AI Website Content and Copyright Laws

Updated October 27, 2026. This article covers U.S. copyright law and practical publishing considerations for law firms and marketing agencies. It provides general information, not legal advice about a particular work or dispute.

An AI tool can produce a practice area page in seconds. Before that page goes on your website, someone still has to answer a few important questions. What did a person actually create? Who owns the rights to that work? Does the finished page reproduce someone else’s protected material?

Those questions deserve separate answers. A decision about an AI company’s training data does not clear every page its product generates. And a provider’s permission to use an output does not establish that you own an enforceable copyright in all of it.

At Blue Seven, our work starts with research and human judgment. We care about this issue because law firms and agencies are investing in content they expect to use, maintain, and stand behind for years. Here is where the law stands and what we recommend checking before publication.

In this article

• Separate authorship from infringement
• What human authorship means for AI website content
• Who owns the content your firm commissions
• What AI training cases do and do not resolve
• Check the actual content before publishing
• What law firms need to know beyond copyright
• A practical publishing workflow for firms and agencies
• Build website content you can stand behind

Separate authorship from infringement

There are three issues to keep straight:

1. Copyrightability and ownership. Is there protectable human expression in the page, and who owns it?
2. Acquisition and training. Did the AI developer have a lawful basis to obtain and use the material involved in training?
3. The published output. Does the text, image, or other material you plan to publish infringe someone else’s rights?

The Copyright Office addresses the first and second issues in separate parts of its AI report. Its training report also discusses the possibility of infringing outputs. A page can lack protection for its machine-generated passages and still contain protected material copied from someone else. Treating an output as uncopyrightable is not a clearance process.

What human authorship means for AI website content

U.S. copyright law protects qualifying original expression fixed in a tangible medium. It does not protect ideas, systems, or methods simply because someone describes them on a website. Section 102 of the Copyright Act supplies that starting point.

The Thaler decision now includes an appeal and a denial of review

Stephen Thaler sought registration for A Recent Entrance to Paradise, naming his Creativity Machine as the sole author. On March 18, 2025, the D.C. Circuit affirmed the refusal to register the work, holding that the Copyright Act requires human authorship.

The Supreme Court declined to review the case on March 2, 2026. That left the appellate ruling in place. A denial of review is not a Supreme Court decision on the merits.

Thaler involved a work submitted as entirely machine-authored. The D.C. Circuit did not decide how much human involvement would be sufficient in a different case, and it recognized that using AI does not categorically disqualify a work from protection.

Human contributions still matter

The Copyright Office’s January 2025 copyrightability report explains that protection may cover original human expression that remains perceptible in a work, creative modifications, or sufficiently creative selection and arrangement. Protection does not automatically extend to the underlying machine-generated material.

The report also distinguishes assistance from authorship. Using AI to brainstorm before independently writing an article need not prevent protection. Under the technology the Office examined, prompts alone generally did not provide enough control over the resulting expression to make the user its author. The assessment is specific to the work and the person’s actual contribution.

For a law firm practice area page, our recommendation is straightforward: have a writer develop the explanation from verified sources, add relevant examples, and make substantive editorial decisions. Keep the drafts that show that work. Adding a byline or changing a few words should never be your team’s basis for promising ownership of an entire AI draft.

Be accurate if you apply for registration

The Office’s registration guidance directs applicants to identify their human contributions and disclose and exclude AI-generated material that exceeds a minimal, or de minimis, amount. An assistive use of AI does not automatically require a disclaimer. If an application or registration omitted relevant AI-generated material, the guidance explains how to correct the record.

Who owns the content your firm commissions

Copyright ownership generally begins with the author or authors, subject to rules such as work made for hire. Copyright ownership can also be transferred. Sections 201 and 204 explain those rules, including the signed-writing requirement for transfers other than those occurring by operation of law.

Paying for a freelance article does not, by itself, answer every ownership question. The work-made-for-hire rules distinguish employees working within the scope of employment from certain specially commissioned works that meet additional requirements.

We recommend that a firm and its agency settle the following in their written agreement:

• Which rights the client receives and when they transfer.
• Whether the work includes licensed images, stock material, or other third-party assets.
• Whether generative AI may be used, for which tasks, and what the writer must disclose.
• Who checks sources, confirms permissions, and approves publication.

Review the AI provider’s applicable terms as well. A contractual allocation of rights cannot supply missing human authorship or grant rights in someone else’s protected material without the necessary authority. Any promise about output ownership should be read alongside the copyright rules and the actual work delivered.

What AI training cases do and do not resolve

Fair use is a fact-specific analysis. Section 107 directs courts to consider the purpose and character of the use, the nature of the copyrighted work, the amount and significance used, and the effect on the work’s potential market or value. Commercial use matters, but it does not resolve the analysis by itself.

The Copyright Office’s Part 3 report on generative AI training considers how those factors may apply. As of October 27, 2026, the Office still labels it a pre-publication report, first released May 9, 2025. It is policy analysis, not a new statute or a court ruling that clears every training use.

