As the title says. Non-expert reviews only evaluate how good the writing was, or how impressed the reviewer was by the ideas; but I typically want something that evaluates the actual scientific claims.
My one trick has been to look at Google Scholar for papers that cite the book and search for "review" within those, which sometimes finds reviews published at academic journals, but this doesn't seem very effective overall.
Elizabeth's Epistemic Spot Checks series comes to mind.
The scholarly literature sometimes features article-type reviews of 'popular' science books.
I will look on ebscohost (google scholar may work as well)
for just the title of the book,
refine my search for publication dates in the first couple years after the original book was published (although I just saw an article-review on Seeing Like a State from 2010, 12 years after the original's publication)
And then there are often reviews from noteworthies in the same field as the work, particularly if the author has published in academia prior to the popular work.
I've found them largely useless for any research, however, due to their brevity and their largely not-serious approach. The review is simply not treated like a scholarly response, it's more of a 'five stars, would recommend' thing usually speaking, and even criticisms in the pieces feel off the cuff and not well considered.