Credible Roots / GEO for Law Firms
Vertical: legal
AI search for law firms, inside Rule 7.1
Every other business optimizing for answer engines is trying to make a model say flattering things about it. A law firm has a professional conduct rule governing what may be said about its services — and no control over what the model actually outputs. That gap is the whole problem, and most agencies selling AEO to law firms have not noticed it exists.
Not legal advice. We are a search agency, not a law firm, and nothing here is an ethics opinion. Every statement below is sourced and linked, and your own state’s rule governs — it may differ materially from the ABA Model Rule. Run anything here past your ethics counsel or your bar before acting on it.
Answer engine optimization for a law firm runs into a constraint no other vertical has: ABA Model Rule 7.1 provides that a lawyer shall not make a false or misleading communication about the lawyer or the lawyer’s services, and the commentary around it treats statements likely to create unjustified expectations about results, and comparisons with other lawyers that cannot be factually substantiated, as the central risks. An AI answer describing your firm routinely does both — it summarises outcomes and it ranks you against competitors — and you did not write it. The genuinely unresolved question, which we could not find squarely answered in any bar opinion as of September 2026, is whether an AI-generated description of your firm that your own optimization work helped shape counts as a communication you are responsible for under Rule 7.1. More than 35 state bars had issued guidance on AI in legal practice by March 2026, and the consistent principle across it is that responsibility stays with the lawyer rather than transferring to the tool. The practical consequence is that a law firm should optimize for accuracy and verifiability rather than for flattery: seed checkable facts, correct the directories and structured data that models draw on, and avoid creating the superlatives you would not be allowed to publish yourself.
- Rule 7.1 governs what may be said about your services. You do not control the model’s output.
- Superlatives and result summaries are the risk, and they are exactly what AI answers generate.
- Whether that output is “your” communication is unresolved. Ask your bar, not an agency.
- Optimize for accuracy, not for praise. It is the only version defensible either way.
On this page
Why law firms are a different problem
ABA Model Rule 7.1 states that a lawyer shall not make a false or misleading communication about the lawyer or the lawyer’s services, and that a communication is false or misleading if it contains a material misrepresentation of fact or law, or omits a fact necessary to make the statement as a whole not materially misleading. The rule and its commentary treat two categories as the recurring danger: statements likely to create an unjustified expectation about results, and comparisons with other lawyers’ services that cannot be factually substantiated.
Now consider what an answer engine produces when someone asks it to recommend a firm. It summarises outcomes. It ranks firms against each other. It uses words like best, leading and top-rated as a matter of ordinary generated prose. In a marketing brochure those would be the exact sentences your compliance review exists to remove.
This is why generic AEO advice is not safely transferable to a law firm. The standard playbook — publish comparison content, seed superlatives, get listed on “best of” pages, encourage the model to recommend you — is, in a regulated profession, a playbook for manufacturing exactly the statements the rule restricts. Most agencies selling this to law firms have simply lifted the SaaS version.
Note also that the Model Rules bind nobody by themselves. Each state adopts its own version, and the variations are material — some states impose filing requirements, mandatory disclaimers and format rules that the Model Rule does not. Your state’s rule is the one that governs.
The question nobody has answered
Open as of September 2026
If your optimization work shapes what an AI says about your firm, is that output a communication you are responsible for under Rule 7.1?
We looked and could not find a bar opinion squarely answering it. The honest positions available today:
The narrow reading: a communication is something the lawyer makes. A model generating prose is not the lawyer speaking, any more than a journalist’s description of a firm is the firm’s advertising. On this view the rule reaches your website and your ads, and stops there.
The broad reading: if you deliberately engineered the inputs so that a system would produce a favourable claim about your services, the output starts to look like an indirect communication, in the way that a testimonial you solicited and placed is treated as yours in many jurisdictions.
Nobody knows which way this lands, and it may land differently in different states. That uncertainty is not a reason to avoid the work — being invisible to answer engines is its own problem — but it is a strong reason to do the version that is defensible under either reading.
Which, conveniently, is also the version that works better: seeding accurate, specific, checkable facts rather than manufacturing praise. A model that describes your firm correctly is doing something you can stand behind at a bar hearing. A model calling you the best firm in the state is a sentence you could not have written yourself.
What the bars have said so far
Not about this question directly, but about AI in practice generally, and the direction is consistent.
- More than 35 state bars had issued guidance on AI in legal practice as of March 2026, with recurring themes of competence, confidentiality, candour, supervision, reasonable fees and client communication.
- The Florida Bar’s Opinion 24-1 covers core duties including confidentiality, oversight and reasonable fees, and addresses compliance with the advertising rules where AI is used in marketing.
- The State Bar of California first published Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law in November 2023, and its Board of Trustees approved updated revisions in May 2026 reflecting developments including agentic AI. In August 2025 the California Supreme Court directed the State Bar to consider folding that guidance into the formal rules.
- The consistent principle across the guidance on AI-generated advertising is that responsibility does not transfer to the tool: where an AI system produces misleading language or omits a required disclaimer, the risk remains the lawyer’s.
That last principle is the one we would build around. It concerns AI the firm uses rather than AI describing the firm, so it does not settle the open question above — but a regulator that consistently declines to let responsibility sit with a tool is not an obvious candidate to accept “the model said it, not us.”
What we build for a law firm
The same five-part engagement as any other client, with the legal constraints applied at each step. The general version is here.
Access, first
Whether GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and Google-Extended can actually fetch your pages, confirmed from server logs. Firm sites sit behind security products more often than most, and an engine that cannot reach you cannot describe you accurately — it will describe you from whatever else it can reach.
