table on this page. The AI images were generated by TimeBinder for this quiz from text descriptions, not as copies of the real photographs. Both sides were cropped to the same 4:3 shape and stripped of metadata so that size and shape do not give the answer away. No real people are shown.
Why this topic?
Underwater footage is a common test of what viewers believe. A news report (The Guardian, 24 October 2026) said a wildlife documentary series would keep murkier, less polished underwater shots to reassure viewers that its footage is real. This quiz uses no footage from any series, and no real person is shown. This quiz is a visual practice test. It cannot tell you whether a photo you found online is real.
Can people tell AI photos from real ones by looking?
Often not. Clues such as odd lettering, over-smooth texture, too-perfect symmetry or a very tidy composition help on a still image, but they are hints, not proof, and each one also appears in real photographs. They fade once a photo is cropped, filtered or re-shared.
Does TimeBinder tell me whether a photo is AI-made?
No. TimeBinder is not a tool for identifying generated media, and it cannot tell you whether AI made a file. A TimeBind shows that an exact file existed no later than a recorded time and that a copy you hold still matches it. That is useful when you can say “here is this file, recorded by this time,” not when you need to know how the picture was made.
Do you store my score or answers?
Your score is worked out in your browser. This page sends a few anonymous events to our analytics, such as the quiz being finished with a score band, never your name or email.
How can I show a photo is the file I kept?
Keep the original file unedited, and make a TimeBind of it when you take or finish it. Later, a copy that still matches the fingerprint is the same file, and the record shows it existed no later than the recorded time. It does not show who took it, how it was made, or who holds the copyright. Read how to prove a photo is not AI for the honest limits.
Related guides and tools
- Cakes: real or AI? (quiz)4 rounds
- Real or AI? wildlife photos (quiz)Six rounds of animal photographs
- Strange animals: real or AI? (quiz)Six rounds of odd animals
- Real or AI? plants (quiz)Six rounds of plant photographs
- Could you spot these viral AI fakes? (quiz)Ten questions about recent cases
- Was this file changed? File comparison checkerCheck whether two files are identical
- How to prove a photo or video isn’t AI-editedThe honest limits of a timestamp
- Protect yourself from AI forgeries and deepfakesPlain-English guide for the public
- All free toolsEvery free TimeBinder tool
Date the original photo before it gets copied
Add a layer of protection against AI and human fraud and disputes: a dated, checkable record of the exact file you hold.
- Open the app and sign in with email or Google.
- Add your files. They stay on your computer.
- Seal them and download your Proof of Time certificates.
First Pooled TimeBind free, no card, then from US$5. See pricing.