How Skilled is ChatGPT?
16 May 2025
Recently, the European Patent Office (EPO) in case T1193/23 concluded that ChatGPT is not a skilled person. This may be unsurprising to followers AI and patent news -, after all, the well-reported 2023 UK Supreme Court DABUS case found that an AI cannot be an inventor because it is not a person. However, the ‘skilled person’ is not actually a real person either…
Who/what is the skilled person?
The skilled person is a legal fiction (or a tool) which may assist patent examiners (and judges) in deciding if an invention is inventive. Case law has established that this fictional entity embodies an experienced practitioner who has average knowledge and abilities. For this article, it is also important to know that the skilled person:
- has common general knowledge in the field of the invention (and in some cases, neighbouring fields), before the invention was applied for at a patent office; and
- possesses no inventive capability.
The use of this skilled person ‘tool’ enables patent examiners to set a baseline when beginning their assessment of inventiveness. It can help with interpreting terms of a patent claim (e.g., a ‘load’ would be interpreted differently by a civil engineer and an electrical engineer), when deciding if two documents are combinable, and when determining the common general knowledge.
What does the EPO think?
In contentious patent cases, the meaning of claim language is often closely scrutinised. In the recently issued decision by the EPO’s Board of Appeal T1193/23 (relating to a spinning machine for producing yarns), the patent applicant argued that [at least] the term “position control” had a well-defined meaning to the skilled person in the art. They supported this by questioning “ChatGPT” about the meaning of the term “position control”, which confirmed their interpretation.
It may be reasonable to assume ChatGPT has access to information in many technical fields and possesses no inventive capability, similar to the skilled person. However, as the Board of Appeal in T1193/23 noted:
ChatGPT’s answer is irrelevant in itself, since the interpretation of the claim is about the understanding of the skilled person (see also T 206/22, Reasons 1.). The general increase in the prevalence and use of chatbots based on language models and/or “artificial intelligence” alone does not justify the assumption that an answer received – which is based on training data unknown to the user and may also depend sensitively on the context and precise formulation of the question(s) – necessarily correctly reflects the expert’s understanding of the respective technical field (at the relevant time).
That is, when using ChatGPT there is no way to scrutinise if it is using information outside of the technical field (or neighbouring fields) of the skilled person, and/or when it acquired that information (for example, if it was acquired before or after the patent application was filed). Thus, ChatGPT cannot be equivalent to ‘the skilled person’.
Lessons learned
This decision indicates that the EPO may not be tolerant of ChatGPT serving as a primary source to evidence the knowledge of the person skilled in the art. As the Board of Appeal say, “ChatGPT’s answer is irrelevant in itself”. However, ChatGPT might still be useful when searching for evidence of common general knowledge (if properly checked), such as the definitions or interpretations of terms in textbooks or other literature which could be used as evidence before the EPO to support the interpretations of terms.
In conclusion, ChatGPT is not a skilled person, but if appropriate, could be wielded skilfully by the person.
Theo Worsley
Electronics, Computing & Physics group
This publication is a general summary of the law. It should not replace legal advice tailored to your specific circumstances.
© Withers & Rogers LLP May 2025
