Wikipedia's guideline for books, WP:NBOOK, sets a low bar on paper: two or more substantial reviews in sources independent of the book, and a title is presumed notable. The guideline rests on one word doing most of the work — independent — and two of the best-known names in trade publishing now sell a service that fails it without ever looking like it should.
Kirkus Reviews runs two separate operations under one masthead. The first is the one people mean when they say 'a Kirkus review': an editor at the magazine chooses a forthcoming title, assigns it to a freelance critic from the regular pool, and the review runs free of charge to the publisher, positive or not. The second is Kirkus Indie, built for self-published authors: the author submits the manuscript, pays a fee up front, and receives a review guaranteed to run on the same website, in the same format, often from the same freelance pool. Publishers Weekly runs an almost identical split between its standard reviews and its paid BookLife imprint.
Nothing about the finished page tells a casual reader which track produced it. The byline format matches. The star rating looks the same. The only difference that matters to a deletion discussion is a transaction nobody can see from the page itself: did an editor choose this book, or did the author's credit card choose the reviewer? WP:NBOOK, like the general notability guideline before it, asks for coverage an editor decided to run — not coverage a subject bought the right to receive.
A debut novelist came to us with three Kirkus Indie reviews, each four stars, each quoted on the back cover of the second printing. None of the three counted toward notability, because all three existed by arrangement rather than by an editor's choice — the guideline doesn't ask whether the praise was earned, only whether anyone besides the author decided the review should exist. A poet published by a small regional press came to us the same month with one review in a Sunday newspaper's books section and a longlisting for a minor regional prize. Two sources, neither glamorous, both chosen by somebody other than the poet. That thinner record cleared a bar the three paid reviews never reached.
A book is presumed notable once it has been the subject of two or more non-trivial published reviews in sources independent of the book itself.
Self-published books do sometimes earn an unpaid, editorially chosen review — small magazines and regional papers still cover them on merit, and a handful make it into the free Kirkus or PW queue without charge. What separates a review that counts from one that doesn't has nothing to do with how the book was published. A human editor somewhere has to pick the book up on their own, before any money changes hands over the review itself.
Checking which track a review came from takes longer than reading the star rating. Look for a disclosure line near the byline, a URL segment that says 'indie' or 'booklife,' or simply ask the author directly whether a fee was paid — most will say so without being pressed, since the arrangement is advertised, not hidden, everywhere except the one place it needs to be disclosed.
A paid review buys a line for the jacket. It has never once bought the two sources Wikipedia's guideline for books actually counts.