{"course":{"slug":"ai-for-students","title":"AI for Students","url":"https://addaly.com/learn/ai-for-students","level":"beginner","track":"work"},"lesson":{"slug":"how-a-citation-gets-invented","title":"How a citation gets invented","url":"https://addaly.com/learn/ai-for-students/how-a-citation-gets-invented","position":39,"of":73,"minutes":8,"answer":"Citations are a rigid form that a predictor reproduces easily, so plausible fabrications and real-paper hybrids both occur: resolve the DOI to check existence, and read the page to check support.","keyIdea":"Citations are a rigid form that a predictor reproduces easily, so plausible fabrications and real-paper hybrids both occur: resolve the DOI to check existence, and read the page to check support."},"body":{"format":"markdown","text":"## Watch it happen\n\nA citation has a rigid form: authors, year, title, journal, volume, issue, pages, DOI. Rigid forms are exactly what a next-token predictor reproduces most easily. Every component is drawn from a strong pattern:\n\n- **Authors** who publish in that area, in the right initials-and-surname format.\n- **A year** in the plausible range for that literature.\n- **A title** assembled from the vocabulary of the field, often a very good title.\n- **A journal** that publishes that subject.\n- **A volume and page range** consistent with that journal's numbering.\n- **A DOI** with a real prefix, since prefixes are publisher-specific and heavily represented, and a suffix that is invented.\n\nEvery part is individually plausible. The combination has never existed. This is not a bug in the citation feature: there is no citation feature. It is the same process that writes the prose, applied to a highly structured string.\n\n## The hybrid, which is worse\n\nPurely invented references are the easy case; a search finds nothing and you delete it. The dangerous ones are hybrids:\n\n- A **real paper with a wrong year**, which breaks your reference list and looks like carelessness.\n- **Real authors, real journal, title that does not exist.**\n- A **real paper attached to a claim it does not make.** This is the worst, because every element checks out and the citation is still false. Only reading the paper catches it.\n- A **real paper whose findings are reversed**: the study found no effect, and the citation is attached to a sentence saying it found one.\n\nThe last two are why *the reference exists* is not the same as *the reference supports this*. Two separate checks.\n\n> Figure: Two separate checks, and only one of them is cheap. Does it exist? Ten seconds.: Paste the DOI after the doi.org address and press enter; Search the exact title in quotation marks in Scholar or the library catalogue; A real DOI lands on the publisher's page; an invented one errors; Check author, year, journal and volume together, not separately; Says nothing at all about whether the paper supports your sentence. Does it support the claim? Three minutes.: Get the full text, then find the page your claim rests on; A real paper attached to a claim it does not make passes every other check; So does a real paper whose finding is reversed in your sentence; Note the page number; that is what makes the citation yours; If you did not open it, you did not check it. Purely invented references are the easy case; a search finds nothing and you delete the line. The hybrids are the dangerous ones, because every element checks out and the citation is still false. The inference a supervisor draws is not that you used a model. It is that you cited something you never read, which was misconduct long before any of this existed.\n\n## Why asking the model does not help\n\n*Is this a real paper?* is a question about the world, and the model has no channel to the world. It answers from the same process that produced the citation. It will often say yes. Occasionally it will say no about a real paper, which is even more confusing.\n\nThis holds when search is switched on, in a modified form. Now it may retrieve something, and the failure becomes a real link attached to a claim the page does not make. Which brings us to the only rule that survives every version of this technology.\n\n## The rule\n\n> **Open it. If you did not open it, you did not check it.**\n\nOpen means: reach the actual paper or book, find the page, and read the sentence that supports your claim. Not the abstract if your claim is about a method. Not a search result snippet. The thing itself.