This post is part of a series commissioned by the Royal Historical Society on ‘AI and History Teaching in UK Higher Education’ in Autumn 2026.
The series includes a range of perspectives on the uses and limitations of Generative AI and LLMs from history educators and research academics.
Participating authors provide ideas and suggestions for teaching practices shaped by their own scholarship and teaching experiences in contribution to the ongoing conversation about the impacts of AI on teaching and learning history at UK universities.
By Dr Miguel Hernández
It is a cruel irony that just as we emerged from the COVID-19 lockdowns, our discipline would be faced with a new challenge to overcome. Generative Artificial Intelligence (GAI) software threatens to upend our approach to historical research and teaching like no other previous technology.
Despite prevailing and urgent fears regarding the environmental and social impacts of GAI, as well as ethical concerns about attribution and copyright, the decision by UK Higher Education institutions to embrace the potential of this technology has been made, and academics have been left to develop teaching and assessments that can incorporate GAI. The almost overnight impact of GAI on universities and scholarship has meant that necessary research on its potential is still ongoing, but for many of us the most sudden effect has been our students’ use and misuse of GAI for their studies. A recent 2026 survey of UK undergraduates found that 95% of students employed GAI software in the past year for their work and in their daily life, with text-generators like ChatGPT and writing software like Grammarly being among the most popular. This prevalence has grown from previous years, a development that cannot be ignored.
As historians, it is imperative that we introduce our students to responsible GAI use. Even those who are not actively seeking to employ it will come across it on search engines like Google or integrated into eBook platforms like ProQuest. GAI’s features – its tendency to respond in a positive fashion to user prompts, its ability to condense complex issues into bitesize summaries, and its inherent functionality to regurgitate the most popular elements of a topic – are all significant concerns that require users to query and question the information provided.
We wanted to go a step further and institute a measure that required students to be cognizant of their use or reliance on GAI at regular intervals throughout their degree via our department’s GAI policy.
Last year, the Royal Historical Society funded a project led by David Clayton, Henrice Altink and Esther Wilson at the University of York for which undergraduate history students were asked to assess critically the responses Gemini produced when it was asked to undertake regular tasks such as preparing answers to seminar questions or developing and refining ideas for a dissertation. The project’s findings suggest that students benefitted from such activities, as they realized the responses produced by this GAI were ‘underwhelming’. The project recommended integrating such practical exercises into teaching practice to make students aware of the capabilities and limitations of GAI. This has certainly been a valuable exercise for my department and our students at Aberystwyth University, but we wanted to go a step further and institute a measure that required students to be cognizant of their use or reliance on GAI at regular intervals throughout their degree via our department’s GAI policy.
A recent study by Sam Illingworth for the Higher Education Policy Institute has established there is significant variance in GAI policies across UK universities and is critical of how these policies often employ the language of surveillance rather than education. In particular, the author frames mandated requirements that students state their use of GAI or retain evidence of prompts and GAI input as problematic, as it creates a relationship built on conditional trust wherein the student is only trusted if they make their GAI use explicit. While Illingworth is correct that GAI policies should do more than provide means for detection of improper use, accurate attribution remains a core aspect of our work as historians.
Whether it is referencing evidence in footnotes or acknowledging the contributions of GAI, students need to understand that attribution is not an afterthought, but a key aspect of the process that allows others to understand what is original about their work. Furthermore, including discussions of penalties and disciplinary procedures for misuse in said policies is necessary considering the student conduct framework in which we operate.
Like many other historians, my colleagues and I had to hurriedly put together a GAI policy for our department that conformed to the wider University’s, but which also served the specific needs of our discipline. At first, our concerns were about rooting out improper GAI use, but we soon realized that this should not be the primary function of our policy. We asked our students to submit their assessments with a template that required a statement acknowledging whether and how GAI was used in the assessment. Initially, we discussed whether we should consider the absence of such a statement to be a form of unacceptable academic practice. However, we appreciated that students often forgot to include a template with their assessments and that treating this oversight in the same category as plagiarism would be draconian. We instead instituted a system where these students would be contacted while assessments were being marked and asked to produce the GAI statement if it had been left out. We explained that their marks would be withheld until the statement had been submitted.
We considered outlining prohibited uses of GAI in our department’s policy but decided against this since we felt it would simply drive such uses underground. Instead, we created two categories of potential use divided into ‘Fair Use’ and ‘Permitted but Inadvisable Uses’.
