Academic integrity at UCD: the similarity result you have to ask to see

The short answer

UCD runs text-based similarity software over submitted work, and whether you ever see your own result is a decision your Module Coordinator makes, module by module. There is no university-wide self-check area. If seeing the result before the deadline would help you, the policy makes that a request you can put to one named person.

The rest follows from that. AI use has to be permitted and acknowledged, the module descriptor is where the permission lives, and there is no opt-out from being checked. UCD is unusually clear about how much weight a detection result carries and about the standard of proof it applies — both of which are set out below, because they matter most if something goes wrong.

The result is at your coordinator’s discretion

Two sentences in the Academic Integrity Policy (Office of the Registrar; approved by the Academic Council Executive Committee, May 2024; effective 1 September 2024) decide this. From §4.2, verbatim:

“Module Coordinators should inform students about UCD’s text-based similarity detecting software. It is at the discretion of the Module Coordinator to allow students to have access to the results of the detection software.”

Unpack that carefully, because both halves are doing something.

The first sentence is a duty owed to you: your Module Coordinator should tell you the software exists. If nobody has, you are not the one who missed something.

The second is a discretion, not a right. Some coordinators release similarity results to students; some do not; the policy leaves it to them. So the practical move is to ask, in the first weeks of the module rather than the night before submission: will we be able to see our own similarity report before the deadline? A yes is worth having, because a similarity report read early is a referencing tool. A no is worth knowing too, so you stop expecting one.

The University’s own undertakings, at §3.1, confirm the software is institution-wide rather than a local experiment:

“c) Use electronic and other detection mechanisms, such as text-matching software, to identify instances of potential academic misconduct.”

UCD does not name the product. Turnitin appears zero times in the Academic Integrity Policy, and the IT Services URL we probed for it returned 404. Nor does UCD state whether any AI-writing detector is switched on. Those are absences in the published record, not conclusions — see the gaps section below.

If a similarity figure does reach you and you are unsure how to read it, what a similarity or AI score actually means is narrower than most students assume.

The AI rule: §4.3, in full

Academic Integrity Policy §4.3, “Use of Generative Artificial Intelligence Technologies”, verbatim:

“If permissible (by faculty/tutor/instructor) and when used appropriately, generative artificial intelligence (AI) tools can offer support across various aspects of the learning process. However, outputs from these tools must always be considered in the same manner as work created by another person/persons i.e. used critically, ethically, cited and acknowledged appropriately.
This academic integrity policy prohibits students from representing work as their own that they did not write, code or create. Accordingly, submission of AI-generated content without explicit permission and attribution is not allowed.”

Note the framing device in the first paragraph: AI output is treated as if it were another person’s work. That single analogy answers most of the questions students ask. Would you cite it if a colleague had written it? Then cite it.

The second paragraph sets a double condition — explicit permission and attribution. Permission alone is not enough, and attribution alone is not enough. Both.

The section then delegates the permission question, verbatim:

“• It should be indicated clearly in the module descriptor whether generative AI will form any part of the learning experience.
• All faculty / tutors / instructors may allow the use of generative AI to complete specific assignments.
• Specific guidance based on disciplinary expectations should be provided to students in the use of generative AI where it is permitted.
• Students are expected to follow each step of that guidance and properly acknowledge the use of generative AI in each aspect of their submitted work.
• If a student has any doubt about whether a specific use of generative AI is permitted for an assignment or course, they are responsible for discussing it with the faculty member / tutor / instructor prior to using it.
• Students must indicate clearly in submitted work the nature and extent of any outside assistance (including the use of professional tutors, machine learning or AI technologies) according to the citation practices set out by the relevant subject/discipline.”

Two of those bullets place the burden on you personally. Doubt is your responsibility to resolve, before using the tool — so email, and keep the reply. And the disclosure required is the nature and extent of the assistance, which is more than a one-line note that AI was used. It is much easier to write that honestly if you kept a record as you worked rather than reconstructing it afterwards.

UCD publishes no single prescribed wording for an AI acknowledgement — unlike Trinity, which prescribes a coversheet declaration. The citation practice of your own discipline governs. Our guide to AI disclosure statements covers what a good acknowledgement generally contains, but your discipline’s convention wins.

The corresponding duty on staff is at §3.5: Module Coordinators will, via the module descriptor, provide students with information on “i. Expectations for citation methods in that module; ii. Whether the use of generative artificial intelligence, or other machine learning technology, is permitted.” Section 4.2 encourages the same clarity about collaborative work and “whether the use of Generative AI is permissible and under what conditions”.

So: the module descriptor is the document to read. If it is silent, §4.3 puts the onus on you to ask before using anything.

