In my little bits of nineteenth century historical research I use digitised newspaper resources a lot.
The various digitised resources I use most often are nineteenth century Scottish newspapers via the SLV's subscription to Gale Newsvault for family history stuff, The Times of London's archives again through the SLV, Welsh Newspapers Online, Papers Past NZ, and above all, the NLA's Trove.
Trove is undoubtedly a great resource, but the quality of the digitised text, to put it politely, is variable.
Trove does provide OCR's summaries of the articles, but the quality of the digitised text can make the OCR'd text read as if it had been transcribed by a Martian - strange combinations of letters and punctuation followed by gobbets of reasonable text.
So, for years, what I have done is use the download option to generate a pdf, download the pdf to an ipad, and then sit and make notes on a 'proper' computer.
Latterly, if the pdf is too hard on the human eyeball, I've used J's old iMac, which now runs Linux, and Okular to give me a bigger image at a decent resolution to work with, and that's worked pretty well as a workflow.
Now, as I'm sure you're aware if you're an Acrobat user, Acrobat now behaves like an enthusiastic puppy, always asking if you want it to generate a summary of the document.
I've tended to ignore it, really because most of the PDF documents I look at on windows are boring things like credit card and electricity account statements, and there's usually only two important bits of information - how much we owe and when is payment due.
But instead of doing the majority of my work on a linux machine as I usually do, I researched the Panjdeh incident on my Windows machine, and typed my notes into Geany on the old Chromebook I installed Linux on, really as a way of assessing the usefulness of the converted Chromebook.
(Answer, very useful, and good battery life to boot).
Anyway, as I was working on Windows, Acrobat came along wagging its little tail, offering to generate a summary of every pdf document I opened.
So, for a number of longer documents, including some with poor quality OCR'd text, I did.
And they were surprisingly good, and the AI summary engine seemed to deal reasonably well with poorer quality scanned text, producing reasonable and good quality précis of the article texts.
Obviously you need to check the text yourself, but using AI text summaries turned out to be a useful way of assessing if the article was worth reading, it's not the first time I've slogged through a report of court proceedings to find that the report didn't add anything to what I already knew.
It's by no means a panacea, but it's certainly a valuable tool...