@grok is this true
Over on X, the move is to reply to any post with “@grok is this true?” — outsourcing reality-verification to Elon Musk’s chatbot, which will dutifully confirm whatever version of events keeps the main character looking good. It’s Ministry of Truth as a service. On Bluesky, the phrase has become shorthand for the whole absurdity: people asking a propaganda machine to fact-check propaganda.
But here’s the thing. We’ve spent years worrying about LLMs “hallucinating” — generating confident, plausible text that doesn’t correspond to reality. What if the more persistent hallucination problem is human?
Consider this NPR post from this morning:
“Trump says he will deploy ICE agents to U.S. airports to help address delays. And, the president said he would delay strikes on Iranian power plants for five days.”
Two stories. Let’s check both against what actually happened.
Story 1: The airports
What Trump actually posted on Truth Social on Saturday:
“If the Radical Left Democrats don’t immediately sign an agreement to let our Country, in particular, our Airports, be FREE and SAFE again, I will move our brilliant and patriotic ICE Agents to the Airports where they will do Security like no one has ever seen before, including the immediate arrest of all Illegal Immigrants who have come into our Country, with heavy emphasis on those from Somalia, who have totally destroyed, with the approval of a corrupt Governor, Attorney General, and Congresswoman, Ilhan Omar, the once Great State of Minnesota.”
And then, asked on camera today whether ICE would arrest immigrants at airports, Trump confirmed it outright: “Yeah. That’s why the Democrats are going crazy. ICE loves it because they’re able to now arrest illegals as they come into the country. It’s very fertile territory.”
“Very fertile territory.” He’s describing airports as hunting grounds. Not helping with delays — using the delays.
Context NPR omitted: The airport delays exist because TSA workers haven’t been paid in six weeks due to a DHS shutdown. The shutdown started after ICE agents killed two U.S. citizens during immigration raids in Minneapolis, and Democrats demanded reforms. Over 400 TSA officers have quit. Callout rates hit 55% at some airports. The president’s “solution” was an ultimatum to deploy armed immigration enforcement agents for ethnic profiling of Somalis at airports, paired with personal attacks on a congresswoman.
NPR’s version: “deploy ICE agents to U.S. airports to help address delays.”
Story 2: Iran
On Saturday, Trump gave Iran 48 hours to reopen the Strait of Hormuz or he would “hit and obliterate their various POWER PLANTS, STARTING WITH THE BIGGEST ONE FIRST!” Targeting civilian power infrastructure is prohibited under the Geneva Conventions. Iran’s UN ambassador formally notified the Security Council that such strikes would be “inherently indiscriminate.” Over 1,300 Iranian civilians have already been killed. The country has been under a total internet blackout for four weeks.
On Monday morning, Trump posted he’d “delay” these strikes for five days, claiming “productive conversations.” Iran denied any talks took place and called it a move to calm markets. Senator Chris Murphy pointed out that Trump wasn’t announcing a pause on strikes — he was “postponing a possible war crime” as “a panicky message to the markets.”
NPR’s version: “the president said he would delay strikes on Iranian power plants for five days.”
Who’s hallucinating?
An LLM hallucination generates text that is coherent but disconnected from reality. The sentences parse fine. They just don’t correspond to what happened.
Read that NPR post again. Both sentences parse perfectly. They just don’t correspond to what happened. An arsonist deploying armed agents to the fire he started becomes “helping with delays.” A president treating war crimes against civilian infrastructure as a dial he can turn up and down becomes a scheduling update.
If an LLM produced this output — plausible-sounding, syntactically clean, calmly authoritative, and fundamentally detached from reality — it would be flagged, red-teamed, and held up as evidence that AI can’t be trusted with news. The human version gets posted to 3 million followers with the NPR logo on it.
Grok hallucinates to flatter one man’s ego. NPR hallucinates to preserve the fiction that we still live in a country where normal things happen for normal reasons. Different clients, same failure mode: confident text, clean grammar, no contact with reality.
@grok is this true?