{"course":{"slug":"ai-actually-explained","title":"AI, Actually Explained","url":"https://addaly.com/learn/ai-actually-explained","level":"beginner","track":"start"},"lesson":{"slug":"not-a-search-engine","title":"Why it is not a search engine, and what changes when you bolt one on","url":"https://addaly.com/learn/ai-actually-explained/not-a-search-engine","position":30,"of":92,"minutes":9,"answer":"A model generates rather than retrieves, so citations from memory are fabricated shapes, and even with search bolted on, a link proves the page was used, not that it says what the answer claims.","keyIdea":"A model generates rather than retrieves, so citations from memory are fabricated shapes, and even with search bolted on, a link proves the page was used, not that it says what the answer claims."},"body":{"format":"markdown","text":"## Two entirely different machines\n\nA search engine keeps an **index**: a vast list of documents, and for each, where its words appear. You send a query, it finds matching documents, it returns them. Whatever comes back existed before you asked. It can be wrong about relevance. It cannot invent a page.\n\nA language model keeps **weights**. You send text, it generates text. Nothing it returns existed before you asked. It cannot look anything up, because there is nothing to look in.\n\nThat is the whole difference, and nearly every disappointment people have with chatbots traces back to expecting the first while using the second.\n\n## Why citations from memory are unsafe\n\nA reference has a recognisable shape: authors, year, title, journal, volume, pages, a DOI beginning `10.`. The model has read hundreds of thousands of them and has learned that shape thoroughly.\n\nGenerating a well-formed citation therefore requires no knowledge of any actual paper. Plausible authors for the topic, a plausible year, a plausible journal, plausible page numbers. The output is indistinguishable in form from a real one, and it may be entirely fictional.\n\nThis is not a rare edge case. It reached court: in 2023 two New York lawyers were sanctioned for filing a brief containing six fabricated case citations produced by a chatbot, complete with quotations from the non-existent rulings. Comparable incidents have followed in several countries since, including from people who knew about the first one.\n\nThe rule is unconditional. **A citation from a model with no search is a hypothesis about a document, not a document.** Check that it exists before repeating it.\n\n## What \"with web access\" actually does\n\nWhen a product searches, something specific happens, and it is worth knowing the steps:\n\n1. Your question is turned into one or more search queries.\n2. A real search engine runs them and returns real pages.\n3. Some of that page text is pasted into the context window.\n4. The model answers from the text now sitting in front of it, and links to the sources.\n\nThis is **retrieval-augmented generation**, RAG, and it is the same recipe whether the corpus is the whole web, your company's documents, or a folder of your own notes. The retrieval step is often the embedding search from an earlier lesson, sometimes ordinary keyword search, usually both.\n\nWhat it fixes is real. The answer is now grounded in text that exists, the cut-off is bypassed, and you get links you can check.\n\n## What it does not fix\n\nThree things, and each has caused public failures.\n\n**Retrieval can bring back the wrong thing.** If the search returns an outdated page, a satirical post, a forum joke or a competitor's marketing, the model answers faithfully from it. Grounding guarantees a source, not a good one. The widely reported episode in 2024 of an AI search feature recommending glue on pizza came from faithfully summarising a joke comment.\n\n**The model can still misread.** It may overstate a hedged finding, merge two sources into a claim neither made, or attribute a sentence to the wrong document. A link next to a sentence means the system used that page. It does not mean the page says that.\n\n**The link may not support the claim.** This is the one to check, and it takes ten seconds. Open the source, search it for the specific figure or claim. Studies of AI search summaries have repeatedly found meaningful rates of citations that do not support the surrounding sentence.\n\n## The practical test\n\nWhen an answer matters, ask two questions.\n\n**Did it actually search?** Look for links, or for a visible indication that a search ran. If neither is present, the answer came from weights and any citation in it is unverified. Some products search only sometimes, deciding per question, and do not always make that obvious.