Goodbye, Search Box. Hello, Chat Box.

I never thought that the biggest shift in user behavior in 2025 would be our reliance on traditional search engines like Google, dropping from near-total information monopoly to almost negligible.

This isn't baseless speculation. I've observed that more and more people's primary way of getting information has shifted from "searching" to "asking AI."

To be honest, I was late to realize this. It wasn't until recently, when I watched the Collison brothers (Stripe founders) interview indie developer Peter Levels, that it hit me. Peter mentioned that his personal blog's traffic from ChatGPT jumped from 5% to 20% in just over a month. That's an astounding number.

This completely upended my assumptions. I always thought that even if users got answers through AI, they probably wouldn't bother clicking through to source links—at least, that's how I usually behave, unless I'm particularly curious about the source.

This also made me reflect on how I've become increasingly dependent on AI myself, sometimes unconsciously accepting all its suggestions. Around May this year, I wanted to replace my car's license plate. I heard that a private organization called the "Alberta Motor Association" (AMA) could get better plates, so I planned to go there. Why did I think a private organization was obligated to serve me? Because I asked AI, and it told me: this is a public service, and all licensed private departments must handle it. Naturally, when I showed up, I was turned away.

While this was just a minor incident, it reflects a massive trend. Even Google itself has had to face this reality—their recent rush to launch AI search mode is precisely because the decline in their traditional search business has reached a point they can no longer ignore.

Behind this lies the fundamental difference between traditional search and AI chat modes.

The story of OpenAI co-founder Greg Brockman is a perfect example. He was previously Stripe's CTO, a top technical expert, but to found OpenAI, he started learning AI from scratch. In an interview, he mentioned that early on when using AI, he was particularly concerned about privacy, even using incognito mode, not wanting AI or Google to know too much about his personal information.

But as AI evolved, he discovered that the more AI knew about him—for instance, after long-term use of the same ChatGPT account, AI understood his personal background, thinking patterns, and preferences—the more precise and useful AI's answers became. This is why many people now find ChatGPT's "memory feature" so excellent.

"Context"—this is the biggest divide between traditional search and AI chat.

This challenge is exactly like the dilemma we face when solving problems in daily life. When people ask questions, they often just throw out a big question without sufficient context. We can certainly try to answer, but a truly valuable answer might require us to ask ten, or even a hundred questions first, to fill in the necessary background information.

For example:

• When you ask a programming question, I need to know your programming background: Have you never written code? Learned but forgot? Or do you use it occasionally? The answers for these three situations are completely different.
• When you say "I'm having problems using AI for something," I want to ask: Which AI tool are you using? What version? Most importantly, how are the "parameters" (like reasoning efforts) configured?

To perfectly solve a problem, the context we need might run thousands of words.

With all that said, I want to offer two specific suggestions:

1. Provide "excessive" context: Whether asking AI or real people, develop the habit of providing as much background information as possible. While Google's search box couldn't handle long sentences, for today's AI, the richer the context, the more the answer will align with your actual needs.

2. Practice an "AI-first" strategy: Except for the rare scenarios where it's unsuitable, we should try making AI our first choice for problem-solving. Whether browsing information (using Perplexity Comet), writing code, or planning trips, the more you use it, the more you'll appreciate the cognitive leap AI brings.

Thank you for reading. If you have any thoughts, please feel free to leave a comment below.