Paraphrasing and translation software

This one works against our commercial interest and we are going to state it plainly, because you need it more than we need the sale.

UCD’s definitions list makes a category of misconduct out of “Inappropriately using digital or information technology to complete an assessment task” — that is, using such technology “without explicit permission from relevant academic staff and / or not acknowledging use of such technology when its use is permitted”. Its two named examples, verbatim:

“i. unauthorised and / or unacknowledged use of artificial intelligence tools to generate content for assessment purposes; or
ii. unauthorised and / or unacknowledged use of paraphrasing or translation software to, for example, disguise plagiarism, collusion, contract cheating or other academic integrity breach.”

UCD is the only institution in this group with a clause naming paraphrasing and translation software directly. Two things follow.

Running your text through a rewriting or paraphrasing tool to make it look less like something is itself named as misconduct at UCD. Not a grey area, not a lesser offence — an example in the definitions list. If your instinct after a similarity result is to reword until the number drops, that instinct is the thing the clause is written about.

Translation software is in the same sentence, and that catches an innocent case too. If you write in another language and translate, that is “use of translation software”. The clause bites where the use is unauthorised or unacknowledged. So if you drafted in your first language and translated, the safe course is to say so and to ask whether it is permitted — not to assume it is invisible.

For the record about our own product: our AI check reads text and reports on it. It does not rewrite your work, and nothing on the result screen rewrites it for you. If a document is largely machine-written, rewording does not fix that, and we would rather tell you so.

What UCD may use as evidence

Section 7 of the Academic Integrity Policy is the evidence section, and it opens with the sentence that removes any doubt about consent:

“Any work submitted for assessment may be subject to electronic or other detection procedures.”

The next paragraph is the one worth knowing if you are ever accused:

“Where employed, electronic detection software, such as text-based similarity or AI detection software, should be used with caution and regarded only as one tool in assisting an examiner to make a judgement about academic misconduct. Any software used to detect academic misconduct in assessed work must be approved for use by the University.”

That is a rule about how a result may be handled, written into the policy itself by the Academic Council Executive Committee. A detector output is positioned as one tool assisting a human judgement — the examiner makes the judgement, the software assists it. If an allegation appears to rest on a score alone, this sentence is in the same document the allegation is brought under.

The section then sets out what an examiner is expected to do instead of stopping at a number:

“• If academic misconduct is suspected by an examiner, the examiner should employ all reasonable means to clarify whether academic misconduct has taken place.
• Where, as a result of a student’s performance in another assessment task within a module or component, an examiner forms the reasonable suspicion that an assessment may not be a student’s own unaided work (excluding legitimate cooperation), the examiner must report the matter…
• Computer and systems records, including details of access to the Virtual Learning Environment, and other student online activity may be reviewed to identify or substantiate potential academic misconduct.
• The examiner or Module Coordinator can, after outlining their concerns, require a student to discuss or explain components of their assessment tasks to determine the authenticity of their work.”

The third bullet is a real processing claim and you should know it exists: UCD reserves the right to review VLE access logs and other online activity. It is used as corroboration — an alternative to relying on a detector — and it can point either way. A record showing you working steadily through a module is evidence in your favour.

The fourth is the one students find most stressful and it is, on balance, good news: you can be asked to explain your own work. That is a conversation you can prepare for. Know your argument, your sources, why you cut the section you cut. See how to prove you wrote it yourself and what to do when you are accused. If you write in English as a second language, how detectors treat non-native writing is worth reading before that meeting.

If a detector score is what started this, our guide to an AI detector flagging your thesis sets out what that number does and does not establish on its own.

Reporting is mandatory once suspicion is reasonable: “Any examiner or assessor who reasonably believes that a student has breached academic integrity requirements must report it to the relevant Module Coordinator and/or the person responsible for dealing with academic integrity.” Your examiner does not have discretion to let it go quietly.

The standard of proof, and why it is unusual

UCD is the only institution in this group of five that publishes the standard by which allegations are judged. Quoting the national Framework, §7 states:

“As per the national Framework for Academic Misconduct Investigation and Case Management, UCD will ‘Investigate suspected [academic integrity] breaches as a lay proceeding, using the standard from civil law, where the “balance of probabilities” is the relevant test to which allegations must be subjected. The balance of probabilities is based on “clear and convincing evidence” that it is more likely than not that the allegation is true. This is less demanding than the legal test of “beyond reasonable doubt”.’”

The Student Academic Misconduct Procedure repeats it operationally: decisions are made “based on the evidence before them and on the balance of probabilities”, and “Decisions will be taken by a simple majority and will be made on the balance of probabilities”.

