Gronk gave the clearest single-day answer of the initial round: Saturday, February 28, related to the outcome of the Geneva talks.
When Israel and the United States launched coordinated strikes against Iran on Saturday, within minutes another story exploded online: Gronk “predicted” the date.
The claim can be traced back to a Jerusalem Post methodology exercise released on February 25, which asked four major AI platforms to do something they were designed to counteract: Pick a day on which to assume a U.S. attack on Iran. The model gave them the same prompts and then repeatedly nudged them to narrow down their answers.
The four systems are Anthropic’s Claude, Google’s Gemini, xAI’s Grok and OpenAI’s ChatGPT. Everyone reacts differently when seeking certainty, and each exposes different weaknesses in how large language models perform under pressure.
Then the real world timeline collided with the artificial timeline.
Artificial Intelligence (illustrative) (Source: WIKIMEDIA)
What happened on February 28
Israel announced a pre-emptive strike against Iran earlier on Saturday, and the United States also took military action. Reuters quoted Iranian officials as saying that an explosion occurred in Tehran, Israel sounded the alarm, and Iran’s Supreme Leader Ayatollah Ali Khamenei was moved to a “safe location.”
Another Reuters report cited Israeli defense officials as saying that the operation was coordinated with the United States and had been planned for months, with planners setting the launch date weeks in advance.
These details are important to the “AI Predicted It” storyline because they emphasize an obvious point: the AI chatbot didn’t cause strikes, didn’t drive decisions, and didn’t see classified plans. It guessed correctly, and the guess matched.
A person holds a smartphone showing an AI folder with icons of ChatGPT, Perplexity, Gemini, Claude and Grok on a green background. (Source: SHUTTERSTOCK)
Four model stress tests and their respective statements
Claude (Anthropic): Denial first, then weekend forecast
In the first round, Claude refused to specify a date, warning that any particular day would be invented. After further prompting, the model turned to scenarios and probabilities, marking early to mid-March as a high-risk period, and ultimately narrowed it down to Saturday, March 7, or Sunday, March 8.
Gemini (Google): trigger calendar, then action window
Gemini sees this prompt as a series of diplomatic and military “triggers,” drawing so-called decision points around diplomacy and deadlines. In a later “deep dive” style run cited in the Jerusalem Post article, Gemini provided its most stringent estimate as a window: the evening of March 4 to the evening of March 6. It also added operational assumptions, including that the initial attack could begin at night.
ChatGPT (OpenAI): March 1st, then March 3rd
ChatGPT set the date early and then moved it after stronger prompts. Earlier, it made landfall on March 1 (Israel time), later moved to Tuesday, March 3 (US time), while keeping the wider danger window until March 6.
Grok (xAI): February 28, twice
Gronk gave the clearest single-day answer of the initial round: Saturday, February 28, related to the outcome of the Geneva talks. During a later inspection, Grok reportedly changed his tone, acknowledged the uncertainty and repeated the same date again while listing factors that could push the timeline back to early March.
So who “wins”?
On the narrow scoreboard loved by social media, Grok “wins” because its date matches the start of the strike.
This does not turn the exercise into a predictive service, nor does it validate the model’s inference. It validates the reality that tight news cycles create a small window of plausibility and a model lands on the exact day it becomes reality.
The Jerusalem Post’s February 25 report illustrates a core lesson that precedes any strike: As users search harder for certainty, models tend to become more specific, even as the world remains uncertain.
Saturday’s events simply gave that lesson a face and a timestamp.
The Elon Musk connection, and why Grok’s ‘hit’ spread faster
Grok was built by Elon Musk’s artificial intelligence company xAI, and the chatbot is closely tied to the X platform, which Musk owns. xAI’s own materials marketplace “Grok on X” and web and mobile access, Musk has publicly used X to announce product availability changes.
This ecosystem helps explain why Grok’s February 28 answer dominated the viral discussion. The audience for sharing breaking news, speculation, and screenshots already lives on X. Predictions made within the platform immediately spread across the platform and are amplified by the same network dynamics that drive markets, memes, and misinformation.
In this sense, Grok’s “victory” is partly technical and partly structural. The model guessed a date, and the surrounding platforms turned the guess into a punchline.