Entity resolution, which for lawyers means the directories
Attorney identity is unusually ambiguous: common names, practitioners at multiple firms, firm mergers and rebrands, and a dense layer of third-party legal directories that models draw on. The work is making the firm and each attorney resolvable to one entity with consistent details, correct practice areas, correct jurisdictions and correct bar admissions — and correcting the records that say otherwise.
Facts, structured to be quoted
Jurisdictions, admissions, practice areas, languages, court experience, published writing, speaking. Specific and verifiable, phrased so a model can lift them without embellishing. This is the substitute for the superlatives the generic playbook would have you seed.
Substantiation kept with the claim
Where a claim could be read as comparative or outcome-based, the substantiation sits next to it rather than in a file somewhere. If a statement cannot be substantiated, it does not go on the page — which is both what Rule 7.1 asks of you and what makes a passage citable.
Monitoring, as a compliance function
A fixed prompt set, re-run on schedule, recording what the major models actually say about the firm. For most clients this is marketing measurement. For a law firm it doubles as an early-warning system: if a model starts asserting a success rate you never published, you want to know before a regulator or an opposing party does. Reported against our published metrics, with the sample size attached.
What we will not do for a law firm
- Seed superlatives. No content engineered to make a model call you the best, leading or top-rated firm. Those are the statements the rule is about.
- Manufacture comparative claims. No “better than [competitor]” framing, substantiated or otherwise. If it is genuinely substantiated, your marketing counsel should approve it, not your search agency.
- Publish outcome summaries as marketing. Case results carry state-specific disclaimer requirements and a live risk of creating unjustified expectations. Not our call to make, and not something to hand to a language model.
- Buy or place directory “awards”. Paid badges that imply peer recognition are a known problem area in several states.
- Guarantee a model names you. Nobody can, in any vertical. The general list of what is not on offer.
- Advise you on your ethics rules. We will flag where the work touches them and tell you to ask someone qualified. That is the entire extent of it.
Who this fits
| A good fit | Not yet | |
|---|---|---|
| Practice | Clients research before they call, often at length | Purely referral-driven, or emergency-response work |
| Substance | Published writing, court experience, defined niches | General practice with nothing specific to state |
| Compliance | Has ethics counsel or a marketing review process | Nobody to sign off on what gets published |
| Appetite | Wants to be described accurately | Wants the model to say it is the best in the state |
| Horizon | Building position over a year | Needs matters this quarter — buy advertising |
Questions firms actually ask
Does Rule 7.1 apply to what ChatGPT says about my firm?
Nobody has definitively answered that. Rule 7.1 governs communications a lawyer makes about the lawyer's services, and an AI answer is generated by a third-party system. The unresolved part is whether output your own optimization work deliberately shaped becomes an indirect communication you are responsible for. We could not find a bar opinion squarely addressing it as of September 2026. Ask your state bar or ethics counsel, and in the meantime do the version of the work that is defensible either way.
Is AEO for law firms even worth doing given the risk?
The risk runs in both directions. A model describing your firm inaccurately, or not at all, is also a problem, and it is the default outcome if you do nothing while competitors publish. The difference is what you optimize for: accuracy and verifiability are defensible under any reading of the rule, whereas engineered praise is the thing that creates exposure.
What if a model says something false about our firm?
There is no field to edit and no vendor can force a correction. What can be done is change the evidence the description is drawn from — correct the directory records, fix the structured data, resolve the entity properly and get accurate statements into sources the models actually retrieve. Then re-measure. It is indirect and slow, which is exactly why monitoring matters: the earlier you see it, the less it has propagated.
Do you write our case results or practice area pages?
We will structure them so they can be quoted accurately, and we will tell you where a phrasing looks like it creates an expectation about results. We will not decide what outcome claims your firm publishes or what disclaimers your state requires. That is your marketing counsel's call, not ours.
Can you get us into the "best lawyers in [city]" AI answers?
We will not try to engineer that, and we would be cautious about any agency that offers to. A superlative you could not lawfully publish yourself is not obviously safer because a model generated it. What we do instead is make the firm accurate, resolvable and quotable, so that where a model assembles such a list from real evidence, you are actually in the evidence.
What does it cost?
$2,000 a month, three month minimum, then month to month. Same price as every other engagement — no vertical premium. The pricing page sets out what it does not buy.
Related reading
Sources
- American Bar Association. Model Rule 7.1: Communications Concerning a Lawyer’s Services.
- American Bar Association. Model Rules 7.1 to 7.3, as amended August 2018.
- The State Bar of California. Ethics and technology resources, including the Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law.
- The State Bar of California. Proposed amendments to the Rules of Professional Conduct related to artificial intelligence.
- Credible Roots. AI search statistics, traced to source. Context for how much traffic these systems actually send.
How this was made: the rule text is taken from the ABA’s own publication of Model Rule 7.1 and the state bar positions from each bar’s own materials, linked above. The open question in the second section is presented as unresolved because we could not find a bar opinion answering it, not because we have a view on the outcome. Nothing here is legal advice or an ethics opinion, and state rules vary materially from the Model Rule. Corrections go in the corrections log. Drafted with AI assistance, reviewed and approved by the named author before publication. Our editorial standards.
We will tell you what the models currently say about your firm
The first call is a diagnostic: whether the engines can reach your site, how the major models describe the firm and its attorneys today, where the directory records disagree, and which of it would concern your compliance reviewer.
Book a 30 minute call