What's the difference between an LLM and a search engine?

A search engine finds and ranks web pages that already exist; an LLM generates a brand-new answer in its own words. Search engines like Google crawl the web, index pages, and rank them by relevance — when you search, you get links to pages other people wrote, and you can click through to check the source.

How does each one actually produce an answer?

A search engine is like a librarian: it doesn't write anything itself, it points you to existing material. It continuously crawls the web, builds an index of what's out there, and when you type a query it ranks the most relevant pages and hands you the links. An LLM like ChatGPT, Claude, or Gemini works the opposite way.

It doesn't retrieve pages at all — it was trained on huge amounts of text and learned patterns well enough to predict, word by word, what a good answer looks like, then writes that answer fresh with nothing to click through to.

Which one can you actually trust and verify?

This is the real practical difference. A search result is traceable: you can open the original page, see who wrote it and when, and judge the source for yourself. A plain LLM answer is synthesized from patterns, so it can be outdated, oversimplified, or flat-out wrong — a mistake known as a hallucination — and there's no source link baked in unless something adds one.

Search makes you do the reading; an LLM does the reading for you but asks you to take its word for it.

Can one tool do both?

Increasingly, yes. ChatGPT's and Gemini's search modes let the model run a live web search first, then write an answer using what it finds — so you get generation plus citations in one place. But under the hood, search (retrieve and rank existing pages) and generation (write new text from learned patterns) are still two different jobs, even when a single product stitches them together.