Editorial methodology
Every book on Readark carries two separate 1–5 ratings — spice and darkness — plus content warnings. This page explains exactly what those numbers mean, how they are produced, and what they cannot tell you.
Most book sites collapse "intense" into a single number. That fails dark romance readers, because two very different things are being measured. Spice is about explicit sexual content. Darkness is about subject matter and moral weight. They move independently.
A romance can be scorching hot and emotionally gentle. Another can be genuinely bleak with almost nothing explicit on the page. Collapsing those into "5/5 intense" tells you nothing useful. Rating them separately lets you say what you actually want: high heat, low darkness. Or the reverse.
Darkness is not a quality score. A 5 is not "better" or "more advanced" than a 2 — it is simply further from safe. Many of the strongest books in the genre sit at 2 or 3.
Ratings tell you intensity; they do not tell you which theme you may want to avoid. That is what content warnings are for. We tag from a fixed vocabulary so the labels stay consistent across thousands of books: dubious consent, sexual assault / non-con, graphic violence, torture, kidnapping / captivity, physical abuse, emotional abuse / gaslighting, blood / knife play, self-harm, substance abuse, pregnancy loss, and death of a loved one.
Warnings are collapsed by default on every book page. Some readers want to know everything before starting; others consider the same information a spoiler. Collapsing it means you choose when to look, rather than having it forced into view.
Tropes are tagged from a fixed list too — enemies to lovers, captive / captor, stalker, forced marriage, age gap, morally grey MC, possessive-obsessive MC, touch her and you die, and around thirty more. A controlled vocabulary is less expressive than free-text tagging, but it makes browsing reliable: a trope filter returns every book carrying that tag, not just the ones where someone happened to phrase it that way.
We are transparent about this: Readark’s ratings are machine-assisted, not the result of an editor reading all 6,000+ books. Anyone claiming otherwise at this catalogue size would not be telling you the truth.
Each rating is generated by a language model working from the book’s factual metadata only — title, author, and subgenre — combined with well-established genre conventions. The model is explicitly instructed that it has not seen any publisher or retailer description, and must not reproduce marketing copy, invent plot events, or fabricate quotes. Where a book has no usable signal, it inherits a conservative default from its subgenre.
What that means for you: treat the numbers as directional, not authoritative. They are reliable for sorting and filtering across a large catalogue — finding the gentler end of mafia romance, or the heaviest end of captive romance. They are not a substitute for an author’s own content notes, and they will occasionally be wrong on an individual title.
Our bestseller lists are not editorial picks or sponsored placements. They are derived from real Amazon Best Sellers Rank, pulled directly from Amazon’s catalogue data for each title’s romance or fantasy category.
Because raw rank is a steep curve — the gap between #1 and #100 matters far more than the gap between #5,000 and #5,100 — we convert it to a 0–100 score on a logarithmic curve, so a book ranked in the low hundreds scores meaningfully higher than one ranked in the thousands. Books whose only ranking sits in an unrelated category are excluded rather than given a misleading score. Ranks are refreshed on a recurring schedule, so the lists move as real sales move.
Readark is free to use and funded by affiliate commissions: as an Amazon Associate, we may earn from qualifying purchases made through links on this site. That relationship is disclosed on every book page.
Crucially, commission does not influence spice, darkness, content warnings, or bestseller position. Every title in the catalogue carries the same affiliate relationship, so there is no title we are paid more to promote — and the bestseller ordering is computed from Amazon rank data, not chosen by us.
We would rather state our limits plainly than overclaim. Ratings are machine-assisted and directional. Content warnings are drawn from a fixed list and may miss a theme outside it. Bestseller data reflects Amazon’s ranking at the time of our last refresh, not live sales.
If a rating or warning on a book looks wrong to you — especially if a missing content warning could affect another reader — tell us at info@revoba.net with the title and what should change. Corrections from readers who have actually read the book are the single best signal we get, and they are applied by hand.
Spice measures explicit sexual content on a 1–5 scale, from closed door to very explicit throughout. Darkness measures subject matter and moral weight, from dark-flavoured to extreme taboo-adjacent themes. They are rated separately because a book can be high in one and low in the other.
No. Darkness is not a quality score — it only describes how far the book sits from safe territory. Many of the strongest books in the genre sit at 2 or 3 on the darkness scale.
No, and we do not claim otherwise. Ratings are machine-assisted: a language model works from each book’s title, author and subgenre together with genre conventions. Treat them as directional for filtering rather than authoritative for an individual title, and check the author’s own content notes as well.
From real Amazon Best Sellers Rank for each title’s romance or fantasy category, converted to a 0–100 score on a logarithmic curve so top-ranked books separate meaningfully from mid-list ones. Rankings are refreshed on a recurring schedule and are never sold or editorially placed.
Because readers disagree about whether warnings are safety information or spoilers. Collapsing them by default means you decide when to see them.
Email info@revoba.net with the book title and what should change. Reports from readers who have read the book are applied by hand and are the most valuable correction signal we have.
78score
