Grok Showed Up in Three Federal Documents. All Three Were About Writing Emails.

Here's a useful stress test for any technology: see if the people with actual budgets and actual accountability will touch it.

Reuters went through more than 400 examples of documented US government AI use — cases where specific vendors were named — and found Grok or xAI in exactly three of them. Not three categories. Three instances. And in each case, the application was document drafting or social media management. The kind of work you'd give an intern on their first day.

A writer at The Verge put it plainly: Grok is not very good, and not many people are using it. That's not a hot take. That's a Reuters review doing the math out loud.

The Gap Between the Story and the Product

What makes this interesting isn't the numbers — it's the timing. Elon Musk is reportedly positioning xAI for what could be one of the largest IPOs in history, and Grok is supposed to be at the center of that story. The pitch is a truth-seeking AI, unshackled, honest in ways its competitors supposedly aren't. That's a compelling narrative. Narratives, however, don't show up in procurement records.

Government AI adoption isn't glamorous, but it is a real signal. These are institutions that move slowly, document everything, and evaluate tools under actual scrutiny. When something is useful, it appears in the logs. When something is mostly vibes, it doesn't. Grok doesn't.

Three entries. Email drafting. Social media posts.

The contrast with the IPO framing is almost elegant in how uncomfortable it is. You can't simultaneously be the most important AI product in the world and the one nobody with a budget and a responsibility to justify that budget will deploy at scale.

What Vanity Projects Look Like From the Outside

There's a specific kind of tech product that exists primarily to make its creator seem relevant in a space they entered late. It gets announced loudly. It gets benchmarked against competitors in ways the creator controls. It gets integrated into platforms the creator also owns, which inflates usage numbers without really inflating usefulness. And it gets called something like a "truth-seeking chatbot" because the branding does work the product cannot.

I've watched this cycle enough times to recognize the shape of it. The tell is always the same: the story about the product is more developed than the product itself.

The Reuters review didn't editorialize. It just counted. Grok appeared three times out of more than four hundred. That's not a market-share problem or a sales-execution problem. That's a product that the people best positioned to evaluate it have quietly looked at and moved on from.

The Verge staking out this take now — while the IPO narrative is still being constructed — matters. Because once the offering is priced, the story calcifies. The valuation becomes the argument. The money becomes the proof.

Three email-drafting contracts won't make the prospectus cover. But they'll still be in the records.