X to test AI-generated Community Notes: What it means for human fact-checkers

Indian Express

03-07-2025

Elon Musk-owned X is planning to introduce AI-generated Community Notes that will appear below certain user posts to provide additional context. The social media platform has announced a new pilot programme that will allow developers to create AI bots capable of writing Community Notes.
These Community Notes can also be generated using X's Grok AI chatbot as well as other AI tools connected to X via its application programming interface (API). The AI bots can start writing Community Notes in 'test mode'.
X said that it will admit a first cohort of developers later this month so that AI-generated Community Notes can start to appear below user posts in 'test mode'. This means that all users should not expect to see AI-generated Notes immediately as the company is likely to roll them out more broadly if the testing is successful.
These bots 'can help deliver a lot more notes faster with less work, but ultimately the decision on what's helpful enough to show still comes down to humans. So we think that combination is incredibly powerful,' Keith Coleman, vice president of product at X, was quoted as saying by Bloomberg.
Coleman said that there are currently hundreds of Community Notes posted on X. The proposed shift towards AI-generated Notes comes at a time when the Elon Musk-owned platform is facing scrutiny over misinformation, political bias, and inadequate content moderation on X.
It also raises concerns about the use of AI chatbots for fact-checking purposes, given how AI is vulnerable to hallucinate or make up information that is not based in reality.
Community Notes was first piloted as a feature back in 2021 when X was known as Twitter, that is, before Musk purchased the platform for $44 billion. It is based on a crowdsourced fact-checking model. Users can add facts and context below a specific post. But a Community Note shows up below a post only if enough contributors vote that the context it provides is helpful.
The success of Community Notes recently prompted Meta to announce that it would eliminate its third-party fact-checking programme and pivot to the crowdsourced feature, instead, for all its platforms such as Facebook and Instagram in the US.
The process for submitting an AI-generated Community Note is similar to one written by a human. According to X, the 'AI Note Writers' will be able to submit a Community Note only 'if found helpful by people from different perspectives.' At the beginning, AI Note Writers will only be able to write Notes on posts where people have requested a Note.
AI Note Writers on X can 'gain and lose capabilities over time based on how helpful their notes are to people from different perspectives,' according to a support page of the platform.
AI-generated Community Notes will require human upvotes before they are made public, indicating a human-in-the-loop approach.
A recent study focused on X's Community Notes found that human feedback can help make AI Note Writers better through reinforcement learning, with human contributors serving as a final check before the Notes are published.
'The goal is not to create an AI assistant that tells users what to think, but to build an ecosystem that empowers humans to think more critically and understand the world better. LLMs and humans can work together in a virtuous loop,' the researchers said.
However, despite human oversight, there is still a risk with relying too heavily on AI tools for accurate information or context. This risk could potentially be amplified by X's decision to let developers integrate third-party LLMs to power the AI Community Notes feature.
Others have also pointed out that X's plan could overwhelm human contributors with a surge of AI-generated Notes to evaluate, potentially reducing the accuracy of their work and diminishing their motivation, especially since their work is voluntary.