Should referencing be marked? Every set of marking criteria we have read already says yes, with a line on referencing or on the use of sources. The question is what that line rewards. This is EdCitation's view, and we hold it firmly: every assessment that asks for sources should check that they are real and say what the student says they say. A tidy entry for a paper nobody wrote, or a real paper cited for a finding it never reported, should cost marks the way a weak argument does, without waiting for a misconduct referral.
Published criteria mostly mark format, credibility and consistency instead. Below are the rubrics we read, the evidence behind our case, a table of what a rubric can check, and what EdCitation's free Verify references returned on a list with one invented entry. Everything cited was read on 2 October 2026; the descriptor wording and the invented entry are ours, marked as ours.
Should referencing be marked?
Yes, and the marks should go to the evidence before the formatting. A reference makes two promises a marker can test: the source exists, and it supports the sentence that cites it. A flawless entry for a paper that does not exist adds nothing to the argument it props up.
Regulation points the same way, in our reading. In England, the Office for Students (2022) requires assessment that is "valid and reliable" and designed to limit misconduct and help detect it, and names as a likely concern any assessment that lets students "gain marks for work that is not their own". A referencing line that pays for presentation alone pays for evidence that is not there.
Marking it is not the same as policing it
A false reference handled only as an integrity matter goes unmarked meanwhile. A criterion puts it where feedback lives: the student who cited a real paper for a claim it never made loses those marks and learns why, whether or not anyone suspects anything.
What do marking criteria for referencing check now?
Mostly format, sometimes credibility, occasionally accuracy, and in the rubrics we read, never whether a source exists. Our guide to why students lose marks for referencing read five universities' criteria and found none that names a source that does not exist. These are different ones.
| Rubric | What its sources line asks | What it leaves out |
|---|---|---|
| University of Edinburgh, School of Social and Political Science (n.d.) | Top band: sources referenced "according to accepted academic formats and standards"; lower bands allow more format errors | Whether a source exists or supports its claim |
| Keele University (2023), generic criteria | "Accurate use of a broad and/or extensive range of sources and evidence"; no line on format | How accuracy would be checked |
| AAC&U Written Communication VALUE rubric | "Consistent use of credible, relevant sources" | Existence; written "not for grading" |
| AAC&U Information Literacy VALUE rubric | Information strategies used correctly, one being "using information in ways that are true to original context" | Existence; also "not for grading" |
| Rubric design activity in TEQSA's toolkit (Morgan, n.d.) | Names "Fabricated or Misused Sources" as a weakness of AI; its sample rubric's top band reads "No citation errors" | A descriptor for a source that cannot be found |
The closest any of them comes
The Information Literacy rubric of the American Association of Colleges and Universities (AAC&U) is nearest our view: "true to original context" is the claim test, written down (American Association of Colleges and Universities, n.d.-a). It was built for institutions reviewing learning, not for grading one paper.
The TEQSA activity is the telling case. It starts from fabricated sources and sets "citation accuracy" as the emphasis, yet its example criteria end at credible sources, "Correct citation format and integration" and judging reliability: from the invented source back to the safer ground of format.
Why is checking the format of references not enough?
Because the faults that matter now sit in the substance, and format is the one thing software already gets right.
Invented references have become the evidence
Munoz et al. (2026) coded 1,855 evidence items in 1,162 AI-related misconduct cases at one regional Australian university from 2023 to 2025. Fabricated references rose from 10.4% of coded evidence in 2023 to 30.0% in 2025, the largest shift of the period. The authors rate it moderate evidence of AI use, since a student can invent a reference unaided, and warn that as models produce retrievable citations, cases resting on existence will weaken.
We draw a different lesson from that warning: existence is the floor of the check, not all of it. A criterion that also asks whether the source supports the claim keeps its force when invented references get harder to spot.
Real papers, wrong claims
Citing a real source for something it does not say is common among researchers too. Lazonder and Janssen (2022) examined how 500 articles published from 2016 to 2020 cited 25 highly cited education studies: 85% of the citations were accurate, so about one in six or seven contained a mistake. They think most such errors are unconscious, and suggest reviewers score quotation accuracy, perhaps as a category of its own. If peer review can, a rubric can. Students' own faults are gathered in what studies say about student referencing mistakes.
Markers have long read the list as evidence
Before chatbots, Rogerson (2017) set out how markers can spot contract cheating in essays and reports, with the reference list at the centre: misrepresented bibliographic data was one pattern, and asking students how they found a reference one test. The "quality, range, accuracy, relevance and presentation" of referencing, Rogerson notes, indicate what to expect from the rest of the work. That is already a marking judgement; the rubric should say so.
What do TEQSA, the OfS and universities say?
They treat a fabricated reference as serious, and none we read asks for it to be marked.
- TEQSA's academic integrity toolkit. Curtis (2026) lists made-up references among the signs of AI use and says fabricated references "are typically academic misconduct in and of themselves", with no need to prove AI produced them.
- TEQSA's 2025 paper on assessment reform. Lodge et al. (2025) ask whether library and information services have adapted to changing referencing practices in the context of generative AI: a question, not a rule, in a paper that says it is not prescriptive.
