This site collects what people on X say about whether AI systems are conscious, and about whether they are owed moral consideration, including arguments for or against that don't depend on consciousness. Each opinion is sorted into one of three stances:
- Affirming AI systems are, or may well be, conscious, or they are owed moral consideration.
- Skeptical AI systems are not conscious, or talk of their welfare and moral status is misplaced.
- Uncertain The question is open, unknowable for now, or badly posed, without leaning either way.
An opinion is an original tweet or a quote tweet. When its author continued it in a thread, the whole thread counts as one opinion. Replies to other people are not collected. Only human opinions are kept: tweets that just relay what a model said about itself, news without comment, and questions are left out.
The tweets come from the Community Archive, an opt-in archive of X, starting 2026-06-01. The share of each stance describes the accounts in that archive, which lean toward particular corners of AI Twitter, not public opinion.
How it works
- The archive is searched for “conscious”, “sentient”, “qualia”, “experience”, “welfare”, “moral status”, “rights”, “self-aware” and related phrases.
- Each match is read together with the tweet it quotes, the rest of its thread and its attached images. Gemini 3.8 Flash decides whether it is a human opinion on the topic, which stance it takes, and how much substance it has, from 1 to 10. Opinions scoring 4 or more appear on the site. Stances are assigned per opinion, not per person, and the grader can be wrong.
- Each opinion on the site is tagged with up to three of the arguments it makes, from a fixed list (version 1), by Gemini 3.8 Flash. The list came from two passes of Claude Opus 5.5 over all the opinions' claims; when it changes, every opinion is tagged again.
- Opinions are ranked by
50 × quotes + 20 × log2(1 + min(retweets, 100)) + 15 × substance, where quotes counts distinct Community Archive accounts that quote-tweeted the opinion. - Claude Opus 5.5 reads each stance's strongest opinions, at most four per account, and writes the summary on its page.
Credits
Built on Model Behavior Reports by Alexandre Variengien (MIT). The code is on GitHub. Data from the Community Archive.