The day a newspaper said Google had hit the alarm over ChatGPT

ChatGPT had been public for exactly three weeks. On December 21, 2022, The New York Times ran a story saying that Google, the company most people mean when they say "look it up," had declared a "code red" over it.

That's a strange sentence to read about the biggest search company on earth. But it was the moment a lot of people outside tech realized the chatbot they'd been poking at for fun was being taken seriously by the companies that run the internet.

What the report said

I couldn't open the Times piece itself, so this comes from outlets that summarized it within a day or two. According to those write-ups, the Times had seen internal memos and an audio recording. Sundar Pichai, Google's chief executive, led meetings on the company's AI strategy and pointed several groups at the threat to search. Teams including Google Research and Trust and Safety were asked to help bring AI products forward, and some employees were told to work on tools that make pictures from text, the way OpenAI's DALL-E does.

The summaries also explain why Google hadn't already done it. Digital ads bring in more than 80% of Google's revenue (Aaj Tak put it at about 81% of Alphabet's 2021 total). A chatbot that just hands you the answer leaves fewer reasons to click an ad. And there was the accuracy problem: these systems can say something false in a perfectly calm voice, which is a risk for a brand people trust to be right.

Slashdot's post on the story picked out the line that Google could face the first serious threat to its main search business. Read that again with three weeks of ChatGPT in mind. That's fast.

The part that still matters to you

Forget the corporate drama for a second. The worry in that report, the confident wrong answer, is the same thing that trips up regular people every day. The tools are better now. They still don't flag their own guesses unless you ask.

So ask. Most people never do, and I think it's the single most underused move in prompting. A model will happily tell you which parts of its answer are shaky if you give it permission to say so. Left alone, it writes everything in the same even tone, and the made-up date sits right next to the real one looking exactly as solid.

Here's what that looks like in practice.