Three things are being said at once, and it is worth separating them because the combination is genuinely unusual.

“A lay proceeding.” This is not a court. The people deciding are academics applying a civil-law standard, not lawyers, and the process is designed around procedural fairness rather than rules of evidence.

“Balance of probabilities.” The test is whether it is more likely than not — over 50%, not near-certainty. The policy says so explicitly: less demanding than “beyond reasonable doubt”. Being honest about that is more useful to you than pretending the bar is higher than it is.

“Clear and convincing evidence.” And here is the part that cuts your way. UCD does not leave the standard at bare probability; it attaches a quality requirement to the evidence that gets you there. Those two phrases usually belong to different standards, and putting them together is an unusual formulation. Whatever satisfies the decision-maker has to be clear and convincing — read alongside the §7 instruction that detection software is “one tool” to be “used with caution”, and the expectation that an examiner will “employ all reasonable means to clarify”.

If you are ever in front of a School Academic Integrity Committee, those are the words to work with: not that the case is unproven beyond reasonable doubt, but that the evidence is not clear and convincing, and that the reasonable means of clarification were or were not employed.

We hold our own tool to the same distinction. Our published error rates are measured on English and German corpora, and no detector — ours included — produces proof. See how accurate AI detectors really are.

What counts as plagiarism at UCD

The policy’s definition is long, but one paragraph inside it is worth every student’s attention:

“Plagiarism is unacceptable in academic work, even where it arises as a result of:
• poor referencing; • error; • inability to paraphrase; or • inhibition about writing in the student’s own words.”

Intention is not a defence to the finding. It does bear on the penalty — the Procedure lists “first-time offenses, lack of intent, genuine [error]” among factors warranting a more lenient penalty, and “repeated offenses, deliberate deception, premeditation” among those warranting a stricter one. But “I did not mean to” does not undo the classification.

The named forms of plagiarism cover using extracts from published or unpublished work (including from the internet) without acknowledgement; presenting direct extracts without quotation marks or other appropriate indication, where “It is not sufficient simply to acknowledge the source”; copying the same or a very similar idea; changing the order of words while retaining the original idea; copying non-written material including images, computer code, musical notation, recordings and performances; and using another student’s work “in a way that exceeds the bounds of legitimate cooperation”.

The fourth of those — “changes the order of words taken from source material but retains the original idea or concept without appropriate acknowledgement” — is the classic paraphrasing trap, and it connects directly to the software clause discussed above.

Two further definitions catch things students often do not classify as offences. Self-plagiarism: “Reusing one’s own work without citing or acknowledging its original use. This could mean submitting one piece of work in more than one course.” And uploading or sharing an assessment: publishing an assessment or part of one, including responses to university assessment questions, to a website, file-sharing site or other online platform without explicit permission from the owner of the material and/or the Module Coordinator — which “may also be a breach of copyright laws”.

That last one is worth taking seriously, and it is a reason to think about where your writing ends up generally. For what it is worth, our own check does not add your text to any third-party comparison database.

The policy grounds itself nationally, and names its sources: the National Academic Integrity Network’s National Principle and Common Lexicon of Terms (2021), the National Academic Integrity Guidelines (2021) and the National Framework for Academic Misconduct Investigation and Case Management (2023). A footnote points to section 43A of the Qualifications and Quality Assurance (Education & Training) Act 2012, which makes it an offence to provide or advertise cheating services.

Penalties, the Tariff, and the appeal clock

The decision-maker is the School Academic Integrity Committee, under the Student Academic Misconduct Procedure (approved by Academic Council 30 April 2025, effective 1 September 2025). Its stated principles, verbatim from §1:

“This procedure is intended to support the application of procedural fairness for all students. It has been established on the principles of natural justice. Those with responsibility for applying this procedure should:
● treat students fairly, consistently and in a transparent manner,
● apply penalties that are fair and proportionate,
● respect the dignity of all persons involved,
● make decisions that are free from bias.”

Beyond a warning, a direction to advice, and recording in the Academic Misconduct Record System, the Committee may under §5.3.6(b):

“i. Permit the student to re-submit the assessment component without academic penalty.
ii. Permit the student to re-submit the assessment component and cap the resubmitted work…
iii. Permit the student to resubmit the assessment component and reduce the grade achieved in the resubmitted work by a specified number of grade points.
iv. Reduce the grade for the assessment without an opportunity to resubmit the assessment…
v. Where an assessment component is graded as Pass/Fail or where academic misconduct occurs in a resit, School Academic Integrity Committees may apply an NM grade… or consider referral to the Student Discipline Procedure.”

Note that the first option is resubmission without academic penalty. The scale starts somewhere survivable.

