# Books and long-form text for AI training

URL: https://aidataforsale.com/data/books-and-long-form-text
Publisher: AI Data For Sale (Marketing Signals Ltd)
Updated: 2026-11-10

How publishers and rights holders license books, journals and long-form archives for AI training, with reported deal values and the rights questions authors raise.

## Summary

- Books and long-form text teach models to follow arguments and structure across thousands of words, which short web text cannot do.
- Wiley reported it expected to earn $44 million from AI rights deals, according to The Bookseller in August 2024.
- The key question for publishers is whether contracts with authors grant the rights needed to license text for training.
- Text licensing rates reported by Reuters were around $0.001 per word through one data vendor, but large archive deals are negotiated as a whole.

## Specification

| Field | Detail |
| --- | --- |
| What it is | Licensed books, journals, magazine archives, reports and other edited long-form writing in clean digital form. |
| Measured in | Titles, words or tokens, plus subject coverage and publication years. |
| Who buys | Foundation model labs and companies building research, education and writing tools. |
| Where it comes from | Publishers, academic and professional societies, magazine and newsletter archives, and authors holding their own rights. |
| Rights to clear | Author and contributor contracts, territorial rights, third-party images and quotations, and any opt-out promises made to authors. |
| Privacy | Usually low, except memoir, case material and reader correspondence. |
| Pricing signal | Defined.ai quoted about $0.001 per word for text (Reuters, 2024). Wiley expected $44 million from AI rights deals (2024). |

## Why long-form text matters

Most of the open web is short. Books and edited archives give models examples of sustained reasoning, consistent style and accurate structure over long spans. That is why publishers have been among the first sellers to reach large licensing agreements.

## Reported deals and rates

- Wiley said it expected to earn $44 million from AI rights deals, and confirmed authors could not opt out ([The Bookseller](https://thebookseller.com/news/wiley-set-to-earn-44m-from-ai-rights-deals-confirms-no-opt-out-for-authors)).
- Defined.ai's CEO told Reuters the company paid about $0.001 per word for text, and that Brazil-based suppliers typically paid rights owners 20% to 30% of deal value ([Reuters via The Star](https://www.thestar.com.my/tech/tech-news/2024/04/26/inside-big-techs-underground-race-to-buy-ai-training-data)).

Per-word rates make small archives look modest. Value rises with exclusivity, subject depth and rights clarity rather than word count alone.

## Rights questions to settle

1. **Do your author contracts cover it?** Older contracts rarely mention machine learning. Some publishers rely on broad electronic rights clauses; others seek author consent or share revenue.
2. **Are there embedded third-party works?** Photographs, charts and long quotations may belong to someone else.
3. **What territory?** Rights acquired for one region may not cover a global license.
4. **What did you promise authors?** Public statements about opt-outs or revenue sharing become part of the deal's risk profile.

The US Copyright Office has published a report on generative AI training that discusses licensing markets and fair use ([US Copyright Office](https://www.copyright.gov/ai/)). Licensing removes the uncertainty for buyers, which is part of what they pay for.

## Delivering text

Buyers want clean text, not page images: EPUB or structured XML is ideal, with metadata for title, date, subject and rights status per item. A rights manifest listing every title and its clearance basis is often the deciding document in diligence.

## Questions

### Can a publisher license its backlist for AI training?

Yes, if author and contributor contracts grant the necessary rights or the publisher obtains them. Many publishers review contracts title by title and license only the cleared portion.

### Can an individual author license their own books?

If you hold the rights, yes. Check your publishing contract first, because the publisher may hold electronic or licensing rights.

### Is out-of-print material valuable?

It can be, especially in specialist subjects that are underrepresented online. Rights clarity is usually the limiting factor.

## Next steps

- [Request this data](https://aidataforsale.com/request-data)
- [License data like this](https://aidataforsale.com/list-your-data)

## Sources

- [The Bookseller: Wiley set to earn $44m from AI rights deals (August 2024)](https://thebookseller.com/news/wiley-set-to-earn-44m-from-ai-rights-deals-confirms-no-opt-out-for-authors)
- [Reuters via The Star: Inside Big Tech's underground race to buy AI training data (April 2024)](https://www.thestar.com.my/tech/tech-news/2024/04/26/inside-big-techs-underground-race-to-buy-ai-training-data)
- [US Copyright Office: Copyright and Artificial Intelligence, Part 3: Generative AI Training (pre-publication, May 2025)](https://www.copyright.gov/ai/)