Imagine that you are working on a paper or thesis chapter and ask an AI tool for help with sources. The tool lists authors, titles, journals, years, and page numbers. You feel satisfied because the list looks complete and you paste those references into your draft. Later a teacher or supervisor asks you to show the original papers. You look for and find nothing - the citations never existed. This problem happens every day to students and doctoral candidates who treat AI output as finished work. AI systems generate text by predicting likely word sequences. They do not hold a database of real publications. When the system lacks a matching source, it invents one that fits the pattern of academic writing. Studies of large language models show high rates of fabricated citations (called AI hallucination) across subjects. You can protect your work by treating every AI-generated reference as untrusted until you verify it yourself.
Start with the title and authors. Copy the full title into Google Scholar, place quotation marks around the exact title so the search looks for the precise string. Add the first author surname and the publication year. If the search returns zero results, the paper almost never exists. Do it several times with different publication years and initials of the author (full spelling, with dot in the end etc.). Try the same search in your university library catalogue or in a subject database such as ERIC for education topics, PubMed for health, or JSTOR for humanities. Still no match means you discard the reference. Do not assume a typing error on your part. AI systems can invent plausible titles that follow real naming conventions yet point to no document.
Next examine the journal or publisher name. Type the journal title into a browser and open its official site. Look for the volume and issue that the AI listed. If the site shows no such issue or the years do not match, the citation is false. Check the DOI if you have one. Paste the DOI into doi.org or just in a browser search field. A working link will take you to the real article page but a broken link or a message that the identifier does not resolve proves the DOI is fake. Many AI tools invent DOIs that look correct in format however lead nowhere. Testing of DOI is very important since publishers request providing DOI in the reference list and will apply DOI to your work also – this is a unique indicator of a publication’s existence within the academic sphere.
Read the abstract or full text once you find a real paper. AI tools often invent a citation and then attribute to it a statement that the paper never makes. Open the PDF or HTML version, search inside the document for key phrases from the statement you wish to verify. If the phrases are absent, the citation fails. Even when the paper exists, the AI may misstate its findings. To ensure your propositions are true, verify that the source actually contains the evidence you need.
Use author names as a further check. Search the lead author in Google Scholar https://scholar.google.com/ or ORCID https://orcid.org/ . Real researchers maintain profiles that list their publications. If the claimed paper does not appear on the author’s list, there is a high probability the reference is invented. Look at the co-authors. People who publish together usually share institutional affiliations or prior joint work. Random combinations of famous names from different fields indicate a forgery.
Apply these steps in everyday study situations. Imagine you write a short paper on teaching reading in primary grades. When you ask an AI tool for recent sources it gives three references. One claims a 2024 article by Martinez in the Journal of Literacy Research. You search Google Scholar with the exact title and author, and no hit appears. You open the journal website and find the 2024 issues contain no such article. You delete the reference and search the real literature yourself. Another reference points to a book chapter. You locate the book through your library and discover the chapter title is different and the page numbers do not match. You should correct the details or replace the source. The third reference turns out real, yet the AI misquoted the sample size. You adjust your text to match the actual numbers. Such checks take only a few minutes and help avoid awkward situations in the future.
Doctoral students face higher stakes. A dissertation chapter on assessment practices may contain dozens of citations. You generate a draft outline with AI assistance and receive a long reference list. You set aside time to verify every entry before you write the paper. Create a simple spreadsheet with columns for title, author, year, source found, DOI status, and claim match. Work through the list one row at a time and mark any entry that fails a check. Replace failed entries with sources you locate yourself in the databases you already use for literature reviews. Supervisors and members of the commission notice when references look too perfect or when DOIs fail - they often open the papers. You must avoid questions about academic integrity by finishing the verification before submission of the work. The same process applies when you write a grant proposal or a conference abstract. Funders and reviewers examine citations. False ones can damage credibility at once.
Build habits that reduce reliance on unchecked AI lists. Keep a permanent, updated file of sources you have already read and verified. When you need support for a claim, first look in your own file. Only after that file does not help do a search in external databases. When you do use an AI tool for ideas, ask it for keywords or topic outlines rather than for complete references. Then use those keywords in Google Scholar or your library search and collect real results. This order keeps the AI in a supporting role and allows you to control the evidence.
Watch for patterns that hint at hallucination. AI systems often produce references with consecutive page numbers that look tidy, journals with slight spelling variations on real titles, or years that fall just after the model’s training cutoff. They can also mix real author names with invented article titles. When you see any of these signs, raise your verification effort. Check the same claim across two or three independent sources. If only the citation supports the claim and no other literature mentions it, consider the result as unsupported.
The primary education sphere contains the same risks. A student teacher prepares a unit plan on environmental science and asks an AI tool for research on outdoor learning effects. The tool provides a citation to a 2023 study in a journal called Primary Science Review. But a quick search shows the journal exists under a different name and the 2023 volume contains no matching article. The student teacher should find a real meta-analysis instead and prepares the unit on verified findings.
Another example. A master’s student writes about inclusive classroom practices and an AI list includes a paper by an author who works in a different field. The ORCID profile of the author shows no education publications. Apparently, the student must discard the entry and find peer-reviewed work from special education journals.
You strengthen your own research skills each time you complete these checks. Searching databases, reading abstracts, and matching claims to evidence become faster with practice. You also learn to recognise the style of genuine academic writing. Real papers contain specific methods, sample descriptions, and limitations. Invented ones stay vague. Over time you need the AI less for source lists because your own search habits improve.
When you submit the work, keep records of your verification steps. Note the date you checked each reference and the database you used. If a question arises later, you show the history. This practice protects you in universities that use plagiarism software or AI detection tools. Those systems find unusual citation patterns. Clean and verified references reduce the chance of false results.
The process remains the same over different disciplines. History students check archival references and secondary sources, science students confirm experimental papers, education students verify classroom studies. In every case you open the source, read the relevant section, and confirm the link between the citation and the evidence. You never paste a reference you have not opened and checked.
Students sometimes feel pressure to produce long reference lists quickly. AI tools seem an excellent instrument to solve that pressure but they create new problems instead. A short list of real, carefully checked sources serves you better than a long list of invented literature. Quality of evidence matters more than quantity since your readers judge the strength of your argument by the sources that support it. You must give them real sources.
Apply the same approach to secondary citations. AI tools can invent a claim and attribute it to a real paper that never made the claim. You open the real paper and discover the mismatch. You either reject the claim or find a different source that does support it. This step prevents the existence of inaccurate statements in student work. In this article, you got a clear understanding of sequence of actions. Search the title and authors in scholarly databases, test the journal and DOI, open the document and match the claim, check author profiles, replace failures with sources you locate yourself, record your checks. These actions take modest time and protect the integrity of every paper, thesis chapter, or presentation you prepare. You remain the person responsible for the accuracy of your references. AI tools generate text and you verify facts.
About the author: Teimuraz Kancheli is Head of the Electronic Resources Division at the Ivane Javakhishvili Tbilisi State University Library. His professional interests include AI for academic research, digital library services, electronic scholarly resources, and scientific publishing.
Further resources: [AI in Higher Education , professional profile ]
#AIcitations #ReferenceCheck #AcademicIntegrity #StudentResearch #DoctoralWriting #CitationVerification #AvoidHallucinations #GoogleScholarTips #ThesisTips #EducationResearch
keywords: AI generated citations, check references, fabricated sources, verify DOI, Google Scholar search, student paper integrity, doctoral thesis references, primary education sources, academic honesty, literature review accuracy