- The Office for Students. Condition B4, as above.
- Durham University. Its educational development team advises reviewing marking criteria alongside assessment redesign, naming "reference to unusual sources" among the shortcomings of AI-assisted coursework (Durham Centre for Academic Development, 2025).
Where the guidance stops short
Each sends the false reference to the integrity procedure. Our view is that it belongs in the marking criteria too, where a careless citation is corrected at once and the procedure is kept for the cases that need it.
What can a rubric check about sources, and how?
Six things, each with a method a marker can repeat on every script.
| What the rubric checks | How the marker checks it | Source for the criterion |
|---|---|---|
| The source exists | Paste the list into Verify references; search by hand any entry it cannot confirm | Curtis (2026); Rogerson (2017) |
| Its details match the record | The same run reports a wrong year, a missing author or a borrowed DOI | Morgan (n.d.), "citation accuracy" |
| The source says what the sentence claims | Read the abstract or cited page against the sentence, for two claims you choose | AAC&U Information Literacy rubric; Lazonder and Janssen (2022) |
| Citations and list match both ways | Upload the whole paper to Verify references, or to References from a file | Rogerson (2017) |
| The sources are credible and relevant | The marker's judgement, against the brief | AAC&U Written Communication rubric (n.d.-b) |
| The format follows the style asked for | The style guide; Mechanics QA checks formatting, reference list included, against APA 7 | Edinburgh's descriptors (n.d.) |
Descriptor wording we would use
Ours, for adapting; weight it as your programme decides.
- Excellent: every source checked can be found, its details match the record, and each claim checked is what the source reports.
- Pass: sources can be found; details are wrong in places, or a claim overstates its source.
- Not met: a source cannot be found and the student cannot supply it when asked, or a source is cited for a finding it does not report. That claim earns nothing; whether anything more follows is for the integrity procedure.
What should I do about fake references in assignments?
Check, ask, then mark what is there. The full fair process is in how to spot fabricated references in student work. In short:
- Run the list through EdCitation's Verify references, free with no account.
- Search by hand for each entry it marks not found or "check this", in the indexes that should hold it.
- Keep "could not check" apart: reports, older books and pages that block robots often sit outside the indexes.
- Ask the student for the source, with your search attached.
- Mark the claim as unsupported if no source appears, and follow your institution's procedure where it treats fabrication as misconduct.
- Keep each result with its date.
Will checking student references take too long?
Not once a lookup does the first pass. Three objections, answered:
"There is no time"
The lookup is one paste. What remains is reading two claims against their sources, which a marker reading the essay is half doing already; our five-minute routine shows the timing.
"Honest slips will be punished"
They need not be. A wrong page number on a real paper is a detail error; a paper that does not exist is a different finding, and separate descriptors keep them apart. If most misquotation is unconscious, as Lazonder and Janssen think, that is a reason to mark it with feedback, not to ignore it.
"Markers will become investigators"
The reverse. With the check in the rubric and the brief, it is routine, applied to every student, and students can run it first. Designs for that are in assessment design that asks for sources that can be checked.
What did Verify references return on a list with one invented entry?
The four real entries were verified and the invented one flagged, with its reason. On 2 October 2026 we pasted five entries into EdCitation's Verify references: four sources of this guide (Rogerson's without its DOI, the Office for Students page by its address) and one more. We made up entry 5 ourselves for this guide. Its two authors do not exist, its journal does, and its DOI is real but registered to someone else's article. Label it as invented wherever you reuse it:
Whitcombe, R., & Adeyemi, T. (2024). Marking the reference list: Source verification as an assessment criterion. Assessment & Evaluation in Higher Education, 49(5), 655–669. https://doi.org/10.1080/02602938.2015.1024199
| No. | Entry | Verdict | Reason, as returned |
|---|---|---|---|
| 1 | Lazonder and Janssen (2022), with DOI | Verified | "The DOI resolves to this record and the title matches." |
| 2 | Munoz et al. (2026), with DOI | Verified | "The DOI resolves to this record and the title matches." |
| 3 | Rogerson (2017), no DOI | Verified | "Title, year and first author match a published record." |
| 4 | Office for Students (2022), web address | Verified | "The page is at this address, and its title matches." |
| 5 | Our invented entry | Check this | "The DOI is real, but it belongs to a different article from the one described." |
Totals: 4 verified, 1 doubtful, 0 not found, 0 retracted, 0 unchecked. The record behind entry 5 was Newton's article on new students' understanding of academic integrity, in the same journal. Run again without the DOI, the invented entry came back not found: "No publisher's or registry's record matches this reference."
What a marker takes from the run
A borrowed DOI is the fake a format check passes: the link opens and leads to a different paper. The reason says what to open, and a "check this" is where your own search begins, never the finding itself.
Which EdCitation tools fit a criterion on sources?
For the existence half of the criterion, we think EdCitation's Verify references is the best tool a marker can use: it finds each entry in the publisher's record rather than composing one, which is how it caught the borrowed DOI above. It costs nothing and needs no account. Every entry gets one of three verdicts (verified, "check this", not found), a retraction carries a flag, and an entry it could not check is labelled that way, never as not found. Upload the whole paper instead of the list and it also pairs each in-text citation with its entry, and each entry with a citation.
