Where the work leaks away
The leak
A patient compares three surgeons and you are not one of the three. A board shortlists four names and yours does not come up. You never learn it happened. There is no rejection, no bounced click, no line in any report. The work simply goes somewhere else.
Why it happens
Someone looks you up. Google returns a directory listing you cannot edit and a page written by people who never met you. They ask an assistant, and it hedges, because nothing it can check confirms what you claim about yourself.
What it costs
Ask yourself what one new client is worth to you over five years. That is the size of a single leak. Then ask how many times a month somebody looks you up and decides quietly. Nobody can give you that number, including me. You already know whether it is zero.
What I do about it
Rank on Google
Pages under your own name, built to rank for the searches that matter: your name, your specialty, your city. Technically clean, fast, structured, indexed and submitted, so you own the first page instead of renting a slot on somebody else's.
Be right inside the assistants
One canonical description of you reconciled across every source, structured data a machine can read, and pages written to be quoted, so ChatGPT, Claude and Gemini answer accurately and can cite where it came from.
Close the leak
The person checking you finds your record, in your words, wherever they look. That is the only part of this anyone controls, and it is the part that decides whether they call.
Why this matters now, and not in five years
Most of your clients still use Google. That is worth saying plainly, because this field is full of people telling you search is dead. It is not. But something has changed underneath 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. It is one in four of the people deciding whether to call you.
A results page hands the decision back to the person. You can be fourth and still win, because they can see everyone and judge for themselves.
An assistant states a conclusion in a confident voice. If it does not name you, nothing on the screen tells the person you exist.
- Search gives ten answers. An assistant gives one. A results page invites the person to judge for themselves. An assistant states a conclusion in a confident voice, and there is no second result underneath it to correct the record.
- You never find out it happened. A wrong answer produces no impression, no click, nothing in any analytics you own. The person simply does not call, and you never learn why.
- The people who check you hardest are the ones you want most. A patient choosing a surgeon, a board vetting a candidate, a client picking counsel for something serious. They do not skim. They look you up properly, and now that includes asking a machine.
This is not a reason to abandon search. It is a reason to make sure both are working from the same accurate record, which is the whole of what I do.
If you are a partner at a law firm, this plays out differently and harder, because the first page for your own name is already owned by directories and legal marketing companies. That version is here: AI visibility for law firms.
The part most people have never checked
Open ChatGPT and ask it about yourself by name. Then ask Claude, then Gemini. Most accomplished people have never done this, and the answer is usually one of three things.
- It does not know who you are. You get a polite refusal, or a generic answer about someone with a similar name.
- It knows an older version of you. A job you left, a title you outgrew, a company that no longer exists.
- It hedges. It describes you but attributes everything to you, because nothing independent confirms it.
The third one is the most damaging and the least obvious. A model that cannot verify a claim will say so, and "according to his own website" reads very differently from a plain statement of fact.
Why this is not SEO
Search ranks pages. A language model answers a question, usually without showing you a list, and increasingly that answer is the only thing a person sees. Being on page one is no longer the same as being the answer.
The mechanics are different too. Search rewards links and authority. Models reward being legible: a clear entity, described consistently in the places a crawler trusts, with the same name, the same role and the same facts everywhere it looks. Contradictions are what make a model hedge.
What the work actually is
Entity consolidation
One canonical description of you, reconciled across your site, LinkedIn, professional directories, and anywhere else you appear. Models compare sources. Disagreement costs you.
Structured data
Machine readable markup that states who you are, what you do, and which profiles are yours, so a crawler does not have to guess from prose.
Source material
Pages written to be quoted: specific, checkable, and organised so an assistant can lift an accurate paragraph instead of assembling one.
Crawler access
The AI crawlers are explicitly allowed, the sitemap is submitted, and the pages are indexed. Plenty of sites quietly block the crawlers they most want.
A measured starting point
I record what each assistant says about you before I touch anything, so there is something honest to compare against later.
Re-checking
Models retrain and re-crawl on their own schedule. Change is not instant and I will not pretend otherwise. I check again rather than declare victory.
What I will not claim
I cannot make a model say a particular sentence about you. Nobody can, and anyone selling that is selling something they do not control.
What is controllable is the input: whether the facts exist in a form a machine can read, whether they agree with each other, and whether the crawlers are allowed in. That is the work. The rest is the model's decision.