@Grok is this true: How X’s chatbot performs as a fact-checking tool

X users love to ask Grok for fact checks. Search for “@grok is this true?” on the platform and you will see that multiple such requests are made every minute.

In fact, this query was the single most common message sent to the AI chatbot in the six months following Grok’s integration into X, according to a new working paper by Thomas Renault, Mohsen Mosleh, and David Rand. The researchers found that 447,083 tweets tagged the bot to request fact checks of other posts between March and September of 2025. Including other ways users asked for verification help, the researchers estimate that almost 1.4 million fact-checking requests were made on X to Grok and Perplexity, another AI tool. That represents 7.6% of all interactions with the two chatbots.

Even if you account for the fact that some requests are meant in jest, there appears to be a large appetite for automated fact-checking on Musk’s social platform.

This surge in AI fact-checking might ultimately affect the popularity of Community Notes, according to a different working paper by Yingxin Zhou and Jingbo Hou. Zhou and Hou used data from the three months before the introduction of Grok to build a hypothetical trendline for user participation in X’s crowdsourced fact-checking program if the chatbot had never been introduced. They found that note requests, contributors, notes written, and note ratings all declined in a statistically significant manner in the three months following Grok’s launch.

The paper from Renault and colleagues also acknowledged “a substantial reduction in the number of Community Notes being proposed” following Grok’s introduction. While their paper “does not provide evidence of a causal effect,” the authors speculate that “LLMs may be more of a substitute, rather than complement, for Community Notes.”

As my reporting has shown, Community Notes is already in a state of gentle decline. This has been somewhat camouflaged by the introduction of AI contributors in September of 2025. Automated writers have gradually increased their share of all “helpful” notes appended to posts on X from an average of 11% in December to approximately 18% in the first few weeks of January (see chart below).

AI authors in Community Notes may sap the incentive structure for human contributors. But Grok could render the entire program redundant.

What would be gained, and what lost, if that were to happen? I combine the findings from Renault, Rand, and Mosleh with a qualitative review of Grok replies and Community Notes posted in January² to sketch out a tentative answer.