\n\n**Does the source say it?** Open one link (the one carrying the claim you would act on) and find the sentence. If the claim is not there, nothing else in the answer has earned your trust either.\n\nThat second habit is worth more than any amount of prompt technique. It converts a fluent paragraph into either a checked fact or a discarded one, in the time it takes to make tea."},"check":{"question":"An AI search tool returns a confident answer with three source links. A user checks the first link and cannot find the claim anywhere on the page. What is the best interpretation?","options":[{"text":"Retrieval worked and generation did not stay faithful to it, so the remaining claims need checking too rather than being assumed sound.","correct":true},{"text":"The link is a formatting error, and the correct source is one of the other two.","correct":false,"tempting":"Assumes a display bug; the more common cause is a claim generated beyond what the retrieved text supports."},{"text":"The page has been updated since it was retrieved, so the claim was accurate when the tool read it.","correct":false,"tempting":"Possible occasionally, but reaching for it first excuses the system rather than testing it, and the remaining claims are still unverified."},{"text":"The model retrieved from a cached copy, so checking the live page proves nothing.","correct":false,"tempting":"Uses caching to make the claim unfalsifiable, which removes the only practical check available to the reader."}]},"prev":{"slug":"the-invisible-instructions","title":"The text you never see","url":"https://addaly.com/learn/ai-actually-explained/the-invisible-instructions"},"next":{"slug":"confidently-wrong","title":"Why it can be wrong and sound completely sure","url":"https://addaly.com/learn/ai-actually-explained/confidently-wrong"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Why it is not a search engine, and what changes when you bolt one on', in AI, Actually Explained. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-actually-explained/not-a-search-engine (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Why it is not a search engine, and what changes when you bolt one on","courseTitle":"AI, Actually Explained","module":{"index":3,"slug":"how-a-chatbot-works","title":"What a chatbot is actually doing"},"answer":null,"keyIdea":"A model generates rather than retrieves, so citations from memory are fabricated shapes, and even with search bolted on, a link proves the page was used, not that it says what the answer claims.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"two-entirely-different-machines","runs":[{"text":"Two entirely different machines"}]},{"t":"p","runs":[{"text":"A search engine keeps an "},{"text":"index","b":true},{"text":": a vast list of documents, and for each, where its words appear. You send a query, it finds matching documents, it returns them. Whatever comes back existed before you asked. It can be wrong about relevance. It cannot invent a page."}]},{"t":"p","runs":[{"text":"A language model keeps "},{"text":"weights","b":true},{"text":". You send text, it generates text. Nothing it returns existed before you asked. It cannot look anything up, because there is nothing to look in."}]},{"t":"p","runs":[{"text":"That is the whole difference, and nearly every disappointment people have with chatbots traces back to expecting the first while using the second."}]},{"t":"h2","id":"why-citations-from-memory-are-unsafe","runs":[{"text":"Why citations from memory are unsafe"}]},{"t":"p","runs":[{"text":"A reference has a recognisable shape: authors, year, title, journal, volume, pages, a DOI beginning "},{"text":"10.","code":true},{"text":". The model has read hundreds of thousands of them and has learned that shape thoroughly."}]},{"t":"p","runs":[{"text":"Generating a well-formed citation therefore requires no knowledge of any actual paper. Plausible authors for the topic, a plausible year, a plausible journal, plausible page numbers. The output is indistinguishable in form from a real one, and it may be entirely fictional."}]},{"t":"p","runs":[{"text":"This is not a rare edge case. It reached court: in 2023 two New York lawyers were sanctioned for filing a brief containing six fabricated case citations produced by a chatbot, complete with quotations from the non-existent rulings. Comparable incidents have followed in several countries since, including from people who knew about the first one."}]},{"t":"p","runs":[{"text":"The rule is unconditional. "},{"text":"A citation from a model with no search is a hypothesis about a document, not a document.","b":true},{"text":" Check that it exists before repeating it."