AI in historical research

This page concerns historical research, scholarship and writing. It is not about the use of artificial intelligence in psychiatric practice.

To learn more about psychiatry in combination with AI and other emerging technologies, please see the webpages for College's Digital Special Interest Group. 

Generative AI is becoming part of the ordinary environment of research, teaching and writing. Members may use it directly through general-purpose chatbots, encounter it within search engines, office software and scholarly platforms, or meet AI-generated material in the work of students, colleagues and published scholarship. 

These technologies are developing rapidly and offer substantial possibilities, but their implications for scholarly practice are only beginning to be worked through.

The resources signposted here are intended to help with this and provide reference points for appropriate practice and procedure.

 

Orientation

If you have not used these systems, or only used them in passing, begin here. Both resources are non-commercial, written for readers without a technical background, and revised regularly.

Use in historical research

Examples of working with sources

  •  Mark Humphries — Generative History: A working historian documenting experiments with AI-assisted transcription and analysis in detail, including open-source tools and a benchmark for model performance on historical handwriting.
  • Transkribus: A platform for recognising, transcribing and searching handwritten and printed historical documents. It's maintained by the member-owned European cooperative Recognition and Enrichment of Archival Documents (READ-COOP).

Uses in all academic work

Drafting and editing, literature searching, responding to reviewers, preparing funding applications and undertaking peer review are not particular to history. Most publishers and funders have explicit policies, chiefly concerning disclosure, authorship, accountability and confidentiality. The relevant policy should be checked for each submission.

  • Committee on Publication Ethics — Authorship and AI tools: The standard publication-ethics position is that AI tools cannot be authors because they cannot accept responsibility, declare conflicts or enter legal agreements. Human authors remain responsible for the work and must disclose how AI tools were used.
  • Cambridge University Press — publishing ethics: AI contributions to research content: Cambridge publishes the College’s five journals and books on the College’s behalf, making its ethics guidance directly relevant to College publishing. AI use must be declared and explained; AI cannot be an author; and human authors remain responsible for accuracy, integrity and originality. Individual journals may impose additional requirements.
  • UK Research and Innovation — use of generative AI in application preparation and assessment: Applies to all UKRI funding opportunities and sets responsibilities for applicants, assessors and UKRI. Applicants are expected to be transparent about relevant use. Since 6 July 2026, assessors may use generative AI only for language refinement: no application content or personal information may be entered, and the tool must not be asked to understand, summarise or evaluate the application.
  • Research Funders Policy Group — joint statement on generative AI in funding applications and assessment: A high-level 2023 statement from eight major UK science and health funders, setting common expectations while individual policies were developed. It identifies risks involving rigour, transparency, originality, confidentiality, intellectual property, copyright and bias; current guidance from the particular funder should also be checked.
  • Jisc — Maintaining academic integrity and AI detection: Jisc is the UK’s not-for-profit digital, data and technology organisation for tertiary education and research. Its National Centre for AI provides practical guidance and resources for universities, colleges and researchers.
  • Academ-AI — suspected undeclared AI use in academic publishing: An independently maintained catalogue of suspected undeclared AI use in journals, conference proceedings, books and theses, identified chiefly through conspicuous chatbot phrasing. An accompanying preprint analyses the first 768 entries; because cases are selected for visible textual traces, the collection demonstrates occurrence rather than prevalence.

What this list does not cover

Historical research - There is currently no authoritative, discipline-wide guidance defining competent AI use in the archive or in historical argument. The AHA committee above may begin to fill this gap.  For now these resources are reference points rather than a settled code.

Environmental, labour and copyright costs - These all raise questions and are not treated in any detail here. The following resource provides a practical entry point on environmental sustainability.

Evidence on actual practice - Most large UK surveys concern student use. Evidence about how historians and other researchers are incorporating generative AI into research workflows remains much thinner.

 

Contributions and revision

This page is maintained by the Special Interest Group’s AI Working Party and will be revised as the field develops. Suggestions for additions, corrections, and accounts of members’ own practice are welcome and should be sent to sigs@rcpsych.ac.uk.

Last reviewed: 16 October 2026.