In a 2025 UC San Diego study, people picked GPT-4.5 as the human 73% of the time, more often than they picked the actual humans. ELIZA, a 1960s chatbot, scored 23%, slightly ahead of GPT-4o without a persona at 21%. Here are 20 tests from the Turing Test directory where you can check whether you still tell human from machine across chat, text, faces, art, voice and video.

The Turing test used to be a thought experiment. Now it's a party game, a research method and a product category. Cameron Jones and Benjamin Bergen's three-party Turing test had 284 participants chat with a human and an AI at the same time and then pick the human. GPT-4.5 with a persona prompt won 73% of the time, and LLaMa-3.1-405B won 56%.

The public versions tell the same story. AI21 Labs' "Human or Not?" game ran more than 10 million conversations, and players guessed right only 68% of the time. They were worse at spotting the bots (60%) than at spotting the humans (73%). And it isn't just chat: in a blind listening study covered on HackerNoon, listeners trying to pick out AI-composed piano music were "essentially guessing," at close to 50% accuracy.

So test yourself. A note on what's included: not everything here is a quiz. Some are games, some are detectors that try to make the call for you, and some are the generators that make the fakes. To understand the Turing test in 2026, you need all three.

Chat: can you tell who you're talking to?

1. Human or Not? by AI21 Labs. You get a two-minute chat with a stranger, then guess whether it was a person or a bot. It became the largest Turing test ever run, with more than 1.5 million participants. The research paper breaks down the tricks players used to unmask bots.

2. Loebner Prize. This was the original Turing test competition, held from 1991 according to its Wikipedia history. It's no longer running; its last winner was Mitsuku in 2019. It belongs on any list because it shows how long "fooling the judges" was the goal, and how easy it looks now.

3. Bot or Not (Botometer) by Indiana University's OSoMe. It scores whether a social media account is a bot. It's a Turing test turned into a classifier. The live version shut down after X ended free researcher access, and the Botometer X archive only covers accounts created before June 2023.

Text: human writer or language model?

4. GLTR (Giant Language model Test Room) from the MIT-IBM Watson AI Lab and Harvard NLP. Paste text and it colors each word by how predictable it was to a language model. It was built in 2019 to spot AI-written text, and it's still the best way to see why AI text "feels" smooth.

5. GPTZero. This is the AI-writing detector that teachers made famous. Run your own writing through it to see how often humans get flagged. HackerNoon's guide on how to detect text generated by AI explains what these tools measure.

6. Poem or AI? You read a poem and decide who wrote it. The research on this is humbling: in Porter and Machery's Scientific Reports study, 1,634 readers couldn't reliably tell ChatGPT poems from famous poets', and they rated the AI poems higher.

7. NYT AI Article Quiz by The New York Times. Short passages, one question: human or AI?

8. WritersBrew AI Detector. A writing tool with a built-in check on whether a passage reads as machine-written.

9. Winograd Schema Challenge. This is the benchmark proposed as a better Turing test. Who does "it" refer to in "The trophy doesn't fit in the suitcase because it's too big"? It's easy for people. For years it was hard for machines.

Faces: real photo or GAN?

10. Which Face Is Real? by Jevin West and Carl Bergstrom at the University of Washington. You see two faces and click the real one. It logged about 4 million plays in two weeks after its 2019 launch, and players were right about 70% of the time against StyleGAN fakes.

11. This Person Does Not Exist. This is the generator, not the quiz. Every refresh shows a face of someone who was never born. Look at the ears, teeth and backgrounds to learn the tells.

12. Sensity AI. A commercial deepfake detection platform for faces, video and identity fraud. It's what the professionals use when the stakes are higher than a game.

Art: human artist or image model?

13. AI or Not. Upload an image and get a verdict on whether it's AI-generated. Try it on your own photos first.

14. Artbreeder. A collaborative image generator where you "breed" images by mixing them. Spend ten minutes making fakes and you'll get better at spotting them.

15. MIT GAN Dissect from MIT CSAIL. A research tool that shows which neurons inside a GAN draw trees, doors or clouds, and lets you switch them off. It doesn't test you. It shows you what's inside the machine you're testing.

Voice and music: who's really singing?

16. Real or Fake Voice. Listen to clips and decide which are human speech and which are AI voice clones. This one matters most in practice, because voice cloning is the deepfake behind phone scams.

17. AIVA AI Music. An AI music composer. Play its tracks for friends without telling them and see who notices.

18. Deepfake Detection Challenge. The Kaggle competition, organized by Facebook with industry and academic partners, to build automatic deepfake detectors. Its dataset is still a standard for training them.

19. Detective (Deepware). Paste a video link and it scans for face manipulation.

20. FakeApp / Faceswap. The deepfake software that started it all. It's on this list because detectors are trained against it.

What these 20 add up to

The pattern holds across formats. People do only somewhat better than a coin flip at spotting AI, and they get worse as models improve. Faces were about 70% in 2019, chat was 68% in 2023, and by 2025 people were choosing GPT-4.5 over real humans. The Turing test has stopped being a bar for machines to clear. It's turning into a skill humans have to practice.

Browse and filter every test by type (Audio, Classic, Game, Modern LLM, Social, Visual), difficulty and access in the Turing Test directory.