Teachers’ use of ChatGPT in lesson preparation can save time

HELEN POET, RESEARCH DIRECTOR, NFER, UK

KATHERINE ASTON, RESEARCH MANAGER, NFER, UK

Introduction

Since the release of various generative artificial intelligence (GenAI) models for public use, there has been great interest in its potential use in education. By August last year, 57 per cent of teachers reported using at least one form of AI tool to help them with school work, according to Teacher Tapp (Hallahan, 2024).

In December 2023, the National Foundation for Educational Research (NFER) was commissioned by the Education Endowment Foundation (EEF) and the Hg Foundation to conduct a randomised controlled trial of the use of ChatGPT by teachers when preparing for lessons (Roy et al., 2024).

Our aim was to provide useful evidence for the education sector on the impact of using ChatGPT on teacher workload, which we know is a particular challenge for the teaching profession compared to other similar graduates (McLean et al., 2024), with implications for retaining teachers within the profession (Lynch et al., 2016).

Our project

During 2024, we recruited 259 teachers of Year 7 and/or Year 8 science from across 68 schools to take part in the project. Almost all our teachers were science specialists, with an average of nine years of teaching experience. Before the trial started, very few of the participating teachers typically used GenAI when planning lessons.

Schools were randomly allocated to either using ChatGPT or continuing the use of their usual approaches and avoiding the use of any GenAI tools (see Figure 1). Teachers were asked to continue with the activities that they would usually complete when preparing for lessons (either using ChatGPT or without GenAI) and not to complete any additional lesson preparation to what they would do typically. Therefore, where teachers already had materials prepared, they did not need to redo or update them, unless they were planning to anyway. While we asked those in the ChatGPT group to use it as much as possible, they were free to use it as much or as little as suited them.

Teachers in the ChatGPT group were sent a guide to using ChatGPT, which had been developed by Bain & Company (https://teachingwithchatgpt.org.uk).

They followed these approaches for 10 weeks during the 2024 summer term and recorded the time that they spent preparing for lessons and the amount of use of ChatGPT or other GenAI in an online diary each week.

We also collected examples of the lesson resources that teachers had produced during the trial. These resources were ranked by a panel of science teachers, who were not aware of the purpose of the project. These ranks were then analysed by the research team to check whether there was a difference in quality between the two groups.

Finally, we collected feedback from participating teachers through surveys and interviews about their experience.

Figure 1: Format of the randomised controlled trial for the use of ChatGPT in lesson planning

Results

Saving time

Teachers in the ChatGPT group saved time – around 25 minutes per week on average (see Figure 2). This is the equivalent of a 31 per cent reduction in lesson planning time, on average, compared to the teachers asked not to use any GenAI. As the schools (and therefore teachers) were randomly allocated to the groups, we can conclude that using ChatGPT led to a reduction in teacher workload. 

Figure 2: Time spent on lesson planning in the two groups

The saving of 25 minutes per week needs to be considered in the context that teachers only used ChatGPT for a relatively modest amount of their lesson planning workload. We only collected data on workload in relation to Year 7 and Year 8 science lessons, which for the average teacher in our trial represented about five hours of teaching per week. Furthermore, teachers typically used ChatGPT for only around a third of these lessons. Despite this, teachers in the ChatGPT group still saved time compared with the teachers in the group who were asked not to use any GenAI. We found that teachers used ChatGPT as another tool at their disposal, selecting it as appropriate but not at the sacrifice of other established sources or approaches. For example, many science departments have extensive pre-prepared resources, including lesson slides, and science has a wide range of external curriculum resources on which teachers can draw. 

We also found that teachers that used ChatGPT for more of their lessons and scored highly on a knowledge quiz about how to use ChatGPT effectively spent less time preparing for lessons than those that didn’t.

Teachers reported using time saved for other teaching tasks, but others chose to reduce their overall workload.

No difference in quality of resources

The analysis showed that there was no difference in the quality of the resources produced by the ChatGPT and non-GenAI groups. However, there is some caution around this finding. We didn’t receive as many resources as we had hoped for so, while we were able to conduct this analysis, it is based on a limited sample. As teachers were able to select which resources they sent us, it is likely that they sent us those with which they were happiest, which may have introduced some bias to this result.