\n\nThis takes about three minutes per source with the free tools in the next lesson. For a ten-source essay that is half an hour, and it is the half hour that stands between you and the most common academic-misconduct finding in the world right now.\n\n## The DOI test, which takes ten seconds\n\nEvery real journal article since roughly 2000 has a DOI. Paste it after `https://doi.org/` and press enter. A real DOI resolves to the publisher's page for that article. An invented one gives you a DOI-not-found error.\n\nThis single check disposes of most fabricated references instantly. It does not tell you whether the paper supports your claim, only that it exists. Both checks are required and this is the cheap one.\n\n## What a supervisor sees\n\nThe reason this matters more than it feels like it should: an academic reading your bibliography recognises the literature. They know the journals, often the authors, sometimes the specific paper. A fabricated citation in a field somebody has worked in for twenty years is as visible as a misspelled name in your own language.\n\nAnd the inference they draw is not *this student used AI*. It is *this student cited something they never read*, which has always been misconduct, long before any of this existed. That is the charge, and it is a fair one.\n\n## Using a model for references honestly\n\nThere is a legitimate version, and it is a search-strategy conversation rather than a citation request:\n\n> What are the main research areas relevant to this question, what search terms would find them, and which journals publish this kind of work? Do not give me citations.\n\nThen search Google Scholar, PubMed or arXiv with those terms yourself. The model is good at knowing the shape of a field and at translating your vague question into the vocabulary a database will match. It is not a bibliography. Used that way, it saves genuine time and generates nothing you then have to un-invent."},"check":{"question":"A student checks every reference in their draft by resolving each DOI. All resolve to real papers. Why is the bibliography still not verified?","options":[{"text":"Because DOIs can be recycled by publishers, so a resolving DOI may point at a different paper than the one cited.","correct":false,"tempting":"Invents a failure mode; DOIs are persistent identifiers and are not reassigned, which is the point of the system."},{"text":"Because resolving a DOI only reaches the publisher's landing page, and a paywall means the full text was never seen.","correct":false,"tempting":"Describes an access obstacle rather than the verification gap; the same problem would remain if every paper were open access."},{"text":"Because the DOI check does not confirm the year, volume and page numbers, which may still be wrong in the reference list.","correct":false,"tempting":"Names a formatting issue that resolution actually does expose, since the landing page shows the correct metadata to compare against."},{"text":"Because existence and support are separate checks, and a real paper can be attached to a claim it never makes.","correct":true}]},"prev":{"slug":"research-and-real-sources","title":"Research, And Sources That Exist","url":"https://addaly.com/learn/ai-for-students/research-and-real-sources"},"next":{"slug":"finding-the-paper-free","title":"Finding the paper, for nothing","url":"https://addaly.com/learn/ai-for-students/finding-the-paper-free"},"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) 'How a citation gets invented', in AI for Students. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-students/how-a-citation-gets-invented (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":"How a citation gets invented","courseTitle":"AI for Students","module":{"index":5,"slug":"research-and-evidence","title":"Research, sources and evidence"},"answer":null,"keyIdea":"Citations are a rigid form that a predictor reproduces easily, so plausible fabrications and real-paper hybrids both occur: resolve the DOI to check existence, and read the page to check support.","hero":{"kind":"figure","figure":{"t":"figure","label":"Figure","figure":{"left":{"title":"Does it exist? Ten seconds.","items":["Paste the DOI after the doi.org address and press enter","Search the exact title in quotation marks in Scholar or the library catalogue","A real DOI lands on the publisher’s page; an invented one errors","Check author, year, journal and volume together, not separately","Says nothing at all about whether the paper supports your sentence"]},"type":"compare","right":{"title":"Does it support the claim? Three minutes.","items":["Get the full text, then find the page your claim rests on","A real paper attached to a claim it does not make passes every other check","So does a real paper whose finding is reversed in your sentence","Note the page number; that is what makes the citation yours","If you did not open it, you did not check it"]},"title":"Two separate checks, and only one of them is cheap","caption":"Purely invented references are the easy case; a search finds nothing and you delete the line. The hybrids are the dangerous ones, because every element checks out and the citation is still false. The inference a supervisor draws is not that you used a model. It is that you cited something you never read, which was misconduct long before any of this existed."