Blog · 2026-10-15 · baryreview editors

Can an AI-written article be copyrighted? What the Copyright Office said

The short answer

Can AI generated content be copyrighted? In the United States, as of the Copyright Office’s January 2025 report: not the parts the AI generated on its own, no matter how detailed your prompt was. Yes, for the parts a human actually wrote, selected, arranged, or substantially changed. Whether your particular article qualifies is decided case by case.

This is not legal advice. We are editors. The Copyright Office’s position is a registration policy and a reading of existing law, and courts may refine it. If copyright in a specific work matters to your business, talk to an attorney. But the report is clear enough that any business publishing AI-assisted content should know what it says, because it changes what editing is worth.

What the report is

The US Copyright Office is publishing its study of AI and copyright in parts. Part 1 (July 2024) covered digital replicas. Part 2, “Copyrightability”, was published on January 29, 2025 and addresses whether and when the outputs of generative AI can be protected. A pre-publication version of Part 3, on training, followed in May 2025. We read the full Part 2 report and the summary published by Skadden on February 4, 2025.

The Office’s starting point is that human authorship is required for copyright. That isn’t new; it is how the Office has treated works for decades. The question was how to apply it to text and images that a machine produces from a human’s instructions.

The eight conclusions, in the Office’s words

The report’s executive summary lists its conclusions. We quote them because paraphrases of this document tend to drift:

• “Questions of copyrightability and AI can be resolved pursuant to existing law, without the need for legislative change.”
• “The use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection for the output.”
• “Copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material.”
• “Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements.”
• “Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis.”
• “Based on the functioning of current generally available technology, prompts do not alone provide sufficient control.”
• “Human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs.”
• “The case has not been made for additional copyright or sui generis protection for AI-generated content.”

The rest of this post works through what each of those means for someone running a company blog.

Prompts alone don’t make you the author

This is the finding that surprises people. You might spend an hour writing a 400-word prompt with your brand voice, your audience, your outline, and your examples. The Office’s view is that you still didn’t write the article.

Its reasoning: “prompts alone do not provide sufficient human control to make users of an AI system the authors of the output. Prompts essentially function as instructions that convey unprotectible ideas.” Copyright protects expression, not ideas, and the prompt describes what you want while the model decides how to say it. Skadden’s summary puts it as: prompts “may reflect a user’s mental conception or idea, but they do not control the way that idea is expressed.”

The report also closes the obvious workaround. Re-running the prompt twenty times and picking the best output doesn’t change the analysis. The Office describes that as “re-rolling” the dice, and says the effort spent revising prompts is irrelevant, “as copyright protects original authorship, not hard work or ‘sweat of the brow.’”

The Office adds that this reflects how current tools work and could change as tools give users more direct control over the output. For now, though, a blog post that is a raw model output, however good the prompt, is not something you can register or enforce as your copyright. Anyone could copy it, and you would have little recourse under copyright law.

What is protected: your contribution

The report describes three ways a human becomes an author of a work that involves AI output.

1. Your own expression is perceptible in the output. If you feed the model your own written draft and it is still recognizable in what comes out, you own that portion. The report says that “where a human inputs their own copyrightable work and that work is perceptible in the output, they will be the author of at least that portion of the output,” with protection “analogous to that in a derivative work.” For writing, this is the case where you wrote the piece and asked the tool to tighten or restructure it.

2. Creative selection and arrangement. If you take AI-generated material and select, coordinate, or arrange it in an original way, the compilation can be protected, even where the individual pieces are not. There is a floor: the report’s footnote notes that the Office “generally will not register a compilation containing only two or three elements.” Picking the better of two drafts is not arrangement.

3. Creative modification. If you change the output enough that your changes meet the originality standard, the result is protectable. The report says that where a tool lets the user “control the selection and placement of individual creative elements,” then “whether such modifications rise to the minimum standard of originality required under Feist will depend on a case-by-case determination. In those cases where they do, the output should be copyrightable.” The originality bar in US law is low; the Supreme Court’s Feist decision, which the report quotes, says “even a slight amount will suffice.” But it has to be your creative expression, not just proofreading.

There is a fourth situation the report treats separately: AI as an assistive tool. The Office “agrees that there is an important distinction between using AI as a tool to assist in the creation of works and using AI as a stand-in for human creativity.” It gives examples of using AI for brainstorming, or to “structure or create a preliminary outline for literary works,” where the human references the output but writes the work. “Using AI in this way should not affect the copyrightability of the resulting human-authored work.” The same is true for a larger work that contains AI-generated pieces: a film with AI-generated effects is copyrightable as a film even if the effects alone are not.

What this means for a business blog

Most company blogs now sit somewhere on a spectrum between “the AI wrote it and we posted it” and “we wrote it and the AI fixed the commas.” Where you sit determines what you own.

A practical consequence follows. If you intend to treat your content as an asset (license it, syndicate it, build a book from it, or stop a copier), the human editing pass is not optional polish. It is what creates the asset. The Office’s 2023 registration guidance, which the report reaffirms, says that if a work contains more than a minimal amount of AI-generated material, “the applicant should disclose that information and provide a brief statement describing the human author’s contribution.” “I prompted it” is not a contribution it will accept.

There is a second consequence that has nothing to do with copyright and everything to do with why we exist. The same editing pass that makes an article yours is the one that catches the invented statistic and the fake citation. Substantial human editing is not a thing you do for the lawyers; it is the thing that makes the content safe to publish at all. Our list of what we check is, in effect, a checklist of the places where a human has to make a decision the model didn’t.

Practical steps: document the human work

If protection depends on what a human contributed, then being able to show what the human contributed is the whole game. A few habits make this easy:

Keep the raw output. Save the model’s draft, unedited, with the date and the tool used. This is your “before.” Without it, you can’t demonstrate what you changed.

Keep your drafts. Save at least one intermediate version and the final published version. Version history in a document tool is enough; so is a folder with dated files. If you use an editorial review like ours, the correction list you receive and the version after you applied it are both useful records, because they show specific human decisions about specific passages.

Log the decisions, briefly. A sentence or two per piece: “Rewrote intro and section 3 in my own words; replaced the model’s three examples with our own customer scenario; cut the FAQ; added the pricing paragraph.” This takes a minute and is exactly the kind of description a registration application asks for.

Write the parts that matter yourself. Opinions, examples, comparisons, and any claim about your own product are the paragraphs where human authorship is most valuable and where the model is most likely to be wrong anyway. Writing them yourself solves both problems.

Disclose consistently. Decide what your AI disclosure says and use it. The Office asks for disclosure in registration applications; your readers may expect it too. A policy you can point to is easier than a case-by-case decision after someone asks.

Don’t rely on a prompt library as an asset. Prompts convey ideas, and ideas aren’t protected. Prompts may be a trade secret, but that’s separate from owning the articles they produce.

What the report does not settle

The Office said it will keep watching the technology and may revisit its conclusions. The analysis is case by case, so there is no fixed percentage of edits that converts an AI draft into your work. Courts, not the Office, make final calls on infringement, and litigation over AI-assisted works is still early. Other countries take different approaches; the report has a section on them, and a business publishing in several markets should not assume the US analysis travels.

What is settled, for practical purposes, is the direction. The more a human decides how the thing is expressed, the more of it the human owns. That is the same rule that applied to a ghostwriter, a stock photo, or a template before AI existed.

Practical takeaway

Treat the human editing pass as the step that creates your copyright, not a formality after the real work. Save the raw AI output, save your edited versions, and write a one-line note of what you changed and why. If an article is worth owning, it is worth being able to show that a person made it yours.