Penalties are guided by a published instrument:

“In cases where it is determined that academic misconduct has taken place, the penalty will be guided by a University approved UCD Plagiarism Tariff. It is noted that the UCD Plagiarism Tariff is not appropriate for guiding penalties in relation to some categories of academic misconduct, such as collusion.”

Some cases skip the Committee altogether:

“c) Refer the alleged instance, without any decision, for resolution under the University’s Student Discipline Procedure. In some contexts, a first instance may require direct referral (e.g. misconduct in single-assessment modules, second or subsequent offences or heavily weighted assessments.”

The unclosed bracket is in the original. Exclusion, expulsion and any question of degree revocation sit in that Student Discipline Procedure, which is a separate document we did not read — so this page says nothing about what it contains. If you are referred there, get that document.

The appeal clock is short.Section 6.1: “An appeal may be made to the University’s Student Appeals Committee within 10 working days from the date of issue of the decision of the Subcommittee.” The Committee may uphold the appeal in full or in part, or not uphold it.

Decisions “will be communicated to students through their UCD email address (copying the Module Coordinator)”, and where a decision is made under §5.3.6b the communication should include reference to the right to appeal. Read your UCD email. Ten working days from the date of issue is not long, and the notice will be sitting in that account.

One procedural asymmetry worth knowing: where a case is referred without a decisionto the Student Discipline Procedure, the Student Appeals Procedure does not apply to the referral itself — the Procedure’s own footnote says so, and adds that students will be given the opportunity to appeal decisions made under that other procedure.

What UCD does not publish

A short access note first, because it affects how to read these. UCD’s Academic Secretariat policies page, its governance resources policies page and its student academic-integrity page all returned 404 when we tried them. The documents themselves are published — in the UCD Governance Document Library — and that is where the policy and procedure quoted here come from. Each page of the policy carries the reminder: “All policies and policy related documents and forms are subject to amendment. Please refer to the UCD Governance Document Library website for the official, most recent version.” Do that if anything turns on the wording.

• The name of the detection product. Turnitin appears zero times in the Academic Integrity Policy, and the IT Services URL we probed returned 404.
• Whether an AI-writing detector is switched on. The policy writes a rule about how AI detection software may be used without saying whether any is in use. Not documented either way.
• Any opt-out or alternative submission route. None. “Any work submitted for assessment may be subject to electronic or other detection procedures”, with no consent mechanism published.
• Where work submitted to detection software is stored or processed. UCD is in Ireland and inside the GDPR, but no data-residency statement is published in the sources we read. The only published control is that the software “must be approved for use by the University”. That is a procurement control, not a location guarantee, and we are not going to infer one from the jurisdiction.
• Any central student-facing self-check service. Access to similarity results is the per-module discretion described at the top of this page; nothing institution-wide was documented.
• A transcript-notation rule. Not documented in the policy or the procedure.
• Case counts. UCD maintains an Academic Misconduct Record System, named in the Procedure, but publishes no figures we could locate.

Before you submit

• Ask your Module Coordinator whether you will see your similarity result. Section 4.2 makes it their call, so it is worth asking early rather than assuming either answer.
• Read the module descriptor for the AI permission. Under §3.5 that is where it must be stated; if it is silent, §4.3 makes asking your responsibility, before you use anything.
• Acknowledge the nature and extent of any assistance— not just that AI was used — in your discipline’s citation style.
• Do not run your text through paraphrasing software to lower a score, and if you used translation software, say so and check it is permitted. Both are named in the misconduct definitions.
• Keep drafts and dated version history. UCD may review VLE and online activity records, and you may be asked to explain your work in person.
• Check your UCD email during and after any process. Decisions are issued there, and the appeal window is ten working days.

If you want a read on your own text before you hand it in, our AI check is free up to 1,500 words, with no account and no name.

Sources

• University College Dublin: Academic Integrity Policy — Policy owner: Office of the Registrar. Approval date and body: Academic Council Executive Committee, May 2024. Effective 1 September 2024. UCD Governance Document Library, document ID 274. Sections 3.1, 3.5, 4.2, 4.3, 7.
• University College Dublin: Student Academic Misconduct Procedure — Approved by Academic Council 30 April 2025. Effective from 1 September 2025. Document ID 275. Sections 1, 3.7, 5.3.6, 6.
• University College Dublin: UCD Governance Document Library — the authoritative index; the policy pages state that the Library holds the official, most recent version.

This page is orientation, not legal advice. What binds you is your module descriptor, the UCD Academic Integrity Policy and the Student Academic Misconduct Procedure — together with the Student Discipline Procedure and the Student Appeals Procedure where a case reaches them.