Firms asking about this are asking one of two unrelated questions, and conflating them is why the answers they get are unsatisfying. The first is whether to use these tools to produce marketing material, which is a supervision and confidentiality problem. The second is how to be the firm an answer engine cites when somebody asks it a legal question, which is a content and evidence problem. Both are real; almost nothing transfers between them.
A generated page is still a communication about legal services
Every rule your jurisdiction applies to firm advertising, claims about results, comparisons, specialisation language, required disclaimers, applies identically to text a machine drafted. There is no category of publication that is exempt because of how it was written. That makes review by a lawyer a required step rather than a quality preference, and it is the single most common gap in firms that adopted these tools quickly.
Client information must never reach a general-purpose tool
Pasting matter details, documents or client names into a consumer AI service to draft a case study or a post is a confidentiality problem regardless of the provider's retention policy. The safe pattern is to write about the type of matter in general terms and to use tools under an agreement that covers your obligations. Decide the rule, write it down, and tell everyone who produces content.
Being cited by an answer engine rewards specificity and sources
Answer engines assemble responses from material they can quote, so pages that state something concrete, attribute it to the statute, rule or decision it rests on, and are updated with a visible date are the material they can use. Generic practice-area pages that could belong to any firm give them nothing to cite. This is the same discipline good legal writing already has, applied to the website.
Expect fewer clicks even when you are the source
A person who gets their answer in an AI response often does not visit the page it came from, so the value of being cited is authority and eventual enquiry rather than traffic. That changes what you measure: mentions and citations, direct enquiries naming something you published, and consultations booked, rather than sessions. Firms judging this work on visits will conclude it failed while it is working.
Where these tools genuinely save a firm time
Summarising, reformatting, drafting a first version of something a lawyer will rewrite, and producing the many small pieces of copy a website needs. What they should not do unsupervised is state the law, describe outcomes, or write anything a prospective client could read as advice. The useful rule is that the machine can produce a draft and a person remains the author.
Questions people ask about AI marketing for law firms
Can a law firm publish AI-written content?
Yes, subject to the same advertising rules as anything else the firm publishes, and with a lawyer reviewing it before it goes out. How the text was produced does not change what the rules require.
Can we put case details into an AI tool?
Not into a general-purpose consumer service. Write about matter types in general terms, and use tools under an agreement that covers your confidentiality obligations. Decide the rule explicitly and tell everyone who writes.
How do we get cited by AI answer engines?
By publishing specific, attributable statements with the authority they rest on and a visible date. Generic practice-area pages offer nothing quotable, which is why firms with substantial written answers are cited and firms with brochure pages are not.
What should we measure?
Citations and mentions, enquiries that name something you published, and consultations booked. Sessions will understate this work badly, because a reader who got the answer in the AI response often never visits the page.