What's great
I work in clinical research and used SciSpace (PubGenius Inc.), which promotes itself as an AI assistant for improving scientific manuscripts, particularly in grammar and accurate bibliography generation. On February 9, 2026, I submitted a manuscript for review and was required to purchase an annual subscription to access the generated files.
After paying, I found that no meaningful improvements had been made to the text. More seriously, many of the references provided appeared to be fabricated or incorrect, including nonexistent citations, wrong co-authors, inaccurate journal details, and even a reference generated under my own name. The service delivered does not match what is advertised, and the inclusion of unreliable references represents a serious concern for scientific work.
What needs improvement
The company should acknowledge that the tool is not yet ready for universal use and that it makes mistakes. They should further improve the citation search function, strengthen their after-sales support, and take better care of their subscribers. They should also adopt a self-critical approach and carefully analyze what they are not executing properly. Not only the company, but all subscribers would benefit from improvements to the product and from more attentive customer care.
vs Alternatives
I contacted the company to inform them of the errors and to request a refund. Regarding the first issue, there was no response at all; I believe no one reviewed the original documents I submitted or the file generated by the AI tool. As for the refund, it was a tremendous disappointment.
How helpful is the multi-agent literature review feature?
That is the worst part of all: about a third of the references do not exist. Of the 24 references, one was generated under my own name: the co-authors were incorrect, the article title does not exist, and the citation details were fabricated; even the page range did not correspond to any other article in the journal it was supposedly referring to.
How well does it break language barriers for papers?
I am not a native English speaker. I write fairly well, but I need the tool to refine my text and improve the grammatical phrasing. I was surprised to see that the revision carried out by SciSpace practically did not modify my text at all.
What limits exist on document size or upload count?
I did not find any limits