X adds Grok account bios, but they keep hallucinating
X just rolled out a Grok-powered "More About This Account" button for paid web users, letting you hover over any profile to generate a quick AI summary. The problem? Early tests show Grok frequently confuses users with celebrities or invents hobbies out of thin air. Why it matters: Even with direct access to platform data, turning social posts into reliable user context is proving harder than expected.
Elon Musk has been turning X into an open-air testing ground for xAI's Grok model ever since it launched.
First it was chatbot interactions, then AI-generated news summaries for trending topics. Now, X has rolled out a new Grok-powered tool called More About This Account. The idea is simple: hover over an account handle or display name on the web, click a button, and get a quick AI-written dossier on who you are looking at.
It sounds practical on paper. But in practice, it is turning out to be a hallucinating mess.
Brunch bloggers and cricket stars
According to hands-on testing by TechCrunch, Grok's summaries swing wildly between generic platitudes and flat-out fiction.
When asked about tech journalists, the AI often spit out vague descriptions that could apply to virtually anyone. Worse, it labeled one reporter an expert on Canada despite having posted almost nothing about the country.
Another TechCrunch editor was somehow categorized as a brunch account. And when queried about an executive at edtech startup Unacademy, Grok confused him with a famous Indian cricket player who happens to share the same name.
X does display a small disclaimer warning that Grok 1.5 is an early feature and can make mistakes. But the real problem is that even when it does not hallucinate, it barely offers anything beyond what anyone could glean from reading a user's bio for three seconds.
Why it matters
If you are paying for X Premium, this is not a tool you should rely on to vet anyone. You are far better off doing a quick manual glance at a profile than letting an unpredictable LLM misinform you about who you are interacting with.
What makes this stumble particularly glaring is that real-time social data was supposed to be xAI's biggest competitive moat. If Grok cannot accurately summarize accounts living directly inside its own parent database, it shows how challenging basic entity disambiguation and retrieval still are.
In our view, weaving AI into micro-interactions across a social platform is a clever product direction. But social profiles represent personal reputation. Slapping fabricated identities and goofy tags onto real people makes the AI look amateurish, and shipping half-baked profile summaries risks irritating the very paying subscribers Musk wants to retain.
Sources (1)
- techcrunch.com(opens in a new tab)Originaltechcrunch.com