What copyright is looking for in an image
U.S. copyright law protects original expression created by a human—not every image that looks creative or took effort to produce. With a conventional illustration, choices about line, color, composition, and detail are made by the artist. An image generator, by contrast, produces those visual choices itself in response to a prompt. If the person supplies instructions but does not determine the image’s expressive details, the result may lack the human authorship copyright requires.
That distinction is about who made the image’s expressive choices, not whether the image is impressive or useful. A long, carefully worded prompt does not automatically make its author the creator of every detail the system generates.
Where human authorship can enter
AI can still be part of a copyrightable work. A person might substantially redraw or edit generated material, or combine it with human-created elements in an original composition. In those cases, copyright may cover the person’s contributions—the new drawing, editing, or arrangement—without necessarily covering the AI-generated material itself. The line depends on what the person actually created, not simply on whether AI was used.
This is why “made with AI” is not a complete answer. The relevant question is whether a human contributed enough original expression to claim protection for some part of the finished work. Merely choosing an output from several generated versions may not establish authorship of the image’s underlying visual content.
The practical consequence
When an image consists entirely of machine-generated expression, its creator generally cannot use copyright to stop others from copying that image in the United States. That does not settle separate questions—such as whether the generation process used someone else’s protected work, or whether a platform’s contract limits use. It means only that copyright protection for the output depends on identifiable human creative contribution. A small human-authored addition may be protectable; the generated image as a whole is not automatically made exclusive by the prompt that produced it.