In the ‘Fair Use’ category, we listed approved uses, and alongside each item we provided examples that specified how best to employ the software. For instance, we included language and writing improvements in this category, but reminded students that GAI suggestions were the equivalent of employing a thesaurus, and that they should critically engage with GAI contributions so that they can both improve their writing and maintain their independent style.
In the ‘Permitted but Inadvisable Uses for GAI’ category we included such elements as using GAI to provide comment on the structure of their essays, where we explained that structure is a critical aspect of their independent argumentation, and an area where relying on such software would likely harm their work as GAI suggestions would likely flatten their argumentation and impede their ability to develop critical skills independently. Of course, our GAI policy also required that students submit a statement discussing their use of the software. To make students more cognizant of how they were engaging with and using GAI, their statement is required to include details about all the separate functions they have employed and the specific software being utilized. So, for example, a student that had employed ChatGPT to research appropriate readings and Grammarly to improve their writing would have to provide details about both uses separately.
Nonetheless, like other historians and disciplines, we have also tried to introduce assessments that are less vulnerable to GAI misuse. In my Special Subject, I have found that short 15-minute presentations where students analyse sources they have selected themselves, followed by a short Q&A, work quite well. Because the sources are fairly specific, there is relatively little written about them and GAI cannot scrape enough information to provide an insightful analysis that students can rely on. The Q&A portion assists further, as the students cannot consult GAI during this exercise. Although I have found this assessment effective in several aspects beyond its capacity to deal with GAI misuse, it only works because this is a module with capped numbers, and even so it takes most of a day to undertake. Such an assessment is unfeasible in a larger survey course or in first year when students are still getting to grips with basic skills of the discipline. We have also considered returning to in-person examinations to ensure no GAI could be employed. However, we decided against this as both students and staff prefer take-home examinations as it assesses a wider range of skills.
The essay must remain a critical part of our assessment portfolio … Instead of doing away with the essay, in my modules I have decided to keep them as is, but to ensure that the remaining assessment is more resistant to GAI misuse.
At a time when our discipline is under significant stress because of the crisis in UK higher education, changing all our assessments presents significant challenges to our ever-shrinking time and resources. Many of us have groaned at calls to do away with the essay as the standard form of assessment in history because of its vulnerability to GAI misuse.
Yet the essay must remain a critical part of our assessment portfolio. Not only does it test all the skills that are indispensable to budding historians, it is also a critical formative practice for what is generally a student’s crowning achievement: their dissertation. Instead of doing away with the essay, in my modules I have decided to keep them as is, but to ensure that the remaining assessment is more resistant to GAI misuse. With this format, the students get to employ GAI to their chosen extent in the essay, but they are warned (and seem to agree) that relying on it will prevent them from developing the skills they need for the remaining assessments. In my Special Subject, this is the aforementioned presentation, but on larger survey courses I have found that a source analysis exam can be effective. This exam has a series of related yet contrasting sources grouped under themes with no question or further guidance beyond the title of the theme. For example, on my survey course on the Reconstruction Era in the United States, one of the themes was ‘Emancipation’, and the sources were a mix of documents from Black men and women at that time and later reflecting on their experiences with their newly gained status.
Without a question or an indication of what they need to focus on, the students who relied on improper GAI uses could only use it summarise the sources, leading to poor marks. This exam format has also led to much better exam scripts than those with standard essay questions, as the lack of guidance tends to produce more original and interesting responses from the students.
And yet, the arms race continues to escalate between staff developing new policies and assessments and the GAI software students employ.
As this post indicates, we’ve continuously had to adjust our policy and assessments and consider new research and the responses from students. This is a process that feels untenable in the long term. Last summer, one of our students correctly filled out their GAI statement, noting they had employed the software for a source analysis exam. The student felt that the technology had advanced to a point where it could highlight general aspects for discussion, and the student could then frame their own work around the GAI generated suggested points. Fortunately, this student’s exam did not ask them to analyse several sources in conjunction so I do not yet have to change the exam format I mentioned earlier, but it is a telling example of how we will need to continually test what GAI can do and adapt.
A colleague and I had a fruitful conversation with the student about their process. The insights they provided have helped us redevelop our GAI policy for this academic year, and highlighted how important engaging our students as partners in these processes can be. Although like many I continue to see GAI as a threat to more than just the discipline of history, working alongside our students in a less adversarial fashion can help us and them align expectations and set standards that permit uses without detracting from developing the key skills expected of historians and needed in their future careers.
About the Author
Dr Miguel Hernández is a Lecturer in American History at Aberystwyth University.
Miguel’s research focuses on the history of organised white supremacist movements and the broader history of racism in the modern United States.
Featured image credit: Wikimedia Commons.