Two developments help explain why the details matter:

• Bartz v. Anthropic. In June 2025, the Northern District of California found the training use before it to be fair use, while treating the acquisition and retention of pirated central-library copies separately. The June 23 order did not approve all copying associated with AI. On July 20, 2026, the court approved the parties’ settlement. That settlement does not establish that every AI training practice or output is lawful.
• Thomson Reuters v. ROSS Intelligence. On September 29, 2026, the Third Circuit affirmed partial summary judgment concerning Westlaw headnotes, including the rejection of ROSS’s fair-use defense. The case involved a competing, non-generative legal-search tool. The court expressly distinguished generative systems, so this decision should not be presented as a blanket ruling about AI-written websites.

For your publishing team, the practical lesson is to review the material you actually intend to use. A favorable training decision for one provider is not a substitute for checking your page.

Check the actual content before publishing

A prompt asking a tool to rewrite a competitor’s article deserves a different review from a writer explaining a legal issue from primary sources. Look for copied passages, distinctive examples, images, and other material that may belong to someone else. The Copyright Office’s training report recognizes that outputs can reproduce protected expression.

Similarity alone does not answer every infringement question. Copyright protects expression rather than the underlying idea, and the analysis can depend on what was copied and whether a license or legal exception applies. A topic shared by two law firms is different from one firm reproducing the other’s original explanation.

There is no universal safe percentage or word count for borrowing. The Copyright Office’s fair-use FAQ emphasizes that the circumstances matter. We recommend using similarity checks to flag passages for human review, rather than treating a clean score as legal clearance.

The same care applies to visuals. Confirm the source and permitted use of an image instead of assuming that an AI label, search result, or vendor download gives your firm every right it needs.

What law firms need to know beyond copyright

Professional responsibilities still apply

ABA Formal Opinion 512, issued July 29, 2024, addresses lawyers’ use of generative AI. It discusses competence, confidentiality, independent verification, and supervision. It also explains why lawyers need to assess a tool’s terms and data practices before entering information related to a representation.

The opinion interprets the ABA Model Rules. Firms must check the rules and guidance that apply in their own jurisdictions. The ABA is not a single nationwide disciplinary authority for every lawyer.

For website content, check the accuracy of legal statements and claims about the firm. ABA Model Rule 7.1 addresses false or misleading communications about a lawyer’s services. Your jurisdiction’s advertising rules should guide the review of results, credentials, testimonials, and other marketing claims.

Search guidance does not replace a rights review

Google’s guidance on generative AI content recognizes that AI can assist research and structure. It calls for accuracy, quality, relevance, and manual verification, including metadata. Its scaled-content abuse policy addresses large-scale content created primarily to manipulate rankings rather than help people, regardless of how it is produced.

Google does not promise rankings because a human wrote a page. Its people-first content guidance emphasizes useful, trustworthy work and clear authorship. Our recommendation is to meet those editorial expectations while separately checking copyright and professional responsibilities.

A practical publishing workflow for firms and agencies

Here is the process we recommend for a new page or a substantial refresh. These are editorial safeguards, not a guarantee against a claim.

1. Start with the brief and the sources. Identify the reader, relevant jurisdiction, purpose of the page, and questions it needs to answer. Use current statutes, court opinions, and official guidance for legal claims.
2. Set the boundaries for AI use. Agree on approved tools and tasks before work begins. Keep confidential client or matter information out of unapproved systems. Escalate questions about consent or data handling to the responsible lawyer.
3. Document the human work. Retain research notes, drafts, and substantive revisions. Be able to explain who wrote the final passages, developed the examples, and organized the page.
4. Review borrowed material. Check quotations, close paraphrases, images, and licenses. Investigate suspicious matches. If the team cannot establish a sound basis for using material, replace it or obtain appropriate advice or permission.
5. Verify the law and the firm’s claims. Open every cited source. Check dates, jurisdiction, case posture, and whether a statement describes an allegation, a ruling, or an appeal. Have the responsible lawyer review legal accuracy and applicable advertising requirements.
6. Approve the complete page. Review the title, headings, author information, links, image descriptions, and metadata along with the body. Record who approved publication. Do not label a page attorney-reviewed unless that review occurred.
7. Keep the record and maintain the page. Save the approved version, source list, rights documentation, and review date. Revisit time-sensitive claims when the law or the firm’s services change.

If you already have AI-assisted pages online, start with the ones containing substantial borrowed text, uncertain image rights, unsupported legal statements, or confidential information. An organized review is more useful than assuming that every page is safe or that every page has to go.

Build website content you can stand behind

The legal questions will continue to develop. Your publishing standards can be clear today. Know who created the content, verify the sources, understand the rights you receive, and give a qualified person responsibility for the final review.

That approach fits the work we do with law firms and marketing agencies. If your website needs original writing or a careful content refresh, talk with Blue Seven about the pages, audience, and review process your team needs.

Written by Allen Watson and Victoria Lozano, Esq., Founders of Blue Seven Content