
I noticed it on a Tuesday. I had a question, my hands moved on their own, and they didn't go to the search bar. They went to an AI assistant. Then I realized this had been true for weeks.
For most of my life, "look it up" meant Google. Now it means ask. And the gap between those two things turns out to be bigger than I thought.
This isn't a "Google is dead" piece. Google is very much alive and I still need it. But the default moved, and that shift changed how I think more than any app I've installed in years.
I reach for an AI assistant before search whenever the job is to understand, draft, or decide something, and I reach for Google when the job is to find a specific source or verify a fact. AI replaced search for thinking-shaped questions. Search still wins for "give me the one true link." Most of my old searches turned out to be the first kind.
When I looked at what I used to Google, a pattern jumped out. Most of my searches were never about finding a webpage. They were about getting unstuck.
"How do I phrase this awkward email." "What's the difference between these two approaches." "Explain this concept like I'm tired." "Is this idea any good." None of those have a perfect link. They have a thought I needed help finishing.
Search answers those with ten blue links and a promise that the answer is somewhere in there. An AI assistant just answers. For thinking-shaped questions, that's not a small improvement. It's a different activity.
Photo by Mariia Shalabaieva on Unsplash
Let me be specific, because vague praise is useless.
That last point is the real shift. Conversation beats query. A query is a slot machine. A conversation is a colleague.
I keep coming back to how unconscious this change was, because I think that's the real story. Nobody sent me a memo saying "stop searching, start asking." My hands just rewired themselves over a few weeks until the new default felt like it had always been there.
That's how big shifts actually happen. Not with a decision, but with a thousand tiny ones you don't notice until you look back. Pew Research has tracked how quickly conversational AI moved from novelty to everyday tool, and it matched my own experience: the habit changed before I ever decided to change it. It's the same quiet creep I described in the five AI habits that reshaped my year, and it sits underneath the honest truth about AI productivity tools. The first time I asked an AI assistant to explain something instead of searching it, it felt slightly indulgent, like cheating. The hundredth time, searching for that same kind of thing would have felt slow and old-fashioned.
What tipped it was the friction. Search makes you do work after the answer: open tabs, skim, compare, synthesize. For a "find me a link" question, that friction is fine — you wanted the link. For a "help me think" question, that friction is the worst part. You didn't want ten articles. You wanted the one thought you were missing, and you wanted to keep talking until you had it.
I think a lot of people haven't made this switch simply because they never examined what their searches were for. They assume searching and asking are the same activity with different interfaces. They aren't. One hands you raw material and makes you the assembler. The other assembles a first pass and lets you direct the edit. For the questions that fill an ordinary day — most of which are thinking-shaped, not finding-shaped — the second is just a better fit, and your hands figure that out long before your brain admits it.
Now the part that keeps me honest, because using AI more than search without this part is how people get burned.
AI assistants are confident in a way that has nothing to do with being correct. When I need a fact — a date, a number, a quote, who said what — I do not trust the first answer. I verify it. Often I go straight back to Google for exactly that.
AI is brilliant at shaping an answer and unreliable at sourcing one.
So my rule is blunt: if being wrong has a cost, I check. A draft email, no check needed. A statistic I'm about to publish, always checked. The skill isn't choosing one tool. It's knowing which tool you're holding.
Early on, I asked an AI assistant for a "fact" to drop into something I was writing. It gave me a clean, plausible, specific number. I almost used it. Something nagged, so I searched. The number was invented. Not malicious — just wrong in that smooth, certain way that's worse than an obvious error because you don't flinch.
That was the day I stopped treating AI as a search replacement and started treating it as a thinking partner that occasionally lies. Both halves of that sentence matter.
Here's the mental model I landed on, and I genuinely use it dozens of times a day.
| If I want to… | I open… |
|---|---|
| Understand an idea | AI assistant |
| Draft or rewrite something | AI assistant |
| Compare fuzzy options | AI assistant |
| Find a specific source | Search |
| Verify a number or quote | Search |
| Buy something / read reviews | Search |
| Get the latest news | Search |
Roughly, AI for thinking, search for finding. The two aren't rivals. They cover different halves of "I need to know something."
Photo by John Schnobrich on Unsplash
The single biggest upgrade wasn't the tool. It was learning to ask better.
Bad: "marketing tips." That's a search query, and the AI gives you a search-flavored mush.
Good: "I run a small newsletter, 800 readers, and growth stalled. Give me three specific things to try this week, ranked by effort." That's a brief. You get a brief-quality answer.
The people who say AI assistants are useless are, almost always, typing search queries into a thing that rewards conversation. Treat it like a smart colleague you're delegating to, and the quality jumps.
I learned this slowly, and a little reluctantly, because it meant admitting my own laziness was the problem. For months I'd fire off two-word prompts and judge the AI by the mush it handed back. The day I forced myself to write a real brief — a couple of sentences of context, the actual goal, the constraints — the output stopped being generic almost immediately. Nothing about the tool had changed. I had. The frustrating, freeing truth is that an AI assistant is a mirror for how clearly you can describe what you want, and most of us are worse at that than we think. Getting good at AI is mostly getting good at asking, and that's a skill that pays off everywhere, not just in front of a chat box.
If you want to push this further, try writing a real brief the next time you'd normally fire off a two-word query, and notice how much sharper the answer comes back.
Q: Are you saying Google is obsolete? No. I probably search less, but the searches I do run are higher-stakes — verification, sources, the latest news. Search got more important per use, even as I use it less often.
Q: Doesn't AI just make stuff up? Sometimes, yes, and confidently. That's exactly why I verify anything with a cost attached to being wrong. Used for thinking and drafting, the failure rate barely matters. Used for facts, you must check.
Q: Which AI assistant do you use? The honest answer is whichever one is in front of me. The shift isn't about a brand. It's about the behavior of asking instead of searching.
Q: Isn't conversational search just Google adding AI too? It is, and that blurs the line further. The point stands: the interaction changed from querying to conversing, whoever ships it.
I didn't decide to use an AI assistant more than search. I noticed, after the fact, that I already did. The default moved while I wasn't looking, because for the questions I actually have all day, a conversation beats a list of links.
Search finds you the answer. An AI assistant helps you finish the question.
Pull up your own search history sometime. Count how many were really "find me a link" versus "help me think." That ratio is the whole story.
I spent years saving the hardest task for when I 'felt ready.' Doing it first instead quietly fixed my focus, my dread, and my output.

I tracked every distraction for a week and was horrified by what I found. Then I fixed the three that mattered most.

I went from 200 to 11,000 subscribers without hiring anyone. AI didn't write my newsletter — it did everything around it.

Comments
Sign in to join the conversation
No comments yet. Be the first to share your thoughts!