Of course, it doesn't produce research, but Paperguide is a valuable tool to explore, give ideas, propose insights, and also help get rid of the white page syndrome.
Results are really not bad both for paper exploration and for the writer. The writer works like a good non-expert and remains quite relevant within the scope of the question asked, and the syllabus alone may be very useful.
Is it perfect? No, but it's sufficiently powerful to give Paperguide the function of a... guide.

As far as my testing goes, I have not found big mistakes in my field, just some wrong references, but that's the case with all AI literature scanning, so double-checking remains needed, in that case it's very easy to do with a mouse over to check if there's a DOI.
Data extraction is also a pretty cool idea for a fast paper review.
All AI tools available for researchers have different strengths and weaknesses; there is definitely no one tool for every use, so Paperguide is useful in its way (quite different for me from Afforai, for example, or Scispace).

the UI/UX is quite good, nothing amazing but everything is straightforward, obvious and fast, so it fits the use. Link to papers, papers availability, or opensource archives could be improved, it sometimes open nothing (yes like SciSpace if you ask). DOI links seems to work better.

The support has been fast and reliable for me, which is a very good point for such tools that proliferate without being properly maintained and supported.

Use cases:

- For a senior researcher, it helps dig into new subjects and gives ideas for paper writing.
- For a young researcher, well, the writer can be, if properly used, quite powerful ;)

In a near futur, the issue of students home essays will really have to be discussed !