In general, generative AI such as ChatGPT should not be cited as a source because the generated text is neither reliable nor possible to reproduce. In some cases, however, it may be appropriate to reference generative AI, for example if you want to present an AI-generated response to discuss it or the technology behind it.
Citing or reproducing texts created by ChatGPT or other generative AI in your work cannot be done in the same way as other texts. The purpose of referencing is to allow the reader to find the original source on which the text is based. Texts created using generative AI generally cannot be reproduced because a unique text is generated each time. Even if the reader were to input the exact same prompt into the same AI tool, the result would not be the same.
There is currently no established practice for referencing ChatGPT or other generative AI in the Harvard style. Different solutions exist. The suggestion below follows the format for referencing software. Here the text generated by the AI model is treated as an output of an algorithm, and instead of referencing the generated text, you refer to the creator of the algorithm in question.
In addition to a reference to the AI model, you should describe in the text how you used it, including the prompt you used. The generated text can be reproduced in the text if it is short enough or attached as an appendix at the end of the work.
Always check with your teacher or supervisor if any specific instructions apply to you and your work.
... (Developer/Organisation, Year).
Developer's Last name, First name, Developer's Last name, First name and Developer's Last name, First name (Year). Name of AI model, Version x.x. Place of publication: Publisher. URL/DOI