How BookLens rates books
Our ratings combine AI full-text analysis with a structured five-category rubric. Every rating is derived from the actual words on the page — never from summaries, reviews, or marketing copy.
The 5 categories
Every book is scored 0–4 in each category. The scores combine to produce an overall rating.
Violence
Physical harm, combat, injury, death, war, and threatening behavior. Includes both realistic violence and fantasy/game-world violence.
No violence present.
Minor conflict, cartoon-style action, or brief implied violence without detail.
Combat, injury, or death described with some detail. Not dwelling, but present.
Sustained violence, significant injury detail, death scenes with emotional weight.
Highly detailed depictions of gore, torture, extreme cruelty, or mass violence.
Language
Profanity, slurs, blasphemy, and coarse language. We track exact counts for every flagged word and phrase.
No profanity or strong language.
Isolated mild profanity (damn, hell). No strong language.
Occasional moderate profanity or isolated strong language.
Frequent strong language, including the f-word or slurs.
Pervasive extreme language throughout the book.
Sexual Content
Romantic and sexual content, from innocent attraction to explicit scenes. Includes nudity and implied acts.
No romantic or sexual content.
Innocent romance, hand-holding, kissing. No sexual content.
Implied sexual activity, tasteful scenes that fade to black.
Some explicit description of sexual activity, though not dwelling.
Explicit, detailed sexual scenes.
Substance Use
Alcohol, drugs, tobacco, and other substances. Includes both portrayal and consequences.
No substance use present.
Passing reference or background presence (wine at a dinner, etc.).
Characters drink or use substances, sometimes to excess.
Substance use is a significant story element; addiction or severe consequences depicted.
Detailed depiction of drug use, addiction spiral, or glorified substance abuse.
Psychological Intensity
Emotional weight, disturbing themes, horror, trauma, mental illness, suicide, and content that may be psychologically distressing.
Light, low-stakes story.
Some emotional content, mild peril, low stakes tension.
Meaningful emotional weight, scary situations, themes of loss or grief.
Dark or disturbing themes, trauma, significant horror or dread, difficult emotional content.
Extreme psychological horror, graphic depiction of trauma, self-harm, or deeply disturbing content.
Overall ratings
The overall rating is determined by the highest individual category score, weighted by frequency and detail.
Suitable for all ages. No concerning content across any category.
Mild content in one or more categories. Fine for most readers; parents may want to preview for young children.
Moderate content in one or more areas. Best for readers 13 and up.
Strong content — intense violence, language, sexual content, or psychological themes. Adults only.
Explicit content. Not appropriate for minors under any circumstances.
AI methodology
BookLens uses Claude (by Anthropic) to analyze book content. For every book, we process the full text in ~20,000-word chunks, running each chunk through our structured content rubric.
A programmatic pass runs first: we scan for known profanity and content keywords using a curated word list, producing exact counts for every flagged term. This pass is fast, deterministic, and not subject to AI interpretation.
The AI pass then analyzes context, tone, and nuance. It identifies violence, sexual content, psychological themes, and substance use that keyword matching would miss — and it understands when context changes the impact (e.g., a death scene that is brief vs. prolonged, or violence that is consequences vs. glorification).
Per-chunk scores are aggregated into a final rating, and a “Parents Need to Know” summary is generated covering the key content considerations in plain language.
We include the Claude model version used in every rating record so ratings can be reproduced and compared as models improve.
Accuracy & corrections
AI analysis is not perfect. If you believe a rating is inaccurate — either too strict or too lenient — you can flag it using the “Flag inaccuracy” button on any book page. We review flagged ratings manually and update them when warranted.
Community notes allow readers to add scene-specific context that supplements the AI analysis. These are moderated and marked separately from the AI rating.