ChatGPT is already able to help medical researchers and scientists to write articles and abstracts, in literature research, to summarize data or information, to provide suggestions for structure, references, and titles, in language reviews to make the text more readable, or even to generate a full draft of a paper [2]. At this stage, although feasible, no article dealing with critical care medicine has been published so far using this application. Moreover, whether writing a manuscript dealing with a topic of critical care medicine would be easier or more difficult when compared to other medical disciplines remains unknown. Which are the limits of possible applications of these tools have to be clarified.

ChatGPT can generate text on a wide variety of topics. Recently, an entire letter to the editor was written by ChatGPT, answering the author's questions [1]. Artificial Intelligence cannot generate new ideas, but it can organize and develop the ones of the researcher, creating an initial draft. However, this appears to be a starting point for a human-based development of the text, as the automatic text generated is far from being a substitute for the knowledge, creativity, and critical thinking of human experts.

For the literature search, ChatGPT and AI research assistants (such as “elicit.org”) can aid researchers in the review process by finding academic papers, summarizing their conclusions, and highlighting areas of uncertainty (i.e., summarizing the recent randomized clinical trial on the use of extracorporeal membrane oxygenation for refractory cardiac arrest [3], Fig. 1). This could help physicians, as an example, to quickly gain an understanding of the current state of knowledge on a particular topic, and to identify potential gaps that need to be addressed, although the provided summary might be quite generic and do not critically analyze differences among studies.

Example of a summary of the existing literature on a specific topic

During the writing process, ChatGPT can help to generate an initial draft of a scientific paper and even suggest titles. By providing raw information, ChatGPT can also assist in composing the section on the methods used in the study, justify sample size and describe data analysis techniques. From author’s experience, when the manuscript has been finalized, ChatGPT is extremely effective for the editing process; formatting and language editing, rewriting a particularly complex sentence in a clearer way, and even summarizing the entire text in order to compose a suitable abstract, are feasible using this approach, although the results are not always satisfactory, but they certainly save time. ChatGPT has been recently asked to prepare regular discharge summaries [4], but the quality of the content has not been specifically evaluated. To our knowledge, a structured evaluation to determine the quality of the output has not been performed yet. More complicated writing processes, such as systematic review and meta-analyses, require the human intervention and ChatGPT might be helpful mainly for editing.

A future potential application of AI is in the automatic generation of figures, tables, and other visual elements of the manuscript, which can aid in summarizing data. These elements are important for the clarity and understanding of the manuscript, but they are often time-consuming to create.

Importantly, the writing process of a scientific paper requires, for the moment, the guidance and supervision of human researchers who are experts in the field to ensure the accuracy, coherence, and credibility of the content before being used or submitted for publication. Chatbots can help but needs the researcher’s input, and inadequate inputs would lead to inadequate results. For this reason, chatbots and AI, in general, should not replace human researchers' expertise, judgment, personality, and—in the end—responsibility.