},"alt":"Figure: Two separate checks, and only one of them is cheap. Does it exist? Ten seconds.: Paste the DOI after the doi.org address and press enter; Search the exact title in quotation marks in Scholar or the library catalogue; A real DOI lands on the publisher’s page; an invented one errors; Check author, year, journal and volume together, not separately; Says nothing at all about whether the paper supports your sentence. Does it support the claim? Three minutes.: Get the full text, then find the page your claim rests on; A real paper attached to a claim it does not make passes every other check; So does a real paper whose finding is reversed in your sentence; Note the page number; that is what makes the citation yours; If you did not open it, you did not check it. Purely invented references are the easy case; a search finds nothing and you delete the line. The hybrids are the dangerous ones, because every element checks out and the citation is still false. The inference a supervisor draws is not that you used a model. It is that you cited something you never read, which was misconduct long before any of this existed."}},"blocks":[{"t":"h2","id":"watch-it-happen","runs":[{"text":"Watch it happen"}]},{"t":"p","runs":[{"text":"A citation has a rigid form: authors, year, title, journal, volume, issue, pages, DOI. Rigid forms are exactly what a next-token predictor reproduces most easily. Every component is drawn from a strong pattern:"}]},{"t":"ul","items":[[{"text":"Authors","b":true},{"text":" who publish in that area, in the right initials-and-surname format."}],[{"text":"A year","b":true},{"text":" in the plausible range for that literature."}],[{"text":"A title","b":true},{"text":" assembled from the vocabulary of the field, often a very good title."}],[{"text":"A journal","b":true},{"text":" that publishes that subject."}],[{"text":"A volume and page range","b":true},{"text":" consistent with that journal’s numbering."}],[{"text":"A DOI","b":true},{"text":" with a real prefix, since prefixes are publisher-specific and heavily represented, and a suffix that is invented."}]]},{"t":"p","runs":[{"text":"Every part is individually plausible. The combination has never existed. This is not a bug in the citation feature: there is no citation feature. It is the same process that writes the prose, applied to a highly structured string."}]},{"t":"h2","id":"the-hybrid-which-is-worse","runs":[{"text":"The hybrid, which is worse"}]},{"t":"p","runs":[{"text":"Purely invented references are the easy case; a search finds nothing and you delete it. The dangerous ones are hybrids:"}]},{"t":"ul","items":[[{"text":"A "},{"text":"real paper with a wrong year","b":true},{"text":", which breaks your reference list and looks like carelessness."}],[{"text":"Real authors, real journal, title that does not exist.","b":true}],[{"text":"A "},{"text":"real paper attached to a claim it does not make.","b":true},{"text":" This is the worst, because every element checks out and the citation is still false. Only reading the paper catches it."}],[{"text":"A "},{"text":"real paper whose findings are reversed","b":true},{"text":": the study found no effect, and the citation is attached to a sentence saying it found one."}]]},{"t":"p","runs":[{"text":"The last two are why "},{"text":"the reference exists","i":true},{"text":" is not the same as "},{"text":"the reference supports this","i":true},{"text":". Two separate checks."}]},{"t":"h2","id":"why-asking-the-model-does-not-help","runs":[{"text":"Why asking the model does not help"}]},{"t":"p","runs":[{"text":"Is this a real paper?","i":true},{"text":" is a question about the world, and the model has no channel to the world. It answers from the same process that produced the citation. It will often say yes. Occasionally it will say no about a real paper, which is even more confusing."