University College London (UCL)

UCL Library guidance requires acknowledgement where GenAI has functionally contributed to the creation of academic work (e.g. drafting, planning, structuring), subject to discipline and assessment instructions. Minimum acknowledgement includes tool name and version, publisher, URL, and single-sentence context of use. Prompts, generation dates, and tracked-changes are additional requirements that apply only when stipulated by departments or assignment briefs. Preparatory reading, search summaries, and basic grammar assistance may not require acknowledgement. While UCL advises treating AI output as a work with no author, students and researchers must follow required departmental guidelines and standardised referencing styles (such as APA or Chicago), which may require citing the GenAI tool as an author.

For an Investigatory Viva, the Secretary must supply the invitation, report and evidence at least five working days beforehand. Attendance is mandatory and the viva cannot be held in the student’s absence. Organisation should occur within 20 working days of the Stage 1 outcome unless reasonable circumstances prevent this. Stage 3 appeals must be submitted within 10 working days of receiving formal notification. Outcomes are normally issued within 20 working days after eligibility notification, with recommendations subject to Exam Board confirmation.

CLIE oral and invigilated in-person assessments are Category 1 unless otherwise stated; non-invigilated written assessments are Category 2 with assistive use only. CLIE Modern Foreign Language modules prohibit language and writing review. Summer School rules are set by the teaching department. CLIE requires acknowledgement of GenAI use.

UCL does not invest in GenAI detection software. Staff must not submit student work into GenAI detection tools or GenAI services themselves, as this compromises students' intellectual property and personal data rights. Furthermore, due to lack of evidence that GenAI can mark effectively and without bias, staff should not use GenAI to mark student work. UCL markers do not use AI detector scores to establish academic misconduct. Initial informal discussions do not replace the formal Student Academic Misconduct Procedure; suspected inappropriate GenAI use or authorship issues may proceed through Stage 1 and an Investigatory Viva.

The March 2024 UCL Copilot announcement describes protected prompts, no saved chat and no model training when signed in with UCL credentials. These are dated source statements, not verification of current tenant settings. It advises staff not to enter personal, confidential or special-category data and to use content for which they have copyright.

Doctoral candidates must list functional external contributions beyond their supervisory team in the Doctoral Candidate Thesis Declaration Form. GenAI drafting, including copyediting rather than proofreading, must be declared. Candidates retain responsibility and critical oversight, and examiners must be able to confirm satisfactory viva answers without AI support.

UCL Press permits certain responsible supporting uses of GenAI in book preparation (brainstorming, language improvement, structure), subject to human oversight, accountability, and disclosure via the Gen AI Disclosure Form. GenAI tools cannot be credited as authors on research-based works. For research monographs, UCL Press does not permit GenAI creation or manipulation of images, figures, charts, tables, or original research data. Editors and peer reviewers shouldn't upload proposals or manuscripts into GenAI tools (advisory guidance), in contrast to the strict mandatory ban for UCL Press journals.

UCL Press journals prohibit listing AI as an author on research-based works. Broadly, AI use in creating, editing, or manipulating images or figures is not accepted; however, an explicit exception permits AI-assisted technologies used as part of research design or methodology (e.g. data visualisation to generate or interpret underlying data), provided it is fully declared and documented in the methodology. Editors and peer reviewers are strictly prohibited from submitting manuscripts or data reported in them through AI tools.