Flow and readability
18%Continuity, sentence and paragraph rhythm, transitions, modulation and the experience of reading sustained passages.
Two rankings, two different questions
Algorithmic assessment. Reader response. Two distinct views.
The default shelf ranks books by an automated assessment of the complete manuscript. Readers can instead choose the reader ranking, built from private questionnaires by people who show credible knowledge of the book.
Style and personality
12%A natural, recognisable authorial presence with tonal range rather than formulaic or repetitive phrasing.
Story craft
20%Conflict, development, momentum, emotional effect and whether the story fulfils its own intentions.
Character depth
15%Specificity, credible motivation, consistency and meaningful change.
Distinct character voices
10%Differences in vocabulary, cadence, humour, emotional register and worldview—not simply clever dialogue.
Structure and pacing
10%Shape, balance, scene development and control from beginning to end.
Technical correctness
10%Grammar, spelling, punctuation, syntax, sentence construction and clear expression under the conventions of the book’s language edition.
Originality
3%Freshness of conception and treatment; originality cannot conceal an uncomfortable or badly constructed reading experience.
Coherence
2%Internal logic and continuity across the complete manuscript.
An automated assessment, with limits
The literary score is generated by our assessment algorithm, not a human editor. It can miss nuance or make mistakes and is not a guarantee of enjoyment, editorial quality or sales. The rubric is genre-aware; each assessment applies it to the complete supplied manuscript. Qualifying books and their results are listed automatically with the author’s submission consent.
Authors can view their result and withdraw using their private management link. To assess a revised edition, withdraw the old listing and submit the revised manuscript. See the author guide for the full process.
Our commitment to human writing
Our assessment algorithm is designed to reward the depth, individuality and sustained craft of human writing. With today’s AI capabilities, our aim is for AI-generated books generally to score lower than strong human-written books. The House of Lit will continue to develop its assessment algorithm with that aim. This is a commitment to literary quality, not a claim that the algorithm can reliably identify how a book was written or guarantee any individual result.
The literary ranking
Every complete book is assessed under the same genre-aware rubric. The assessment judges the cumulative experience of reading passages across the beginning, middle and ending—not the number of individually clever or quotable lines. Persistent choppiness, monotonous cadence, mechanical paragraphing and interchangeable character voices reduce the rating, and poor flow cannot be offset by originality alone.
Grammar and construction are part of the reading experience, not cosmetic polish. Repeated faults in syntax, punctuation, spelling or sentence construction reduce the score substantially. A book with seriously defective technical writing cannot receive a high literary rating, regardless of the strength of its premise.
Every manuscript is judged against the same literary criteria, without an automatic penalty based on how it was written. Popularity, sales, advertising and the author’s identity do not affect the result. A low rating does not itself disqualify a complete, readable book.
The reader ranking
Readers never enter a number. They answer questions based on how much they actually read. Specific memories, character observations and internally consistent answers establish reading confidence; liking the book is never evidence of having read it. A thoughtful negative response counts exactly as a positive one.
Private answers, public aggregates
Reader answers and email addresses are never published. A public reader score appears only after at least three credible questionnaires, limiting the influence of any single response.