AI search: a written answer instead of a list of links
AI search is when an assistant reads web pages on your behalf and writes a summarized answer, citing its sources. It saves time, as long as you keep the habit of checking.
Data as of October 22, 2026
Key figures
| Level | Beginner |
|---|
How it works, in plain terms
You ask a question in everyday language: « What time does the Met close this Sunday? ». The assistant runs one or more searches, reads the pages it finds, then writes an answer with links to its sources. Technically, it’s an application of RAG: the model retrieves documents before answering, which ties its response to real text (this is called grounding).
You’ll find this technology in several forms: « answer engines » such as Perplexity, search built into assistants like ChatGPT or Gemini, and AI-generated summaries shown at the top of some search engines’ results. The AI assistants comparison shows which ones can search the web.
Why it’s useful
A language model on its own only knows the world up to its knowledge cutoff. For news, prices, opening hours or a recent law, it may answer with outdated information. Web search fills that gap by fetching today’s information.
Concrete examples:
- Everyday life: checking an airline’s cancellation policy, comparing two vacuum cleaners, finding a recipe for what’s left in the fridge.
- At work: getting a first overview of a market, tracking down the text of a standard, summarizing a competitor’s recent announcements.
Pitfalls to know about
- The AI can get the summary wrong. Even with good sources, it may mix up two figures or draw a conclusion the page doesn’t support. That’s a form of hallucination.
- Source selection isn’t neutral. The tool keeps a handful of pages out of thousands; a page that ranks well isn’t necessarily the most reliable.
- Sensitive topics call for extra care. For health, legal or financial questions, treat the answer as a starting point, then check an official source or a professional.
- Your questions are sent to the provider. Avoid including personal or confidential details.
The right habit
Read the answer, then click at least one source. If the information isn’t there, or not in the same way, trust the source over the summary.
Related terms
- RAG (retrieval-augmented generation) · A technique that first retrieves relevant passages from your documents, then hands them to the model so its answer is based on them.
- Grounding · Making an AI base its answer on supplied sources (web results, documents, databases) rather than only on what it memorized in training.
- Hallucination · An AI answer that sounds confident and plausible but is false or made up: a figure, a quote, a source or a reference.
- Knowledge cutoff · The date after which an AI model received no more training data, so it knows nothing about later events.
- AI browser (agentic browser) · A web browser with a built-in AI assistant that summarizes pages and can sometimes act for you: clicking, filling in forms, booking.
Try it
Full glossary
Frequently asked questions
How is it different from a regular search engine?
A regular search engine gives you a list of links to open yourself. AI search reads several pages and writes the answer for you. You save time, but you rely on the sources the tool picked.
Are the cited sources always reliable?
No. The AI may cite a low-quality site or an outdated page, or summarize a source incorrectly. For important decisions, open at least one or two of the sources.
Why doesn't the assistant always search the web?
Depending on the tool and settings, the assistant decides on its own whether it needs to search, or waits for you to turn search on. Without search, it answers from its training knowledge alone, which stops at a certain date.
Can I ask it to cite its sources?
Yes, and it's a good habit. Explicitly ask for links to the pages it used, then click through to confirm the information is really there.
Data sources
Data as of October 22, 2026
Report an error
Does a figure, a date or a fact look wrong? Tell us: every report is checked.