I searched for a laptop setting and noticed something strange about my behaviour.

The answer was already sitting at the top of the search page. I read it, changed the setting and closed the tab.

I never visited the website that supplied the information.

This is becoming normal. Search engines are moving from lists of links toward generated answers, follow-up questions and increasingly agent-like help. You can ask a messy question in ordinary language instead of reducing it to six keywords and opening eight tabs.

As a user, I like that more than I expected.

As someone who cares about the web, I also think it creates a slightly uncomfortable problem.

We are losing the little research ritual

Traditional web search involved a routine.

Search. Scan results. Open a few pages. Decide which one looks credible. Return to search because the first page spent seven paragraphs avoiding the answer. Eventually collect enough information to form a view.

AI search compresses that process. The system reads across sources and gives you a response directly, usually with links if you want them.

For straightforward questions, this is excellent. I do not miss hunting through a recipe blog for an oven temperature.

For complicated questions, convenience can hide an important step: noticing disagreement.

When five sources are shown separately, differences are visible. When a model blends them into one smooth paragraph, uncertainty can disappear into good writing.

Fluency is not consensus.

The web still has to produce the answer somewhere

AI search depends on information created by people, companies, researchers, forums and publishers.

If users increasingly read the summary without visiting the source, the economics become awkward. Websites use traffic to sell products, attract subscribers, earn advertising revenue or simply justify the cost of publishing useful information.

A search engine can make the user experience better while sending fewer visits outward.

Those two things can both be true.

I don’t think the web disappears because of this. But publishers will care more about being the source that systems trust, not merely the result that ranks first for a keyword.

Original reporting, firsthand expertise, tools and distinctive data become more valuable when generic explanation is easy to summarise.

Pages written mainly to catch search traffic may become less rewarding. I won’t mourn all of them.

Asking better questions suddenly matters

Keyword search trained us to speak badly to computers.

“best phone battery camera under 500”

We learned that full sentences were unnecessary and sometimes less effective.

AI search encourages the opposite. Describe the actual situation: “I need a phone for travel, I care more about battery and camera than gaming, and I want something that will last four years.”

That extra context can produce a much more useful starting point.

The skill is not prompt engineering in the theatrical sense. You do not need a 400-word formula. You need to include the constraints that would matter if you asked a knowledgeable person.

I think search becoming conversational will make people realise how many bad results began with vague questions.

Shopping answers deserve suspicion

The most dangerous place for an answer machine is where the answer has commercial value.

If an AI system recommends a hotel, laptop, insurance plan or pair of running shoes, I want to know why. Was the choice based on independent information? Sponsored placement? Availability? My past behaviour? A retailer partnership?

Search engines have always mixed information and advertising. AI makes the boundary feel less obvious because the recommendation arrives in the same conversational voice as everything else.

Users will need stronger instincts here.

Treat the generated answer as a shortlist, not a final purchase order. Open sources. Check current prices. Look for recent complaints. For expensive decisions, go outside the search engine’s summary.

Convenience is great until it becomes the only opinion in the room.

Personal search is useful and slightly creepy

Search is also becoming more personal.

An AI system that can connect to your mail, photos, calendar or previous activity can answer questions that ordinary web search could not. “Which hotel did I stay at in Tokyo two years ago?” is not really a web question. It is a personal data question.

That can be incredibly useful.

It also turns search into one of the most powerful windows into your private life.

I would only connect personal sources when the benefit is clear, and I would check which data is being used. The ability to disconnect should be obvious, not buried under twelve settings screens.

Personalisation usually feels harmless when it works. People begin asking questions after it surprises them.

Brands may have to write for humans again

There is an odd possibility here.

For years, companies wrote pages partly for search engines. Keywords appeared in headings. Articles answered questions nobody had genuinely asked. The first paragraph stretched because the page needed to capture more queries.

If AI systems can extract meaning more effectively, stuffing exact phrases into a page becomes less useful. Clear information, structured data, strong reputation and genuine expertise may matter more.

That is a healthier direction.

It also creates a new optimisation industry. People are already trying to understand how brands appear inside AI-generated answers. Agencies will sell audits. Someone will create a dashboard with a percentage score.

Search errors feel more confident now

A bad traditional search result looked bad. You could skip it.

A bad generated answer looks like an answer.

That difference matters. AI systems can misread sources, combine facts incorrectly or provide outdated information. Search tools are improving, but links underneath do not make the risk zero.

For medical, legal, financial or serious technical questions, I would open the source. Check dates. Prefer primary documents. If the answer affects money, safety or a serious decision, verify it somewhere else.

The easier the interface becomes, the more important this habit is.

We may browse less and discover differently

I used to find interesting websites by accident while looking for something else.

AI search is more efficient, which means fewer accidental detours. It can answer the question without sending you wandering through a forum, someone’s personal blog or an obscure university page.

Efficiency always removes some wandering.

Maybe discovery shifts elsewhere: newsletters, communities, social feeds, recommendations from people we trust, or the source links AI systems surface deliberately. Maybe search products themselves become better at showing original voices instead of only summarising them.

I hope so.

I like getting the answer quickly. I also don’t want the web reduced to a hidden database that machines read on my behalf.

Sometimes the useful part is the page you were not looking for.