Scite: a tool to find out if a scientific paper has been supported or contradicted since its publication

Originally published at:


Its an interesting site, and I really like the idea.
Basically it seems to be using the tone of the sentence in which an article is cited. So there were papers that contradicted each other listed as contradicting, but there is also more basic comparative work listed as contradicting. For example if a paper on one species/condition finds a value of 20 au and another paper says that in their work on another species/condition the same value was 50 au, that is scored as contradiction.

Ultimately to be really useful the analysis would examine if the articles are contradicting the main hypothesis or conclusion of each other, not just disagreeing on a particular value.


It’s stalling out for me with complete title searches, even unambiguous ones, but is lightning-fast for DOI searches.

Really wish it had indication, species, or reagent filters.


It was terrible at title and combined title/author searches, so I switched to DOI.

1 Like

Very cool idea! We borked it with too many requests though (as of 09:30 PDT), so wait a while before taking a look.


On the papers I tried, the sentiment analysis was really poor. On the other hand, the way they lay out the reports (showing the chunk of text where the citation appears) is really useful. That already makes it one of the best citation lookup sites out there. They could ditch the ML part altogether, but then I guess nobody would pay them any attention…:frowning_face:


Interesting, like legal opinion validity checking. Shepardizing science studies.


Ooo! Handy! Just used it on a real job.

When a pamphlet was published entitled “100 Authors Against Einstein” , Einstein retorted “If I were wrong, one would be enough .”

For those of you who have not had the joy of wiring a scientific paper, it is pretty much de rigeur to review the subject to the point. This can be necessary to show what work is yours, and what went before. But I recently had a paper rejected because I did not include all the previous history only to say ‘…but what if we ignore all of this and start from scratch’. So, every paper will tend to have lots of citations of other papers, and maybe the citations of those papers too. These are mostly uncritical citations. Most research is in small fields, so you try and be polite as you are going to meet the same people at conferences. But if something can find critical reviews, there will not be many, so I can sort them by hand, and I can live with a few false positives.


Thank you everyone for the analyses. Gonna keep an eye on this; although I’m not in the paper writing business, being able to see if something pegs the BS-o-meter is very useful, especially in fringe writing.

1 Like

They missed the opportunity to name it Cite or Shite.

But that’s probably a good thing.


Yes, I spent the inordinate amount of time trying to engage Disconnect before a (5 best ideas of the day from Aspen Inst.) page on or or something had a default behavior in Firefox of READING ITSELF ALOUD to the deprecation of non-gibbering CPU functions. Suggestions for type and scale of purgative flames, and direction, please…

:paperclip:-yo it like this site needs to be left indoors for Sukkot, in the oven, a’ight?

1 Like

This topic was automatically closed after 5 days. New replies are no longer allowed.