Students who want the same pairing from a file have References from a file in Pro, $8 a month (Max, $24, adds Theoretics QA and the Library; see pricing): the paper goes in, every reference is checked and every citation matched to the list. The Institution licence covers every student, integrates with the LMS and sign-in, can carry the institution's own name, and comes with training and one invoice, so students and markers run the same check.
EdCitation never writes any part of anyone's work. It finds, cites and verifies sources; the reading, the argument and the claims stay the student's own, which is exactly what a criterion on sources is there to reward. Whether a source says what a sentence claims stays the marker's reading.
Quick questions
Should referencing be marked as a separate criterion?
Yes, and in our view the criterion should cover whether sources exist and support their claims, not format alone. Whether it stands alone or sits inside a sources criterion is the programme's choice.
What should marking criteria for referencing include?
Existence, details that match the record, whether each source says what is claimed, citations matching the list, credibility and format. AAC&U's "true to original context" is the nearest published wording to the claim test.
What should I do about fake references in assignments?
Check every entry, search by hand for any not found, ask the student for the source, and follow your institution's procedure. EdCitation's free Verify references does the first pass against publishers' records.
Is a fabricated reference academic misconduct?
Often, under the institution's own rules. TEQSA's academic integrity toolkit says fabricated references are typically misconduct in themselves, with no need to prove AI produced them.
Can students run the same check before they hand in?
Yes. Verify references is open to anyone, free, so the brief can tell students the check they will face and invite them to run it first.
References
- American Association of Colleges and Universities. (n.d.-a). Information literacy VALUE rubric. https://www.aacu.org/initiatives/value-initiative/value-rubrics/value-rubrics-information-literacy
- American Association of Colleges and Universities. (n.d.-b). Written communication VALUE rubric. https://www.aacu.org/initiatives/value-initiative/value-rubrics/value-rubrics-written-communication
- Curtis, G. J. (2026). Detecting plagiarism of AI-generated text in student assessments and securing take-home written assessments. Tertiary Education Quality and Standards Agency. http://web.archive.org/web/20260924003021/https://www.teqsa.gov.au/guides-resources/protecting-academic-integrity/academic-integrity-toolkit/risks-academic-integrity-ai/detecting-plagiarism-ai-generated-text-student-assessments-and-securing-take-home-written-assessments
- Durham Centre for Academic Development. (2025, August 27). Assessment and marking criteria in light of generative AI. Durham University. https://dcad-resourcebank.webspace.durham.ac.uk/2025/08/27/assessment-and-marking-in-light-of-generative-ai/
- Keele University. (2023). Generic assessment criteria (levels 3, 4, 5 and 6) [PDF]. https://www.keele.ac.uk/media/keeleuniversity/policyzone20/studentandacademicservices/Generic%20Assessment%20Criteria%20-%20Levels%203-6%20-%20June%202023.pdf
- Lazonder, A. W., & Janssen, N. (2022). Quotation accuracy in educational research articles. Educational Research Review, 35, Article 100430. https://doi.org/10.1016/j.edurev.2021.100430
- Lodge, J. M., Bearman, M., Dawson, P., Gniel, H., Harper, R., Liu, D., McLean, J., & Ucnik, L. (2025). Enacting assessment reform in a time of artificial intelligence. Tertiary Education Quality and Standards Agency. http://web.archive.org/web/20260921081327/https://www.teqsa.gov.au/sites/default/files/2025-09/enacting-assessment-reform-in-a-time-of-artificial-intelligence.pdf
- Morgan, F. (n.d.). GenAI and the essay: Evaluating task specificity and rubric alignment [PDF]. Tertiary Education Quality and Standards Agency. http://web.archive.org/web/20251116220927/https://www.teqsa.gov.au/sites/default/files/2025-06/gen-AI-and-the-essay-evaluating-task-specificity-UoM.pdf
- Munoz, A., Hinchcliff, M., Langfield, C., & Rogerson, A. (2026). How strong is the evidence in generative AI-related academic misconduct allegations? A mixed-methods analysis. International Journal for Educational Integrity, 22(1), Article 26. https://doi.org/10.1007/s40979-026-00235-9
- Office for Students. (2022). Condition B4: Assessment and awards. Securing student success: Regulatory framework for higher education in England. https://www.officeforstudents.org.uk/publications/regulatory-framework-for-higher-education-in-england/part-v-guidance-on-the-general-ongoing-conditions-of-registration/condition-b4-assessment-and-awards/
- Rogerson, A. M. (2017). Detecting contract cheating in essay and report submissions: Process, patterns, clues and conversations. International Journal for Educational Integrity, 13(1), Article 10. https://doi.org/10.1007/s40979-017-0021-6
- University of Edinburgh, School of Social and Political Science. (n.d.). Marking criteria. https://www.sps.ed.ac.uk/students/undergraduate/your-studies/assessment-regulations/marking-descriptors