"Most read" is a claim that needs a measurement, so this site states its method in full. Every title is a complete public-domain text, obtained from free resources on the internet and gathered into a 3,400-work candidate pool. Roughly 300 of those works carry enough external evidence of readership to be ranked. The top hundred are published here with their complete text.

The score

Each work gets a score out of 100, from three weighted components.

Component	Weight	Source
Rating volume	60%	Goodreads, via the public goodbooks-10k dataset — log-scaled, so the difference between 50k and 500k ratings counts more than 1.5M vs 2M.
Live downloads	20%	Project Gutenberg's rolling Top-100 download chart, captured across three windows (yesterday, 7 days, 30 days).
Canon presence	20%	How many of three documented lists include the work.

The lists used

The three syllabus and canon lists are quoted verbatim from their published sources: the 1984 National Endowment for the Humanities core reading list (34 works), a 1,000-voter Goodreads survey of essential high-school reading (50 works), and a consolidated current-curriculum survey (51 works). A fourth check used the same 2019 Goodreads "classics" dataset as a fallback where live rating data was missing.

What this ranking is not

It measures recorded readership — the people who log, download, and are assigned books. That skews English-language and it skews institutional. A novel read widely in informal translation across three continents may rank lower here than its real audience deserves, and the opposite applies to books that sit on every syllabus and are read by few.

What was excluded

Ten titles that would have ranked highly — Don Quixote, Middlemarch, Crime and Punishment, Emma, Les Misérables, Gulliver's Travels, The Brothers Karamazov, The Age of Innocence, North and South and A Study in Scarlet — could not be obtained as complete texts from the free sources used. Rather than list a book nobody can read, they were dropped and replaced by the next qualifying titles. Thirty-seven of the 298 ranked works failed this same text check.