What possible ways are there to roughly detect if an answer or question has been wrote by ChatGPT or other artificial intelligence (AI) applications?

Any advice is appreciated and will go nicely with the "Should AI answers be banned?" featured discussion where some support banning and some support flagging/allowing with declarations/notices etc.


1 Answer 1


For a start, there are multiple ChatGPT detectors. From the politics.meta.se on this topic, the website AI-content-detector was provided that could scan text to see if it was produced via AI. Additionally, on another forum (who I cannot remember sadly), the website provided was named GPT-2 Output Detector Demo that supports Q's (questions) or A's (answers) being copy and pasted into the provided box to determine the likelihood of the Q or A being produced by ChatGPT.

Additionally, if you are using Google Chrome and want to install the extension Tampermonkey, then you can have a button at the bottom of every question or answer to have them evaluated for AI likelihood (using @Simon Dispa as an example as this question was at the top of the feed):

enter image description here

where the Detect OpenAI button is present you can click on it to receive a score, in this case 38.4% which is reasonable as much of TeX code has already been published.

enter image description here

Further advice on this script is contained on the Stack Apps site but it may be helpful to you if you want an automatic script to determine whether or not to raise a moderator flag.

  • 1
    It might help to analyse text and code (and maybe comments) separately, text probably flags AI likelihood much more clearly. Code especially the more likely give away I've seen seems to just be non-working answers, maybe changes to any MWE that don't contribute directly to the (supposed) answer (particularly if superfluous rather than best practise).
    – Dai Bowen
    Commented Apr 26, 2023 at 15:31
  • @DaiBowen that's a very good point, TeX code seems to raise the %age high. I do not think the Detect OpenAI can only filter with text and not code so the comments/resources in Werner's answer are probably best.
    – JamesT
    Commented Apr 26, 2023 at 17:25
  • 1
    Tampermonkey is an extension that does exist for many (desktop) browsers and only provides an interface for userscripts, it doesn't magically provide that “Detect OpenAI” link. Maybe move the link to the userscript above your screenshot and remove the restriction to Chrome. It works for me on Firefox as well (and reports 52.48% on my own top answer :|). Commented Apr 28, 2023 at 16:56

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .