What Is Fair Use for Scraped Data? – Definition and How It Works
Learn how the four fair-use factors apply to collected web content, and why facts and expression are treated differently.
Fair use is a United States copyright doctrine permitting some unlicensed use of protected work. It is a defence assessed case by case, not a category of permitted activity. This page is a general explanation, not legal advice.
1. Copyright Protects Expression, Not Facts
The most important distinction, and the one that resolves most questions before fair use is even reached.
Copyright protects original expression. It does not protect facts, ideas, or data as such. So:
| Content | Protected? |
|---|---|
| A product's price | No — a fact |
| A phone number or address | No — a fact |
| A specification value | No — a fact |
| An article's text | Yes — expression |
| A photograph | Yes |
| A creatively written product description | Yes |
| A review's text | Yes |
| A bare listing of facts | Generally not, absent creative selection |
A great deal of collected data — prices, availability, specifications, ratings as numbers — is factual and outside copyright entirely. Fair use is not needed for facts, because there is nothing to excuse.
2. Compilations Complicate It
A database of unprotected facts can attract protection in its selection and arrangement, where those involve originality. Copying the facts is one thing; reproducing a distinctive curated selection is another.
Note also that the EU has a separate database right protecting substantial investment in compiling a database, independent of copyright and with no fair-use equivalent. A US-centric fair-use analysis does not address it.
3. The Four Factors
Where protected expression is involved, fair use is weighed across four factors, none decisive alone.
1. Purpose and character. Transformative use — serving a different purpose from the original — weighs in favour. Commercial use weighs somewhat against, though it is not disqualifying. Research, analysis, criticism, and commentary weigh in favour.
2. Nature of the work. Factual works are weaker candidates for protection than highly creative ones, so using them weighs more favourably.
3. Amount and substantiality. How much was taken, and whether it was the heart of the work. Less weighs in favour; the whole work weighs against, though not fatally where the purpose requires it.
4. Effect on the market. Often the most influential factor. A use that substitutes for the original, or harms a market the rightsholder could exploit — including a licensing market — weighs heavily against.
4. How the Factors Fall for Common Uses
Extracting factual data for analysis — usually no copyright question at all, because facts are not protected.
Storing full page copies for internal processing — intermediate copying for a transformative purpose has fared reasonably well, particularly where the copies are not published.
Republishing article text — poor. It substitutes for the original directly.
Publishing short quotations with attribution — much better, especially in commentary or criticism.
Building a searchable index with snippets — has been treated favourably where the index serves discovery rather than substituting for reading.
Training a model on protected text — actively litigated and genuinely unsettled. Anyone told this is clearly resolved either way is being told something the case law does not yet support.
5. Practical Risk Reduction
- Take facts, not expression, wherever the requirement allows it.
- Store rather than republish. Internal use is a far better position than publication.
- Quote briefly, and attribute where expression is needed.
- Do not substitute for the source. If a user could use your output instead of the original, the market factor turns against you.
- Link back, which supports rather than replaces the source.
- Respect explicit licences — many sites publish terms permitting specific uses, which removes the question entirely.
- Consider licensing where the requirement genuinely needs protected content at volume. It is cheaper than a dispute.
- Take advice for anything commercially significant.
6. Fair Use Is Jurisdictional
Fair use is a US doctrine. Other jurisdictions differ substantially:
- The UK has fair dealing, limited to enumerated purposes rather than an open-ended test.
- The EU has specific exceptions, plus text-and-data-mining provisions with an opt-out mechanism.
- Many jurisdictions have narrower exceptions and no general flexible test.
Collection is inherently cross-border, so a fair-use analysis covers one jurisdiction's position and not the others.
7. Summary
Copyright protects expression rather than facts, so much collected data — prices, specifications, contact details, numeric ratings — raises no copyright question at all and needs no fair-use defence.
Where protected expression is involved, fair use is a case-by-case weighing of purpose, nature, amount, and market effect, with market substitution usually the most influential — so prefer facts to expression, store rather than republish, quote briefly with attribution, and remember that fair use is a US doctrine that says nothing about other jurisdictions or the EU database right.