AI marking

In addition to multiple choice questions that are marked automatically, most questions in Tasks requiring a text input from the student can now be marked automatically too.

When setting a task the teacher has four marking options:

  1. Marked by teacher.
  2. Marked by AI.
  3. Self assessment.
  4. Peer assessment.

This can be changed at any time, even after the task has been set or after the marking process has already begun.

After a student submits their task if AI marking is selected it is automatically sent for marking even if other students are yet to submit their work, reducing the server response time.

Our bespoke language model

Smart Revise uses its own large language model (LLM) provided by OpenAI that is fine tuned to the Smart Revise context. It is not ChatGPT.

Teachers can monitor the progress of AI marking in the Active Tasks screen. Responses are usually returned within a few minutes, but it depnds on the size of the marking queue.

AI marking progress

Artificial Intelligence (AI) in education

"AI will have the power to transform a teacher’s day-to-day work."

"When used appropriately technology has the potential to reduce workload and free up teachers’ time. We [DfE] recognise that teacher workload is an important issue and are committed to helping teachers spend less time on non-pupil facing activities, including through use of generative AI." - Gillian Keegan, Education Secretary, March 2023.

AI marking of tasks significantly reduces a teacher's workload and can create discussions in class about the accuracy of the marking. This causes the student to reflect on their answer and the AI response, which is a valuable learning experience even if the marking is not completely correct.

Remember that research has shown that self and peer assessment are still extremely valuable in raising attainment. Teachers are advised not to be overly reliant on AI marking.

Generative atificial intelligence in education. Department statement (2023), Department for Education.

Education Secretary addresses BETT 2023.


Once marking is complete, students and teachers can review the marking. AI will award a mark for each question and provide a few sentences of written feedback justifying the mark awarded.

AI response

There are additional precautions to ensure that the feedback does not contain inappropriate responses to students.

For the initial launch, AI will not mark calculate command words, questions containing images or answers requiring a level of response framework to be applied.

As the number of questions sent to the LLM grows, the model will be tuned to provide more accurate responses. This will include alternative models for different command words.