Thanks @Austin! It was very exciting to see the coverage ourselves, and very happy to give Manifund its first citation!

Hi all,

Thank you again so much for your support! After completing the exploratory phase which was funded by this grant we completed two experimental studies based on the results.

We evaluated two prompts on 4 different LLMs and found that GPT-4.5 (with one prompt that we had validated through the exploratory phase) was judged to be human significantly more often than real humans (73% of the time). We believe this to be the first evidence that any system passes a standard (3-party) 5-minute Turing test.

You can read the rest of the results in our preprint:

Jones, C. R., & Bergen, B. K. (2025). Large Language Models Pass the Turing Test. arXiv preprint arXiv:2503.23674. (https://arxiv.org/abs/2503.23674)

The site is still live and will continue to run until we spend the rest of the money from this grant/project.

This project would most likely have languished without your support, so thanks very much for keeping it going and helping us to get to this stage! I will probably close the project soon after the funding runs out. Please let me know if you have any questions or ideas for further experiments to run.

Cameron

@Austin Thanks so much!

This is a great point and something we went back and forth about a lot. I am going to post in more places today and hopefully we'll see a bit more traffic. If we are seeing consistently high traffic in those windows we will extend the times where it's playable.

But because you need other people to be online while it's being played, and currently we're not seeing very high traffic even in those periods, then hopefully this format at least maximises the chance that people can play every day.

I like the idea of adding a share button to the homepage. There is currently one after you complete a game, but it could be good to incentivise people to share so that they can play straight away.

On the point of playing vs a friend, this might undermine the Turing test to some extent. I think if you know that the other person is your friend you have a lot of insider knowledge that would allow you to beat even a very good LLM agent (or another human)!

Thanks very much for your support and these suggestions!

The site is finally back up at turingtest.live.

The new site uses a 3-party format where you chat with both a human and an LLM at the same time, and your task is to decide which is which. This setup is closer to Turing's original idea, and we think it will be much harder for the models to pass.

We've also added a range of models, including GPT-4o, Claude, and LLaMA, along with new prompting techniques that allow the models to interact in different ways.

The site will be live every day from 12–1 PM and 8–9 PM GMT (8 AM and 3 PM ET, 5 AM and 12 PM PT). We've done this in the hope of increasing the density of users online at the same time.

Thanks so much to everyone for your help, patience, and support while getting this back going. Please let me know if you have any comments or thoughts on the site as I'll be continuing to make updates. And please feel free to share the site as it works best when we get enough traffic for many people to be live at the same time.

We’re planning to relaunch the site soon and we’re running a pilot test on Friday at 8pm GMT (3pm ET, 12pm PT).

You can access the new site here: 3p.turingtest.live. The pilot will be accessible here on Friday.

Any feedback you can provide about how the site and your interactions work would be greatly appreciated!

If you have any questions or comments, please let me know at cameron@ucsd.edu.

Hi all,

Thank you for your patience and apologies for the lack of updates. I have had to focus on other things over the last months including finishing my PhD and starting a postdoc. However, I am now able to put my full focus on this for the next couple of months. I'm hoping to have the updated experimental design finished by the end of the month and to start collecting data in October.

Thanks again and please let me know if you have any questions!
Cameron

@dominic Thanks very much, Dominic! I'm glad you had a chance to try it out and I appreciate the support!

@Alyssa Thanks so much, Alyssa!

@AntonMakiievskyi Thanks so much Anton! I really appreciate the support.

1. Participants are randomly assigned to be witnesses or interrogators. The lack of humans online is a definite issue, as there were periods where a player would be repeatedly matched with AI if no humans were online. I'm considering only making the game available for ~1hr a day to maximise the density of humans online while keeping costs down.
2. I have applied for OpenAI but didn't hear back. Will try Claude & OpenAI again.
3. This could be a good backstop if we run out of credits again. I'm a little nervous about handling the data but I'm sure there's a secure way to do this.

Yes, I will probably take a couple of weeks to make changes to the site and then hopefully update just before we relaunch. Thanks again and let me know if you have more questions or would like to chat more.

@Austin Thanks very much for your support & thoughtful comments, Austin!