Answer-first: the same advertising rules still apply
AI search is not a separate legal category with its own rulebook. When ChatGPT, Gemini, Google AI Overviews, or Perplexity summarize a law firm's website and repeat it back to a prospective client, the underlying content is still subject to the same attorney advertising rules that apply to a brochure, a billboard, or a firm's own webpage. The risk is not the technology; it is publishing content that overstates outcomes or drops a required disclaimer, then having that content amplified by a tool the firm does not control.
How attorney advertising rules touch AI-surfaced content
Attorney advertising rules generally require that public statements about a lawyer's services be truthful, not misleading, and free of unsubstantiated claims about results. These rules were written for print ads and websites, but they apply equally to any content a firm publishes, regardless of where that content later shows up. When an AI tool pulls a sentence from a firm's "About" page or a case-result summary and presents it as a direct answer to a searcher's question, it is surfacing the firm's own words to a wider audience than the firm may have anticipated. That does not create new liability, but it does raise the stakes on getting the underlying page right the first time, since the firm cannot edit what an AI engine has already generated in a prior response the way it can edit its own website.
Avoiding claims engines might overstate
AI tools summarize and compress. That process can strip context from a carefully worded sentence, turning "results vary by case" into a flat statement of what a firm "does" for clients. The safest practice is writing outcome-related content so plainly that no reasonable summary can misrepresent it. Avoid phrases like "we win every case" or unqualified references to settlement amounts, and pair any mention of past results with the same qualifying language every time, so that no version of the sentence exists without its caveat attached.
Superlatives are the most common source of this problem. Language such as "best," "top-rated," or "guaranteed" is often restricted or requires substantiation under advertising rules, and it is also exactly the kind of punchy phrase an AI summary is likely to lift verbatim. If a claim would need a footnote to be accurate, it is safer to rewrite it as a factual, unqualified statement about what the firm does (areas of practice, years handling a matter type, or client service approach) rather than a comparative claim about how good the firm is.
Keeping disclaimers and accuracy intact
Disclaimers only work if they survive editing, formatting, and now, AI summarization. A disclaimer buried in a footer or a separate "legal notices" page may never be read by a person and cannot be reconstructed by an AI engine summarizing the main content of a page. If a firm needs a phrase like "no attorney-client relationship is formed by contacting us through this site" or "past results do not guarantee future outcomes" to make a statement accurate and compliant, that language should sit next to the claim itself, not several scrolls away.
The same logic applies to case results and testimonials, which many jurisdictions regulate closely. If a settlement figure or client quote appears on a service page, the required qualifying language belongs in the same paragraph, ideally the same sentence. This does not guarantee that every AI-generated summary will include the disclaimer, but it maximizes the odds, and it also protects the firm's own website from being non-compliant regardless of how the content is later used elsewhere.
A conservative approach to AI visibility
The most reliable answer to AI search visibility is not aggressive marketing language; it is precise, well-organized, factual content that reads correctly whether a human or an AI engine is the one reading it. Firms that treat their website copy as a source of truth (accurate practice area descriptions, plainly stated attorney credentials, disclaimers attached to the claims they modify) reduce ethics exposure while also giving AI tools clean, quotable material to work with. Overstated marketing copy is now a double liability: it risks a bar complaint and it risks being repeated, out of context, to a much larger audience than a single webpage visitor.
A useful practice is reviewing existing website content the same way a bar association reviewer would: line by line, asking whether each claim is substantiated, whether each disclaimer is attached to the statement it modifies, and whether any sentence could be quoted in isolation and still be accurate. Content that passes that test is content that works safely no matter how it is surfaced, summarized, or repeated by a search tool the firm does not operate.
Which of your existing pages is already doing this work
Some content a law firm already has online is doing more of this compliance-safe visibility work than others, and it is worth checking which. Client reviews that mention specific practice areas in the reviewer's own words tend to be quoted accurately because they are first-person statements, not firm claims, which sidesteps much of the advertising-rule risk tied to superlatives. FAQ sections that answer plain questions ("What does a personal injury attorney do during a claim?") in neutral, factual language are also low-risk and highly quotable, since they describe a process rather than promise an outcome.
Service pages are the highest-risk and highest-reward asset: they are where outcome language and disclaimers most often collide, so they deserve the closest line-by-line review described above. Photos and attorney bio pages carry the least ethics risk but also the least AI-search value on their own, since there is little factual claim for a tool to summarize. The fastest way to check which asset is carrying the most weight is to search a specific question a client might ask (a practice area plus "how does" or "what happens") in an AI search tool and see which page, if any, gets pulled into the answer. Whatever gets quoted first is the page that most needs the disclaimer and accuracy review described here.