AI generated content copyright 2026: the Supreme Court left AI-only work unprotected, but edited, arranged human writing still qualifies. What to log now.
The Supreme Court didn't write new AI copyright law on March 2, 2026, but it did settle the central question in AI generated content copyright 2026: whether a work with no human author can hold a copyright at all. It denied certiorari in Thaler v. Perlmutter, meaning it declined to hear the case, which left the D.C. Circuit's ruling exactly where it was: the Copyright Act requires a human author, and a work with no human author can't be registered, per Baker Donelson's summary of the denial. No new opinion, no new test, no new statute. The rule that was already in place is now settled for the foreseeable future, because the country's highest court had the chance to revisit it and chose not to.
That sounds like a story for patent lawyers and AI researchers, not a SaaS blog. It isn't. If your blog runs unedited AI output straight to the CMS, the practical read of this ruling is that those posts may not be protectable intellectual property at all. A competitor can lift them, word for word, and you have no copyright claim to stop them with. This is the same territory covered in does Google penalize AI content, and it's worth separating the two questions early: ranking and ownership are not the same risk, and this post is about the second one. It's also distinct from the labeling duties in the EU AI Act's Article 50 and FTC AI disclosure rules, which govern whether you have to say a post used AI. This is about whether you own the words at all.
The Court's non-decision preserves a specific, narrow rule: a work with zero human authorship can't hold a copyright. It does not say AI-assisted work is unprotectable, and it does not touch the much larger, unresolved question of how much human involvement is enough. That gap is exactly where most AI-assisted blog content lives.
Stephen Thaler filed for copyright on an image titled "A Recent Entrance to Paradise," listing his AI system DABUS as the sole author and disclaiming any human creative contribution whatsoever. The D.C. Circuit's reasoning rested on textual features of the Copyright Act that all presume a human on the other end: ownership provisions assume the author can hold property, duration is measured against the author's lifespan, joint authorship requires shared intent, and registration requires a signature, all capacities only a human possesses, per Baker Donelson's breakdown of the court's reasoning. Thaler's own framing, that no human touched the work at all, made this close to the easiest possible case for the Office and the courts to rule against. Even the Trump administration's Department of Justice urged the Supreme Court to deny cert, framing the case as an unusually narrow statutory question not worth the Court's time, according to Newsweek's coverage of the denial.
Law professor Dennis Crouch called the practical effect of the ruling "a vast new public domain," pointing out that AI-only works now enter the public domain immediately on creation, since no one holds a copyright in them to license or restrict, per the same Newsweek piece. That phrase is worth sitting with if your blog publishes fully automated, unreviewed posts. Nobody owns them, which means nobody can stop a competitor from republishing them wholesale.
Thaler's case never tested the question that actually matters for a working blog: how much human control is enough. The D.C. Circuit's ruling explicitly does not prohibit copyright for a work "made by or with the assistance of artificial intelligence," it only rules out a work with no human author, per Baker Donelson's analysis. Allen v. Perlmutter is the case built to test that distinction, and it's still working through the courts.
Jason Allen entered prompts into Midjourney at least 624 times, specifying genre, tone, palette, and composition, used the tool's variation and upscaling features, and hand-edited the final image in Photoshop. The Copyright Office still ruled that the AI's contribution had to be disclaimed, reasoning that the model, not Allen, determined the final expressive output regardless of how much iteration went into the prompts, per Mondaq's coverage of the case. A ruling from the District of Colorado was expected in the second or third quarter of 2026. Whatever it decides, the case already answers the question most content teams ask first: no, prompt volume alone doesn't buy you authorship.
The rule that actually governs a working blog isn't "was AI involved." It's "did a human make the expressive choices that ended up on the page." The Copyright Office's Part 2 report on AI and copyrightability, published January 2025, draws that line directly: AI-generated material alone isn't copyrightable, but a work combining human and AI-generated elements can be registered for the human-authored portions, including a human's selection, coordination, and arrangement of AI output, according to the Copyright Alliance's summary of the report.
The Office's own language is unambiguous: "Simply providing prompts, even detailed ones, does not constitute human authorship," per the Copyright Office's Part 2 report as summarized by Ludwig IP Law. This is the finding that trips up teams who assume a long, carefully engineered prompt earns the same protection as writing the paragraph yourself. It doesn't. Allen's 624 prompts and Thaler's zero prompts land in the same bucket for this specific purpose: neither one is the kind of human control over expression the Office is looking for. A detailed content brief handed to an AI writer is closer to a prompt than to authorship, on its own.
What does count is a human's actual editorial hand on the output: rewriting sentences, restructuring sections, cutting and adding claims, choosing which facts to keep and how to arrange them, adding original analysis or a first-hand result the model couldn't have generated. That's the same standard this blog already argues for on ranking and trust grounds, not just legal ones. As of April 2026, the Copyright Office had issued more than 6,000 registrations for works that combine human and AI-generated material, per Ludwig IP Law's coverage of the milestone, proof that the human-plus-AI combination is a normal, registrable category, not an edge case nobody bothers with. Registration applicants have to identify which parts of a work are AI-generated and disclaim them specifically; the Office has cancelled registrations over undisclosed AI content, so the disclosure step isn't optional paperwork, it's what keeps the registration valid.
