@ruedigergad @carolannie @annaleen I expects it will be abysmal. I tried on several very well known, very documented short texts (article, short essay). Really, things like "the tyranny of structurlessness". The AI (all of the common ones I tested: Claude, ChatGPT, Gemini, local ollama/deepseek...) gave extremely deep sounding and assured summaries, which were also horrible results, missing extremely important, often foundational nuances to the point of misinterpretation.
For things they must have been trained on.
So, AI (llm) summary is the same silver bullet than llm for defending your case in court. Worse than useless
@ruedigergad @carolannie @annaleen I haven't tried the one in Acrobat, but generally ChatGPT or Gemini do a good work at summarising. I don't think hallucinations are a big problem honestly (for summarisation , they are a big problem for other things) I maybe haven't used them extensively but I don't recall seeing any obvious errors when summarising a document. It's the critical evaluation that is probably a bit more limited, but again if you want a quick summary, probably you're not looking for that either, and the tool does the job.
Honestly, aside fron the initial "oh that's cool" novelty-item moment I would say that mostly it's faster and more efficient to quickly scan papers manually yourself but there are other uses.