Teachers reported benefits of using ChatGPT and intended to continue using it after the trial

Three-quarters of the ChatGPT group felt that using ChatGPT had a positive impact on their teaching (Figure 3). Almost half said that it had helped them to teach more creatively or tailor their teaching to pupils; however, similar proportions were ambivalent on these points.

Figure 3: Perceived benefits of ChatGPT for lesson preparation

Examples of ways in which science teachers found ChatGPT useful included:

• generating new ideas for teaching, including creating quizzes and question-and-answer activities
• tailoring a lesson, or elements of a lesson, for a different class more quickly, using prompts detailing year group, ability (e.g. reading age) and prior learning
• adapting existing resources for other teachers to cover a class, or creating relevant activities that could supplement a cover lesson (e.g. in place of a practical).

Over three-quarters (78 per cent) of teachers in the ChatGPT group said that they would continue to use ChatGPT to support lesson preparation in the future – but typically only for some lessons.

Implications for practice

Teachers who were more confident in using technology saved more time using ChatGPT. Therefore, teachers who are less confident or experienced at using technology may need additional support to make the most out of AI.

Teachers who were less confident in their subject matter (this could include non-specialists and those new to teaching a subject or topic) also saved more time. However, a key consideration is that AI-generated outputs need to be checked for accuracy, which may be a bigger task for these teachers.

The ways in which teachers in the trial most commonly used ChatGPT reflected the most common planning tasks for Key Stage 3 science: creating questions and quizzes, finding activity ideas and adapting materials. Other uses recommended by the teacher guide, such as creating model answers or worked examples, may be more relevant for other subjects.

It was rare for teachers in the trial to refine their prompts in order to ask ChatGPT to tweak the output that it had provided, but this is one of the ways in which users can get more out of ChatGPT. Experimenting with the tool to learn how to craft prompts well, and asking ChatGPT to amend what it generated, made a big difference to the outputs. For example, teachers who successfully adapted materials using ChatGPT typically specified a reading age, as the default writing style of ChatGPT was too dense for many lower secondary pupils.

Sharing effective prompts or ways of interacting with ChatGPT helped to save time. Teachers who are less familiar with using ChatGPT and other Gen-AI tools may benefit from peer support or training to make the most of them.

Teachers developed their own approaches to using ChatGPT. As expected, some teachers were focused on saving time – outsourcing time-consuming activities that they found that ChatGPT could do well. Some teachers were focused on tailoring resources for their pupils – for example, by generating question sets on an area where pupils needed further practice or simplifying explanations and texts. As ChatGPT could create these resources quickly, teachers could provide a more bespoke lesson with similar planning time. Other teachers focused on using ChatGPT to refresh their lesson content – for example, by generating activity ideas.

Limitations of using ChatGPT and other GenAI for teaching

Everyone using a GenAI tool needs to be aware that what they enter into ChatGPT and other AI tools (e.g. as part of the prompts or questions) is usually saved by the tool and may be used to train the models and inform responses given by AI to future users. Therefore, it is essential that users do not enter personal details of students or teachers. For this reason, using ChatGPT and other GenAI is not recommended for writing reports.

One of the challenges found by teachers during the trial period was limitations of ChatGPT at the time (summer 2024) in creating images and diagrams, which can be important for conveying scientific concepts. This element of GenAI is developing, but other sources may continue to be better in terms of accuracy and clarity, at least in the short term.

Another limitation of GenAI is that all outputs need to be checked for accuracy. ChatGPT and other GenAI models acknowledge that they can make mistakes, including ‘hallucinating facts’ where information is missing, so time for checking content always needs to be considered.

For the full study report, see Roy et al. (2024).

The examples of AI use and specific tools in this article are for context only. They do not imply endorsement or recommendation of any particular tool or approach by the Department for EducationThe ministerial department responsible for children’s services and education in England or the Chartered College of Teaching and any views stated are those of the individual. Any use of AI also needs to be carefully planned, and what is appropriate in one setting may not be elsewhere. You should always follow the DfE’s Generative AI In Education policy position and product safety expectations in addition to aligning any AI use with the DfE’s latest Keeping Children Safe in Education guidance. You can also find teacher and leader toolkits on gov.uk .