Books and backlists

Sell books to AI companies: what a book earns per title

Updated

Our table carries two per-title figures for books in AI training. In the HarperCollins deal, and by default in the settlement for trade and university press books, the author and the publisher share the money.

Describe my language data Language, kind of data, hours and consent decide which buyers would look at it.

What do AI companies pay per book?

HarperCollins' licensing deal pays a $5,000 fee per title, as the Authors Guild reports it. That is the one licence fee per book on the table.

Source: Authors Guild on the HarperCollins AI licensing deal

The Bartz v. Anthropic settlement is a court settlement rather than a sale, and its site puts the award at approximately $3,000 per work, plus interest and less the court-approved costs and fees.

Source: Bartz v. Anthropic settlement, frequently asked questions

How is it split between author and publisher?

In the HarperCollins deal the $5,000 is split 50-50 between the author who chooses to take part and HarperCollins.

Source: Authors Guild on the HarperCollins AI licensing deal

In the settlement, trade books and university press books are split 50-50 between authors and publishers by default.

Source: Bartz v. Anthropic settlement, frequently asked questions

Princeton University Press says it is fully aligned with that default split for trade and university presses.

Source: Princeton University Press on the Bartz v. Anthropic settlement

Does the publisher need the author's permission?

The Authors Guild says publishers must obtain permission from authors before including their books in AI licensing deals, because a typical trade contract reserves those rights to the author.

Source: Authors Guild on the HarperCollins AI licensing deal

Who on the table deals in books?

3 of the 12 rows on the table are about books, read up to 2026-10-22; the other rows are the data sellers and platforms the board read first. Each book row quotes its source's own sentence.

Source: Speech and language data prices

Tell us about the language data you hold

Language, kind of data, hours and consent decide which buyers would look at it.

The language and the corpus

What kind of language data do you hold?

  • No audio or text files are asked for
  • Answers set against the language data sellers tabled here

Sources

This site's own dataset, every figure quoted from its source and dated:

  • Speech and language data prices, seller by seller: 12 sellers of speech and language data were read for the prices they publish and what they pay contributors, each answer quoted from their own pages, listings and announcements, read 22 October 2026.

Describe the language data you hold

Language, kind of data, hours and consent decide which buyers would look at it.

Describe my language data