}]},{"t":"h2","id":"what-with-web-access-actually-does","runs":[{"text":"What “with web access” actually does"}]},{"t":"p","runs":[{"text":"When a product searches, something specific happens, and it is worth knowing the steps:"}]},{"t":"ol","start":1,"items":[[{"text":"Your question is turned into one or more search queries."}],[{"text":"A real search engine runs them and returns real pages."}],[{"text":"Some of that page text is pasted into the context window."}],[{"text":"The model answers from the text now sitting in front of it, and links to the sources."}]]},{"t":"p","runs":[{"text":"This is "},{"text":"retrieval-augmented generation","b":true},{"text":", RAG, and it is the same recipe whether the corpus is the whole web, your company’s documents, or a folder of your own notes. The retrieval step is often the embedding search from an earlier lesson, sometimes ordinary keyword search, usually both."}]},{"t":"p","runs":[{"text":"What it fixes is real. The answer is now grounded in text that exists, the cut-off is bypassed, and you get links you can check."}]},{"t":"h2","id":"what-it-does-not-fix","runs":[{"text":"What it does not fix"}]},{"t":"p","runs":[{"text":"Three things, and each has caused public failures."}]},{"t":"p","runs":[{"text":"Retrieval can bring back the wrong thing.","b":true},{"text":" If the search returns an outdated page, a satirical post, a forum joke or a competitor’s marketing, the model answers faithfully from it. Grounding guarantees a source, not a good one. The widely reported episode in 2024 of an AI search feature recommending glue on pizza came from faithfully summarising a joke comment."}]},{"t":"p","runs":[{"text":"The model can still misread.","b":true},{"text":" It may overstate a hedged finding, merge two sources into a claim neither made, or attribute a sentence to the wrong document. A link next to a sentence means the system used that page. It does not mean the page says that."}]},{"t":"p","runs":[{"text":"The link may not support the claim.","b":true},{"text":" This is the one to check, and it takes ten seconds. Open the source, search it for the specific figure or claim. Studies of AI search summaries have repeatedly found meaningful rates of citations that do not support the surrounding sentence."}]},{"t":"h2","id":"the-practical-test","runs":[{"text":"The practical test"}]},{"t":"p","runs":[{"text":"When an answer matters, ask two questions."}]},{"t":"p","runs":[{"text":"Did it actually search?","b":true},{"text":" Look for links, or for a visible indication that a search ran. If neither is present, the answer came from weights and any citation in it is unverified. Some products search only sometimes, deciding per question, and do not always make that obvious."}]},{"t":"p","runs":[{"text":"Does the source say it?","b":true},{"text":" Open one link (the one carrying the claim you would act on) and find the sentence. If the claim is not there, nothing else in the answer has earned your trust either."}]},{"t":"p","runs":[{"text":"That second habit is worth more than any amount of prompt technique. It converts a fluent paragraph into either a checked fact or a discarded one, in the time it takes to make tea."}]}],"check":{"question":[{"text":"An AI search tool returns a confident answer with three source links. A user checks the first link and cannot find the claim anywhere on the page. What is the best interpretation?"}],"options":[{"text":[{"text":"Retrieval worked and generation did not stay faithful to it, so the remaining claims need checking too rather than being assumed sound."}],"correct":true,"tempting":null},{"text":[{"text":"The link is a formatting error, and the correct source is one of the other two."}],"correct":false,"tempting":[{"text":"Assumes a display bug; the more common cause is a claim generated beyond what the retrieved text supports."}]},{"text":[{"text":"The page has been updated since it was retrieved, so the claim was accurate when the tool read it."}],"correct":false,"tempting":[{"text":"Possible occasionally, but reaching for it first excuses the system rather than testing it, and the remaining claims are still unverified."}]},{"text":[{"text":"The model retrieved from a cached copy, so checking the live page proves nothing."}],"correct":false,"tempting":[{"text":"Uses caching to make the claim unfalsifiable, which removes the only practical check available to the reader."}]}],"order":[2,1,0,3],"why":null},"prev":{"slug":"the-invisible-instructions","title":"The text you never see","minutes":8},"next":{"slug":"confidently-wrong","title":"Why it can be wrong and sound completely sure","minutes":7}}}