One. The referral that never arrives
Problem
A friend hands someone your name. They do not call. They search it first. Google returns a directory profile you cannot edit, a review page written by strangers, and two competitors paying for the slot above you. Nothing on that screen is yours, and nothing on it was written by you.
Action
Pages under your own domain, built to rank for the searches that actually happen: the firm name, the partner names, the practice area, the city, and the plain question the client is typing at midnight. Technically clean, fast, structured, submitted and indexed.
Result
The person checking you finds your own words first, in your own voice, on property you own. The referral survives the search instead of dying quietly inside it.
Two. The shortlist you were never on
Problem
A prospective client opens ChatGPT and asks which firms handle their matter in your city. It names three. If nothing checkable confirms who you are and what you take on, you are not one of the three. No impression, no click, no line in any report. The work simply goes somewhere else.
Action
One canonical description of the firm reconciled across your site, the partner profiles, LinkedIn, the bar listing and the directories. Structured data a machine can read. Pages written to be quoted rather than skimmed. And the AI crawlers explicitly allowed in, which a surprising number of firm sites quietly block.
Result
When an assistant is asked, it has something verifiable to work from, so it describes your practice accurately instead of hedging or reaching for whichever firm it can confirm.
Three. The version of your firm that is out of date
Problem
Assistants answer from whatever they have. Often that is an old firm name, a partner who left three years ago, a practice area you dropped, or a bar profile nobody has touched since it was created. It reads as fact, because it arrives as one confident paragraph with no second result underneath it to correct the record.
Action
I record what each assistant says about the firm before I touch anything, fix the contradictions at the source rather than arguing with the model, and check again once the corrected sources have been crawled.
Result
An honest before and after: what they said then, what they say now, and which sources changed. Not a promise about wording, because nobody controls that. A documented change in the inputs.
The whole thing on one page
Four stages. The first one is already happening to your firm today, whether or not anyone has built anything.
Someone looks you up
A referral, a board vetting counsel, or a plain question typed into an assistant at midnight. You are not in the room.
They check the sources
Your site, your bar listing, LinkedIn, the directories. Right now those disagree with each other, and disagreement is what makes a model hedge.
I reconcile the record
One canonical description, structured data a machine can read, pages written to be quoted, and the AI crawlers explicitly allowed in.
The answer names you
Google ranks your page instead of a directory. The assistant describes your practice accurately and can point at where it got it.
Stage three is the only part anyone controls. Stage four is the search engine's decision and the model's decision, which is why I will show you what they decided rather than promise you a sentence.
Why this is more urgent for lawyers than for most professions
Two things make legal different. The first is that the decision is high stakes and low frequency. Nobody picks a lawyer casually, and almost nobody has done it recently, so they research harder than they would for anything else. The second is that a large part of the first page for your name is already owned by somebody else: directories, review aggregators and legal marketing companies who out rank you on your own name and sell the slot beside it.
Those figures come from a survey of 1,110 American adults run in March 2026 by Orbit Media. One in four is not everybody. It is also not nobody, and it is one in four of the people deciding whether to call you.
- Search gives ten answers. An assistant gives one. A results page invites the client to judge for themselves. An assistant states a conclusion, and there is no second opinion on the screen.
- You never find out it happened. A wrong or missing answer produces nothing in any analytics you own. The client simply does not call.
- The clients who check hardest are the ones worth most. A company choosing counsel for something serious does not skim. They look you up properly, and that now includes asking a machine.
The check that takes four minutes
Before you speak to anyone about this, including me, do it yourself. Open ChatGPT, then Claude, then Gemini, and ask each one three questions.
- Who is your name, the attorney in your city? See whether it knows you at all, and whether the version it knows is current.
- Which firms handle your practice area in your city? See whether you are named, and who is named instead of you.
- Is your firm any good? See what it reaches for, and whether it can cite anything you actually wrote.
Most partners have never run this. The answer is usually one of three: it does not know you, it knows an older version of you, or it describes you but attributes every claim to your own website, which reads very differently from a plain statement of fact. The third one is the most damaging and the hardest to notice.
If you want the longer version of that test, the free self audit walks through it in ten minutes, and there is no step eleven you can only get by paying me.
Questions partners ask
- Why does it matter what ChatGPT says about my law firm? A search results page gives a prospective client ten options to judge for themselves. An assistant gives one answer in a confident voice, with no second result underneath it to correct the record. If nothing verifiable confirms who your firm is and what it handles, the assistant either hedges or names someone else, and you never find out it happened.
- Is this the same as SEO for lawyers? It overlaps but it is not the same. Search ranks pages and rewards links and authority. Language models reward legibility: a clear entity described consistently across the sources a crawler trusts, with the same firm name, the same practice areas and the same facts everywhere. Contradictions between your site, your bar listing and the directories are what make a model hedge.
- Can you guarantee my firm will rank first on Google? No, and nobody can. What is controllable is whether the facts about your firm exist in a form a machine can read, whether those facts agree with each other, and whether the crawlers are allowed in. That is the work. The ranking is the search engine's decision.
- Does this handle bar advertising compliance? No. Attorney advertising is governed by the rules of the jurisdictions you practise in, and responsibility for compliance stays with the firm. I build the technical visibility layer and I will follow any disclaimer, wording or review requirement you give me, but I do not advise on the rules themselves.
What I will not claim
I cannot make a model say a particular sentence about your firm, and I cannot promise a position on Google. Nobody can, and anyone selling either is selling something they do not control.
I also do not advise on attorney advertising rules. What you may say about results, specialisation and client outcomes is governed by the jurisdictions you practise in, and responsibility for that stays with the firm. Give me your wording requirements and disclaimers and I will build to them. I will not write around them.
What is controllable is the input: whether the facts about your firm exist in a form a machine can read, whether they agree with each other across every source, and whether the crawlers are allowed in. That is the work. The rest is the model's decision, and I will show you what it decided.