If you are writing a research article or if you are just trying to stay up-to-date in your research field, you need to read publications—lots of them. Of course, you know that ChatGPT can summarize these for you. But often the summaries are much too long and may lack the focus you need.

Recommendation: Provide ChatGPT with a structured template for your paper summary. This way your summaries remain short and comparable, so you can easily keep an overview of all your papers.

Prompt

Here is a prompt that you can copy-paste for a first try. This one is specifically tailored to a research publication with an empirical study. I ask for “markdown” as an output format which is just a very simple formatting (e.g. # for headlines). Of course, you should reduce or extend the prompt according to your needs. You may need to upload the PDF of the publication but the publication may also be inside the model or the model may be able to access it with its own search just on the basis of the title.

Summarize the paper [PAPER TITLE AND POSSIBLY OTHER IDENTIFYING INFORMATION LIKE FIRST AUTHOR AND YEAR] according to the following structure (use markdown)

SUMMARY

KEY POINTS

AUTHOR LIST

INSTITUTIONS

ABOUT THE PAPER (title, year, conference/journal)

STUDY DETAILS (if there was a study, data, participants etc.)

STUDY FINDINGS

LIMITATIONS

THOUGHTS

MOST IMPORTANT RELATED WORK (cite authors, year, title, venue)

IDEAS FOR FOLLOW-UP WORK / RESEARCH GAP

Results

Here is what I get when using “Your Brain on ChatGPT” as the PAPER TITLE (this is by now quite a famous paper by MIT Media Lab, so ChatGPT likely “knows” it). I didn’t need to upload the PDF and I used GPT-4o (included only in the paid version).

I will show parts of the reply I got from ChatGPT.

In the first part, the summary and key points are handy to get a quick understanding:

For an empirical study it can be tedious to pick out the key facts like number of participants, number of sessions, etc. And, of course the key findings.

I asked ChatGPT for its own thoughts at the end which encourages it to go beyond summarization and provide a critical or interpretative perspective.

Why this template?

So some of my thinking for these particular choices in the template were:

• Try to get a crisp notion what the paper is about first
• See where the paper is coming from (authors and institutions); maybe I know an author, maybe it’s a famous institution
• Then, of course, the major takeaways
• But also details, especially if it’s an empirical study
• Finally, get more interesting judgments under “Thoughts” and generate some ideas for follow-up research

I found ChatGPT to be particularly useful when used on a single paper, so to proceed paper by paper. ChatGPT is not so good finding many other materials in the internet that relate to this paper. This may change over time but for now (Aug 2025) I recommend exploiting the fact that ChatGPT can save you massive amounts of time in the reading phase.

Always check!

Of course, always make sure you really double check the information by comparing it with the original source! ChatGPT may invent facts, although in this scenario it is not so likely. But always double check with the original, never rely blindly on the AI.

You may also find it useful to run the exact same prompt on a different ChatBot like Gemini, Claude or Lumo. Compare the results. You may get valuable information and, as a side effect, get to know the pros and cons of various models.