A search engine gives you sources and lets you draw your own conclusions. An AI gives you a conclusion and leaves you to guess at the sources. This isn't a matter of convenience: these are two tools answering two different needs, and most disappointments stem from conflating them.
Many people have switched entirely from one tool to the other. In either direction, that's a mistake. Here's a breakdown that works.
What AI does better
Questions that have no dedicated page. "Explain the difference between these two concepts in my specific context" has no answer on any website. You'd have to read five pages and synthesise them. That's exactly the job AI does well.
Adapted rewording. Getting the same explanation in simpler terms, more technical terms, or with an analogy. A web page is written once for an average reader.
Reasoning over your own data. Analysing a document you provide, sorting your notes, comparing options against your criteria. No search engine can do this, since the information isn't public.
Poorly phrased questions. When you don't know the field's vocabulary, a search engine finds nothing because you're not using the right terms. An AI understands the intent and gives you the vocabulary, which then lets you search properly.
What the search engine does better
Anything current. Prices, opening hours, availability, news, the status of a person or company. The knowledge cutoff date makes AI structurally unreliable on this, even with built-in search.
Precise, verifiable facts. A date, an official figure, a piece of legislation. You want the source, not a rephrasing of the source.
What requires a plurality of viewpoints. On a contested topic, an AI produces a synthesis that smooths over disagreements. A search engine shows you that three camps are opposed, which is sometimes the most important piece of information.
What needs to be defensible. If you have to cite, justify, or prove something, you need a document that someone else can consult.
The most effective practice is to chain them rather than choose. Use AI to understand the topic and identify the right terms, then the search engine to verify and source. You get the former's speed of understanding and the latter's traceability. This works particularly well on technical or legal topics where you don't know where to start.
The trap of integrated answers
The distinction has blurred since search engines started displaying a generated summary above the links, as we described regarding AI Overviews.
The risk is real: you think you're consulting a search engine and you're reading an AI. The displayed summary is a probabilistic output, with the same limitations as any generated response. It cites sources, which is progress, but nothing guarantees that what it draws from them matches what they actually say.
The useful reflex is simple: when the stakes matter, scroll past the box and open at least one source. That's thirty seconds, and it's the difference between information and the impression of information.
What to remember
These two tools aren't competitors but complements, and the question isn't which one wins but which one fits what you're looking for.
A simple rule to close: if you need to understand, start with AI. If you need to know, end with a source. Most serious questions require both, in that order.