}]},{"t":"p","runs":[{"text":"This holds when search is switched on, in a modified form. Now it may retrieve something, and the failure becomes a real link attached to a claim the page does not make. Which brings us to the only rule that survives every version of this technology."}]},{"t":"h2","id":"the-rule","runs":[{"text":"The rule"}]},{"t":"quote","runs":[{"text":"Open it. If you did not open it, you did not check it.","b":true}]},{"t":"p","runs":[{"text":"Open means: reach the actual paper or book, find the page, and read the sentence that supports your claim. Not the abstract if your claim is about a method. Not a search result snippet. The thing itself."}]},{"t":"p","runs":[{"text":"This takes about three minutes per source with the free tools in the next lesson. For a ten-source essay that is half an hour, and it is the half hour that stands between you and the most common academic-misconduct finding in the world right now."}]},{"t":"h2","id":"the-doi-test-which-takes-ten-seconds","runs":[{"text":"The DOI test, which takes ten seconds"}]},{"t":"p","runs":[{"text":"Every real journal article since roughly 2000 has a DOI. Paste it after "},{"text":"https://doi.org/","code":true},{"text":" and press enter. A real DOI resolves to the publisher’s page for that article. An invented one gives you a DOI-not-found error."}]},{"t":"p","runs":[{"text":"This single check disposes of most fabricated references instantly. It does not tell you whether the paper supports your claim, only that it exists. Both checks are required and this is the cheap one."}]},{"t":"h2","id":"what-a-supervisor-sees","runs":[{"text":"What a supervisor sees"}]},{"t":"p","runs":[{"text":"The reason this matters more than it feels like it should: an academic reading your bibliography recognises the literature. They know the journals, often the authors, sometimes the specific paper. A fabricated citation in a field somebody has worked in for twenty years is as visible as a misspelled name in your own language."}]},{"t":"p","runs":[{"text":"And the inference they draw is not "},{"text":"this student used AI","i":true},{"text":". It is "},{"text":"this student cited something they never read","i":true},{"text":", which has always been misconduct, long before any of this existed. That is the charge, and it is a fair one."}]},{"t":"h2","id":"using-a-model-for-references-honestly","runs":[{"text":"Using a model for references honestly"}]},{"t":"p","runs":[{"text":"There is a legitimate version, and it is a search-strategy conversation rather than a citation request:"}]},{"t":"quote","runs":[{"text":"What are the main research areas relevant to this question, what search terms would find them, and which journals publish this kind of work? Do not give me citations."}]},{"t":"p","runs":[{"text":"Then search Google Scholar, PubMed or arXiv with those terms yourself. The model is good at knowing the shape of a field and at translating your vague question into the vocabulary a database will match. It is not a bibliography. Used that way, it saves genuine time and generates nothing you then have to un-invent."}]}],"check":{"question":[{"text":"A student checks every reference in their draft by resolving each DOI. All resolve to real papers. Why is the bibliography still not verified?"}],"options":[{"text":[{"text":"Because DOIs can be recycled by publishers, so a resolving DOI may point at a different paper than the one cited."}],"correct":false,"tempting":[{"text":"Invents a failure mode; DOIs are persistent identifiers and are not reassigned, which is the point of the system."}]},{"text":[{"text":"Because resolving a DOI only reaches the publisher’s landing page, and a paywall means the full text was never seen."}],"correct":false,"tempting":[{"text":"Describes an access obstacle rather than the verification gap; the same problem would remain if every paper were open access."}]},{"text":[{"text":"Because the DOI check does not confirm the year, volume and page numbers, which may still be wrong in the reference list."}],"correct":false,"tempting":[{"text":"Names a formatting issue that resolution actually does expose, since the landing page shows the correct metadata to compare against."}]},{"text":[{"text":"Because existence and support are separate checks, and a real paper can be attached to a claim it never makes."}],"correct":true,"tempting":null}],"order":[2,0,3,1],"why":null},"prev":{"slug":"research-and-real-sources","title":"Research, And Sources That Exist","minutes":8},"next":{"slug":"finding-the-paper-free","title":"Finding the paper, for nothing","minutes":9}}}