Which Community Note is more helpful?

You'll see posts from X, each with two proposed Community Notes. Community Notes add context to posts that might be misleading. Your job is to say which of the two notes is more helpful.

What this involves

• 27 comparisons, about 25 minutes. There's a progress bar, and you need to reach the end to be credited.
• You won't be told which note came from where, and there is no answer key we're scoring you against.
• Some posts are a reply to or a quote of another post. When they are, that other post is shown above for context — but the notes are about the post being rated, the one at the bottom.
• Where a post has an image, it is shown with the post — judge the notes against what the image actually shows. Video is not played; you get its first frame, labelled as such.
• Check the sources if you're not sure. Click the links in the notes to see what they cite. This is encouraged: opening a link is never counted as leaving the survey.

Please don't use ChatGPT or any other AI

We're measuring what people find helpful. An answer produced by an AI assistant is worse than no answer at all, because it quietly biases the result we're trying to measure. Please answer as yourself, even when you're unsure — "I'm not certain but I'd pick this one" is exactly the data we want.

Data and quality checks

• We record your answers, how long each one took, and basic page events (clicking a source link, copying text, leaving the tab). We do not record what you type anywhere else, and we never store your name or email. Your IP address is not stored either: it is converted to a scrambled value we cannot reverse, used only to stop automated abuse, and discarded.
• A few comparisons are attention checks with an obviously off-topic note. They're there to catch clicking-through.
• If your responses don't pass these checks we may exclude them from the analysis. This does not affect your payment — if you reach the end you are credited in full either way.
• Your answers are used for research on automated fact-checking and reported only in aggregate.

Taking part is voluntary

• You can stop at any time by closing the tab. Nothing is held against you, and there is no penalty for stopping.
• Questions, or want your responses removed afterwards? Email research@example.org, quoting the id in your invitation link. We can delete your responses up to the point the dataset is analysed.

Start

Tick the box above to begin.