Somebody quotes a paragraph, clips ten seconds of a song or builds a video essay around a film, and the reassurance arrives immediately in the comments: that is fair use. Understanding how does fair use work starts with the uncomfortable part, which is that nobody in that conversation is in a position to know.
Fair use is not a permission granted in advance. It is a defence raised after someone sues, and it is decided by a court weighing four statutory factors against the specific facts of the specific use.
There is no threshold, no percentage and no safe quantity anywhere in the law. The Copyright Office says so itself, in as many words, on its own website.
A defence, not a permission
The distinction sounds pedantic and is the most practically important thing in the subject. A licence is permission obtained beforehand from the owner; fair use is a legal argument that the thing you already did was not infringement, and it becomes authoritative only when a judge agrees.
That is why the honest answer to “is this fair use?” is almost always that it depends, and why anyone offering a firm yes before a dispute exists is describing a prediction rather than a fact.
It also explains why large platforms remove content so readily. They are not adjudicating fair use and have no mechanism to do so; they are avoiding the position of having to defend somebody else’s legal argument at their own expense.
What the statute actually says
The text is short, and two of its phrases do enormous work. Under section 107 of the Copyright Act, the fair use of a copyrighted work, for purposes such as criticism, comment, news reporting, teaching, scholarship or research, is not an infringement of copyright.
The statute then sets out the factors to be considered.
| Factor | What the statute asks |
|---|---|
| 1 | The purpose and character of the use, including whether it is commercial or for nonprofit educational purposes |
| 2 | The nature of the copyrighted work |
| 3 | The amount and substantiality of the portion used, in relation to the work as a whole |
| 4 | The effect of the use upon the potential market for, or value of, the copyrighted work |
Now read the two phrases carefully. The purposes are introduced with “such as”, and the factors are those that “shall include” the four listed.
Neither list is closed. Other purposes can qualify, and courts may weigh considerations beyond the four — which the Copyright Office confirms, noting that other factors may also be considered depending on the circumstances.
The listed purposes are examples, not categories
This is where a great deal of confidence goes wrong. Because teaching appears in the statute, it is widely assumed that classroom use is automatically fair. It is not: the purpose named in the preamble is a starting point, and the full four-factor analysis still runs afterwards.
The same applies to criticism, commentary and news reporting. Labelling a video an essay or a review does not settle the question, and calling a channel a commentary channel settles even less — though genuinely doing the thing may help under the first factor.
Conversely, a use that fits none of the listed purposes is not automatically unfair. The words governing the list are “such as”, and courts have found fair use in situations that appear nowhere in the statutory examples.
How the factors actually interact
The four are weighed together rather than counted, and no single one decides an outcome.
The first factor asks what the use is doing. Courts ask whether the new work adds something with a further purpose or different character, and commercial use weighs against fair use without ruling it out.
The second asks what was taken from. Copying from a factual or already published work supports the defence more readily than copying from a creative one, and the statute adds explicitly that the fact a work is unpublished does not by itself bar a finding of fair use.
The third asks how much, in both senses at once. A very short extract can defeat the defence if what was taken is the heart of the work, and a substantial amount can survive it where the stated purpose genuinely requires that much.
The fourth asks about harm, and it asks a narrower question than people expect. Not whether the original author is annoyed, nor whether they would have preferred to be asked, but whether the use substitutes for the original in its own market — and whether copying of that kind, if it became widespread, would damage that market.
The fourth is generally the heaviest, and it is the one most people never consider.
There is no percentage
Every version of the rule you have heard — ten percent of a book, thirty seconds of a song, eight bars of music, one image at low resolution — is invented.
The Copyright Office addresses this directly, and its wording leaves no room. Because courts evaluate claims on a case-by-case basis and the outcome of any given case depends on a fact-specific inquiry, there is no formula to ensure that a predetermined percentage or amount of a work — or a specific number of words, lines, pages or copies — may be used without permission.
Those numbers do have an origin. Some began as internal guidelines a university or publisher adopted to keep its own staff out of trouble, and a deliberately cautious internal policy is not a statement of what the law permits.
Others are simply repetition. A figure gets quoted in a forum, then in a guide summarising the forum, and eventually in an article citing the guide as authority — and nowhere in that chain is there a statute, a regulation or a court decision.
Credit does not create fair use
Attribution is a strong professional norm, and it is not a copyright doctrine at all.
Naming the source addresses plagiarism, which is an ethical and academic failing. Copyright concerns whether you had the right to reproduce the work at all, and those are different questions with different remedies.
A fully credited, carefully cited reproduction of an entire article is still an unlicensed reproduction of an entire article, and the citation does not change what was copied.
Credit can help slightly at the margins, since it supports a claim that the use is genuinely critical or scholarly rather than a straightforward substitute for the original. On its own it settles nothing.
Why the Office publishes a database of court cases
The most revealing thing about fair use is not in the statute. It is in what the Copyright Office built to explain it.
Rather than issuing a rule, the Office maintains a searchable index of judicial decisions, sorted by category and type of use — music, digitisation, parody and the rest — covering the Supreme Court, the circuit courts of appeal and the district courts.
Its stated purpose is to help lawyers and non-lawyers understand the types of uses courts have previously determined to be fair, or not fair. The Office also notes that the index does not include every judicial opinion on the subject.
Consider what that implies. The agency responsible for copyright in the United States cannot tell you whether a particular use is fair, so instead it shows you what judges have decided in comparable situations and leaves you to reason by analogy.
An index of outcomes is what a regulator produces when the answer genuinely depends on the case. It is an unusually honest piece of public administration, and it is also the clearest evidence that the confident answers circulating elsewhere are not available.
How this sits beside the rest of copyright
Fair use is one of two very different ways the law limits an owner’s control, and the contrast is instructive.
The other is the compulsory licence, where the owner cannot refuse and cannot set a price because a government body sets it — the arrangement that governs half of every song, as how music royalties work describes.
That mechanism is completely predictable and completely rigid — a rate published years ahead, applying whether anyone likes it or not. Fair use is its opposite in every respect: no rate, no procedure, no advance answer, and enormous flexibility.
Both exist because unrestricted owner control produces outcomes nobody wants, and they solve the same problem in opposite ways. One replaces negotiation with a fixed price set by the state; the other replaces permission with an argument made afterwards.
And both eventually run out, because copyright itself expires — after which none of this matters, as when does copyright expire sets out.
What the case law tends to reward
Patterns are visible across decided cases, though none of them is a rule and none of them guarantees an outcome in the next one.
- Uses that add commentary, analysis or criticism fare better than uses that simply present the original again
- Taking no more than the purpose requires fares better than taking everything available
- Uses that do not compete with the original in its own market fare better than substitutes
- Factual and published source material supports the defence more readily than creative or unpublished material
None of that is legal advice, and this article is not a substitute for it. It is a description of how a doctrine is structured and what its own regulator says about its predictability.
The single most useful correction is the one at the top. Fair use is not a box you can be safely inside; it is a case you would have to win, and the difference matters most to the people most confident it does not.