No. Copyright status and Google's ranking treatment are two entirely separate systems answering two entirely different questions, and conflating them is the most common mistake in this whole conversation. Google's own guidance on AI-generated content states plainly that "using automation, including AI, to generate content with the primary purpose of manipulating ranking in search results" violates its spam policies, while "rewarding high-quality content, however it is produced" doesn't depend on authorship at all, per Google's Search Central guidance. Nothing in that policy references copyright ownership, and nothing in copyright law references search rankings.
Ranking risk is about whether Google thinks your page helps the person who searched. Public-domain exposure is about whether anyone, including you, holds an enforceable right to stop someone else from copying your page. A post can carry zero ranking risk and full public-domain exposure at the same time: it ranks fine because it's genuinely useful, and a competitor can still republish it verbatim tomorrow with no copyright claim available to you, because no human ever touched the draft.
The data on the ranking half of this is well established. Across 600,000 pages, Ahrefs found a correlation of just 0.011 between a page's AI-content share and its ranking position, effectively noise, and 86.5% of top-ranking pages already contained some AI-generated content, per Ahrefs' own analysis. Our own does Google penalize AI content breaks that dataset down in full, including the four edits that separate a page that ranks from one that gets filtered as scaled, unhelpful content. None of those four edits were ever about copyright. They're about whether the page helps a reader. The ownership question sits entirely on top of that, unaffected by how well the page performs.
If your blog already runs on a pull-request workflow, most of what the Copyright Office looks for is stuff you're already producing, you're just not saving it anywhere. The fix isn't a new process. It's logging the process you already have before you merge, the same discipline this blog applies to fact-checking every claim before a post ships and to keeping an audit trail for AI content governance more broadly. It's also the default output of a PR-based writer like Lyra: every draft ships as a pull request with the diff against the raw output, the reviewer, and the edit history already attached, so the authorship record exists whether or not you ever go looking for it.
A pull request already captures a diff between the raw draft and the merged version, a reviewer's approval, and a timestamp. That's most of the record the Copyright Office would ask you to reconstruct after the fact if a registration or an infringement claim ever depended on it. Before merging an AI-assisted post, make sure the PR shows:
A pull request is also where you'd already be checking the disclosure duties in Article 50 of the EU AI Act or FTC AI disclosure rules if either applies to your audience. Those are separate obligations from copyright, about telling readers a post used AI, but the record-keeping habit is identical: log it in the PR once, and it answers three different compliance questions instead of one.
Most blog posts never need a formal copyright registration. US copyright exists automatically the moment a qualifying work is created; registration is a separate, optional step that mainly matters if you'd need to sue someone over it. Federal registration is a prerequisite for filing an infringement lawsuit in the US, and it unlocks statutory damages and attorney's fees that aren't available otherwise. That makes it worth the filing fee for content that's genuinely valuable to defend: a flagship guide that took real original research, a proprietary framework or dataset you built the post around, anything you'd actually go to court over if a competitor copied it outright.
For routine posts, the practical goal isn't a registration certificate. It's a PR history clean enough that, if it ever mattered, you could point to specific commits and reviewers and show exactly which parts of the post were human-authored. That record is worth more than most teams realize, and it costs nothing extra to keep if the review step already exists.
A pull-request workflow that fact-checks every claim before merge is also the record that proves a human wrote your posts, not just an AI. Lyra opens every draft as a PR you review and merge yourself, so the authorship record and the editorial one are the same diff.
FAQ
No. The Court denied certiorari in Thaler v. Perlmutter, which means it declined to hear the case at all. That left the D.C. Circuit's ruling in place: the Copyright Act requires a human author, so a work with no human author can't be registered. The denial confirms the existing rule rather than writing a new one.
Not on its own. The Copyright Office's Part 2 report on AI and copyrightability holds that purely AI-generated material isn't protectable, but a work combining human and AI-generated elements can be registered for the human-authored portions: substantive edits, and the selection, coordination, and arrangement of AI output. Prompting alone, however detailed, doesn't count as that human contribution.
No, and the two questions have nothing to do with each other. Google's ranking systems judge whether a page is helpful, not who or what wrote it; an Ahrefs study of 600,000 pages found a 0.011 correlation between AI-content share and ranking. Copyright is a separate legal question about who, if anyone, can stop someone else from copying your post.
There's no prompt count that qualifies on its own. Jason Allen ran more than 624 prompts through Midjourney, used variation and upscaling tools, and hand-edited the result in Photoshop, and the Copyright Office still required the AI-generated image itself to be disclaimed. The line the Office draws is control over expression, not iteration count.
Built by the tool you're reading about
Lyra finds the topics worth ranking for, writes them in your repo's voice, fact-checks every claim, and opens a pull request scored and ready to merge. You review and hit merge. Want to see what she'd write for you? Start free with three posts, no card.
Keep reading
EU AI Act Article 50 content rules took effect August 2026 with fines up to 15 million euros. See when AI blog posts need a label, and when review exempts you.
A trademark risk comparison blog post carries real exposure: nominative fair use limits, disparagement traps, and the legal check to add before you merge.
The best AI content detector claims 99%+ accuracy, but independent tests show 60-96%, with real false positives. Here's why a flagged score isn't an SEO risk.