If a genuine source does not appear to support a learner’s claim, check the cited passage and its context before deciding what the mismatch means. Then apply the published marking criteria to the work submitted: credit what the learner has demonstrated and explain where the evidence falls short. Do not assume the mismatch is misconduct.

A citation helps a reader find a source; it does not, by itself, show that the source supports the point being made. Oxford’s referencing guidance explains that transparent citation also allows readers to check an interpretation. York’s assessment guidance distinguishes using source information to support an argument from the mechanics of referencing. ([ox.ac.uk](https://www.ox.ac.uk/students/academic/guidance/skills/referencing?utm_source=openai))

First, check what the source actually says

Read the learner’s sentence with the surrounding paragraph. Identify the precise claim: is the learner reporting a finding, offering their own interpretation, or citing a source as background? Then:

  1. Find the cited passage. Check the page, section or other locator, as well as the reference-list details. If the learner cites a secondary source, check what that source says about the original.
  2. Read beyond the extracted sentence. Look for qualifications, the population or setting, the date, and whether the statement is a finding, recommendation or opinion.
  3. Compare like with like. A source about staff preferences, for example, does not necessarily demonstrate improved productivity. An association is not automatically evidence of a cause.
  4. Check the learner’s wider argument. Another cited source or explanation may support part of the claim. Do not treat every statement in the assignment as unsupported because one citation is problematic.

If you cannot access the cited version, do not conclude that it says nothing relevant. Note the access problem and seek an approved route to check it, or ask the assessment lead how to proceed under the applicable rules. A missing page number or incorrect reference detail may make checking harder, but it is not the same finding as a source that has been checked and does not support the claim.

A fictional example: a source exists, but the claim goes too far

A learner recommends flexible working in a management assignment. Their submission says:

“The staff survey shows that flexible working increases productivity across the organisation.”

The citation points to a fictional internal staff survey report. The assessor reads the cited section. It says that 24 staff responded and 15 selected flexible working as a preferred working arrangement. The report did not measure productivity. Elsewhere, the learner discusses implementation costs and practical risks.

The source is relevant to staff preference. It does not establish a productivity increase, and the respondents’ preferences do not automatically describe everyone in the organisation. The learner may have drawn an over-broad conclusion from relevant evidence; the assessor cannot tell why from this mismatch alone.

Apply the published criterion

Suppose this fictional assignment has a four-mark criterion for using evidence to justify a recommendation:

  • 0: No relevant evidence is used.
  • 1: Evidence is mentioned, but the recommendation is largely unsupported.
  • 2: Relevant evidence is used, but a significant claim is unsupported or a significant limitation is missed.
  • 3: Evidence generally supports the recommendation and a relevant limitation is recognised.
  • 4: Relevant evidence supports a well-reasoned evaluation, including its limitations.

In this example, the assessor could award 2 out of 4: the preference data and discussion of costs and risks are relevant, but the productivity claim is significant and unsupported. The assessor should record that reason alongside any other criterion evidence considered. A different rubric or fuller submission could justify a different outcome; this example is not a universal marking rule.

If referencing accuracy is assessed separately, check that criterion too. In this example the source is identifiable, so the unsupported interpretation is not automatically a referencing-format error. Avoid deducting twice for the same issue unless the published scheme requires separate judgements.

Write feedback the learner can use

Keep the comment focused on the claim, the evidence and the next step. For the fictional assignment, feedback might say:

“You use the survey to identify staff interest in flexible working and discuss implementation risks. However, the cited section reports preferences, not productivity, so it does not support your claim that productivity increased. Revisit the wording: distinguish what the survey shows from what you infer, and use relevant evidence if you want to make a productivity claim.”

If the assessment permits revision or resubmission, explain the permitted next step and deadline under the published rules. Otherwise, frame the comment as guidance for future work rather than an invitation to change the submitted assignment.

When should you use a separate integrity process?

A source mismatch can arise from misunderstanding, an incorrect locator, an overstatement or another cause. Marking the quality of evidence is different from deciding whether assessment rules have been breached.

Follow your organisation’s and, where applicable, awarding body’s integrity procedure if the available evidence meets its referral threshold—for example, a quotation attributed to a real source cannot be found in the checked version, or there are several materially inaccurate attributions that warrant a closer look. The University of Sussex’s guidance includes fabricated or altered quotations and citations among its examples of fabrication; local definitions and procedures may differ. ([sussex.ac.uk](https://www.sussex.ac.uk/skills-hub/referencing-and-academic-integrity/?utm_source=openai))

Record the specific claim, citation, version and passages checked, including any access limits. Refer the concern to the designated person rather than making an allegation in routine feedback. Follow the process for giving the learner an opportunity to respond, and check whether marking or result release should pause. An AI or similarity indicator, if available, is information for human review—not proof of misconduct or of how the work was produced.

Keep the marking decision connected to the work

In EduMark Studio, assessors can review the submission alongside the criteria and record feedback against the work. Use that context to explain which claim was checked, how it affected the criterion and what the learner can do next. Optional AI assistance is in beta; a person reviews and confirms the marking decision. Source checking and any integrity referral remain human responsibilities. ([edumark.ai](https://www.edumark.ai/marking-studio?utm_source=openai))

Sources & further reading

A little less admin. A lot more possibility.

See how EduMark connects learning, assessment and your team.

Explore EduMark →