4 AI chatbots tried to fact-check Rubio on Iran. They couldn’t agree

A simple question about Trump’s Iran war goals produced four different answers from Grok, Claude, Gemini, and ChatGPT.

[Photo: Brendan Smialowski/Pool/AFP via Getty Images]

On Monday, March 30, Secretary of State Marco Rubio told Good Morning America’s George Stephanopoulos that the U.S. objectives in the Iran war are exactly the same as those laid out by President Donald Trump on February 28, when the U.S. and Israel launched airstrikes across the country. The consistency of the objectives over the past month of fighting is a sensitive issue; the White House doesn’t want to appear to have whittled down its strategic goals so that it can depart the conflict sooner.

“This operation is about a set of very specific objectives; the president laid them out on the very first night of operations,” Rubio told Stephanopoulos. “I’ll repeat them to you now because I hear a lot of talk about we don’t know what the clear objectives are.” Rubio then listed four objectives: 1. The destruction of Iran’s air force. 2. The destruction of their navy 3. The severe diminishing of their missile launching capability 4. The destruction of their factories.

We asked four chatbots—xAI’s Grok, Anthropic’s Claude, Google’s Gemini, and OpenAI’s ChatGPT—whether Rubio’s statement was true. Their answers reveal how competing AI models can have very different ways of interpreting political news.

We first asked Grok, X’s built-in AI chatbot, to fact-check whether Rubio’s objectives indeed match those laid out by Trump in February. It replied: 

“Yes, Secretary Rubio confirmed today that President Trump laid out these exact four objectives on the first night of the operation (with preventing a nuclear Iran as the overarching goal). We’re on pace or ahead of schedule.”

“Yes, Secretary Rubio confirmed today that President Trump laid out these exact four objectives on the first night of the operation (with preventing a nuclear Iran as the overarching goal). We’re on pace or ahead of schedule.”

Instead of fact-checking the consistency of the two objectives lists, Grok simply confirmed that Rubio said what he said to Stephanopoulos Monday morning. We then asked Grok to address the real question by comparing the actual transcript of Rubio’s Monday claim against the transcript of Trump’s Truth Social video, but the chatbot again merely confirmed Rubio’s comments.

After several more turns, Grok finally compared the transcripts and stated that Trump had indeed mentioned every item in Rubio’s list. That was false: Trump hadn’t mentioned the objective of destroying Iran’s air force. More importantly, Grok failed to mention that Trump announced objectives that don’t appear in Rubio’s list, including regime change, complete nuclear disarmament, and the destruction of regional Iranian “proxies.”

Grok then tried to explain the discrepancies by saying that while Trump was describing “operational” objectives, Rubio was describing “military” objectives. But neither Rubio nor Trump made that distinction, only Grok. 

Artificial Intelligence

John Summit's rise from accountant to DJ

© Fast Company