Fair Use, the Four-Factor Test, and When It Doesn't Protect You

By Hank FasthoffJuly 4, 2026

Fair use is the most commonly invoked and most commonly misunderstood defense in copyright law. Under 17 U.S.C. § 107, the fair use of a copyrighted work for purposes such as criticism, comment, news reporting, teaching, scholarship, or research isn't infringement. But fair use is an affirmative defense, meaning the defendant bears the burden of proving it applies. It's determined case by case, by weighing four statutory factors that often point in different directions. And the Supreme Court narrowed it significantly in Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508 (2023), which limited the scope of transformative use in commercial contexts.

The Four Statutory Factors

Section 107 identifies four factors courts must consider in determining whether a use is fair. No single factor is dispositive, and all four must be weighed together in light of the purposes of copyright. Campbell v. Acuff-Rose Music, Inc., 510 U.S. 569, 578 (1994).

Factor one asks about the purpose and character of the use, including whether it's commercial or nonprofit educational. Factor two asks about the nature of the copyrighted work. Factor three asks about the amount and substantiality of the portion used in relation to the copyrighted work as a whole. Factor four asks about the effect of the use on the potential market for or value of the copyrighted work.

Factor One, Purpose and Character of the Use

Factor one is where most fair use disputes are decided, and it has undergone the most significant evolution in recent Supreme Court decisions.

In Campbell v. Acuff-Rose Music, Inc., 510 U.S. 569 (1994), the Supreme Court held that 2 Live Crew's parody of Roy Orbison's "Oh, Pretty Woman" could constitute fair use, and in doing so introduced the concept of transformative use into the fair use analysis. A use is transformative if it adds something new, with a further purpose or different character, altering the first work with new expression, meaning, or message. Justice Souter's opinion emphasized that the more transformative the new work, the less significant other factors, including commercial purpose, become.

After Campbell, lower courts expanded the transformative use concept substantially. Courts in some circuits found transformative purpose when a secondary work served any new aesthetic, meaning, or message, even without commentary on the original work. Appropriation art, documentary film, search engine thumbnails, and text mining were all found transformative under this broad reading.

In Google LLC v. Oracle America, Inc., 593 U.S. 1 (2021), the Supreme Court held that Google's copying of approximately 11,500 lines of code from Oracle's Java API declarations was a fair use. Justice Breyer's opinion found that Google's use was transformative because it used the declarations in a new context (smartphone platforms) and for a new purpose (allowing programmers to use their existing Java skills to write new programs for Android). However, the decision was closely tied to the unique nature of API declarations, which fall at the boundary between uncopyrightable ideas and copyrightable expression.

In Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508 (2023), the Supreme Court significantly narrowed the transformative use inquiry. Justice Sotomayor's opinion held that the first fair use factor weighed against AWF's licensing of Andy Warhol's "Orange Prince" image to Condé Nast for a magazine cover about Prince, because the licensed use served substantially the same purpose as Lynn Goldsmith's original photograph, a portrait of Prince used in a magazine to illustrate a story about Prince. Both were commercial uses serving the same function, and the Court affirmed the Second Circuit's judgment that the use wasn't fair, the other three factors having gone unchallenged.

Goldsmith established that new expression alone isn't dispositive of transformative use. A secondary work must have a purpose or character that is "distinct from the original." When the secondary work and the original serve substantially the same commercial purpose, adding new aesthetics or new expression doesn't make the use fair. Critically, the Court limited its analysis to AWF's specific licensing to Condé Nast, and noted that a different use, such as a teaching use or an appearance in an art magazine alongside an article about Warhol, might yield a different result. Justice Gorsuch, concurring, added that display in a nonprofit museum or a for-profit book commenting on twentieth-century art might likewise point toward fair use.

After Goldsmith, the central question under factor one is whether the accused use has a different purpose from the original, not merely a different expression. Commercial uses that serve the same purpose as the original face a steep burden.

Factor Two, Nature of the Copyrighted Work

Factor two considers whether the copyrighted work is creative or factual, and whether it's published or unpublished. Creative works receive broader protection than factual compilations, and unpublished works receive broader protection than published ones, because the right of first publication is particularly significant. Harper & Row, Publishers, Inc. v. Nation Enterprises, 471 U.S. 539, 564 (1985).

In practice, factor two rarely determines the outcome of a fair use case. Courts frequently note that it's of limited significance when the secondary use is transformative. But when a court has already found the use non-transformative under factor one, the creative nature of the original work reinforces the conclusion that the use wasn't fair.

Factor Three, Amount and Substantiality Used

Factor three asks both how much of the copyrighted work was used (a quantitative inquiry) and whether the portion used constitutes the "heart" of the work (a qualitative inquiry). Copying an entire work doesn't automatically preclude fair use, nor does copying a small portion guarantee it. In Campbell, the Supreme Court held that a parody may need to copy enough of the original to "conjure up" the work being parodied, and that copying the original's most distinctive elements can be justified by the parodic purpose.

In Harper & Row, the Supreme Court held that copying only 300 words of a 200,000-word memoir wasn't fair use because the 300 words taken were the "heart of the book," the most powerful and newsworthy portion. Factor three asks not just how much was taken, but whether what was taken is the most valuable part of the original.

Factor Four, Effect on the Market

Factor four considers the effect of the challenged use on the potential market for the copyrighted work, including the market for derivative works. This factor asks whether the secondary use serves as a substitute for the original, displacing sales or licensing revenue the copyright owner would otherwise receive.

In Campbell, the Supreme Court clarified that market harm must be evaluated in relation to the transformativeness of the use. A transformative work is less likely to substitute for the original than a work that merely copies it. But when the secondary work competes directly with the original in the same market, factor four weighs heavily against fair use.

In Goldsmith, factor four also cut against AWF. The Second Circuit found that AWF's licensing encroached on Goldsmith's market for licensing her photograph for magazine covers and editorial use, a determination AWF didn't challenge in the Supreme Court. Justice Sotomayor's factor-one analysis for the majority stressed that AWF's license to Condé Nast for the same purpose (a magazine cover about Prince) superseded the objects of her photograph. The use displaced revenue Goldsmith would otherwise have received.

Common Misconceptions About Fair Use

Fair use doesn't depend on crediting the original author. Attribution is a professional courtesy, not a legal defense. Copying a copyrighted photograph and providing credit to the photographer is still infringement if the use doesn't satisfy the four-factor test.

Fair use isn't a function of how much you copy. Copying 10 percent of a work can be infringement if the 10 percent taken is the most important part. Copying 100 percent of a work can be fair use if the purpose is sufficiently transformative and the use doesn't substitute for the original.

Commercial use isn't presumptively unfair, and noncommercial use doesn't automatically qualify as fair. Language in Sony Corp. of America v. Universal City Studios, Inc., 464 U.S. 417, 451 (1984), suggested that every commercial use of copyrighted material is presumptively unfair, and the Sixth Circuit applied that language as a presumption in Campbell. The Supreme Court rejected that reading, holding that the court of appeals erred in giving virtually dispositive weight to commercial character by way of a presumption, and that Sony itself called for no hard evidentiary presumption. Commercial character is one element within factor one, not a thumb on the scale. Nonprofit and educational use receives no automatic pass either. Courts weigh all four factors regardless of commercial status.

"Fair use" doesn't mean "any use I think is fair." A belief that copying is justified doesn't create a legal defense. Fair use is determined by the statutory factors applied to the specific facts of the specific use.

What Goldsmith Changed

Before Goldsmith, defendants could argue that adding new expression, meaning, or aesthetics to a copyrighted work was transformative, even when the secondary work served the same commercial purpose as the original. After Goldsmith, purpose weighs as heavily as expression. A use that serves the same purpose as the original, in the same market, for the same audience, faces a heavy burden on factor one, regardless of how much new expression the defendant added.

For businesses, the practical takeaway is that licensing a copyrighted work and modifying it for the same commercial purpose (a photograph altered for use in the same type of publication, a song sampled for use in the same genre, a design element adapted for the same product category) is unlikely to qualify as fair use after Goldsmith. The safest approach is to obtain a license when the intended use serves the same purpose as the original, and to reserve the fair use argument for uses that serve a materially different function, such as commentary, criticism, education, parody, or research.

Fair Use and AI Training

Whether training an AI model on copyrighted works is fair use is the largest open question in the doctrine right now. District courts have split. Judge Alsup held in Bartz v. Anthropic (N.D. Cal. 2025) that training on lawfully acquired books was fair use, but that downloading and retaining pirated copies wasn't. The court in Thomson Reuters v. Ross Intelligence (D. Del. 2025) found no fair use where the AI product competed directly with the copyrighted legal research product it was trained on. The Third Circuit heard argument on that ruling in June 2026, the first appellate test of the question, and hasn't yet issued a decision.

Businesses that use generative AI tools, or whose content those tools consume, are operating without settled law. The Goldsmith purpose inquiry is doing heavy work in these cases, since a model trained to produce outputs that compete with the training data looks different from one trained for a materially distinct function.

Fair use is powerful when it applies, but it's narrow, fact-specific, and in flux. Relying on it without analyzing all four factors against the current case law, including Goldsmith, is a risk that frequently produces litigation rather than avoiding it.