A growing share of the questions that used to be typed into a search box are now asked of an assistant instead. "Who does emergency plumbing near me and are they any good" used to produce a page of links to evaluate. Increasingly it produces an answer: a couple of named businesses, a sentence about each, and no obligation to click anything.
That shift has produced a lot of new vocabulary and a certain amount of nonsense. It is worth separating what has actually changed from what is being sold.
What has genuinely changed
Three things, concretely.
First, the answer is now assembled rather than listed. Something is reading a set of sources and writing a summary. Your business is either in that summary, described correctly, or it is not.
Second, the description is generated from whatever the system can read about you. It is not choosing from your marketing copy. It is synthesising from your site, your listings, your reviews, directories, news, and anywhere else your name appears. Gaps get filled with inference, and inference is where things go wrong.
Third, the visitor may never see your site at all. They may act on the summary — call the number in it, or simply take the recommendation. That raises the cost of the summary being wrong, and it lowers the value of a website that is beautiful but unreadable to software.
What AI visibility work actually consists of
It is less exotic than the name suggests. In practice it is four things:
- Finding out how you are currently described. Ask the assistants about your business and your category the way a customer would, and read the answers. This is free and most owners have never done it.
- Making the facts unambiguous. What you do, where, for whom, hours, service area, how to reach you — stated in text, in structured data, and identically everywhere you control.
- Answering the real questions in writing. Assistants summarize sources that address the question. If nothing on the open web answers "does anyone in this area do same-day repairs", nothing can be summarized in your favor.
- Fixing contradictions. Two different phone numbers, a closed location still listed, an old business name — each one is a reason for a system to hedge or to pick the wrong version.
What it is not
This is where the marketing gets loose, so it is worth being blunt about the limits.
- It is not a ranking system you can climb. There is no "position three in ChatGPT". Different people asking the same question get different answers, and the same person can get a different answer twice.
- It is not something anyone can guarantee. Nobody outside those companies controls what a model says, and anybody selling a guaranteed mention is selling something they cannot deliver.
- It is not achieved by writing pages addressed to the AI. Stuffing a site with text aimed at models is the 2005 keyword-stuffing playbook wearing a new hat, and it degrades the page for the humans who do arrive.
- It is not separate from the fundamentals. Almost everything that helps here also helps ordinary search, because the underlying requirement is the same: be clear, be consistent, be readable.
Why it overlaps so heavily with ordinary SEO
People sometimes expect AI visibility to be a separate discipline with its own tooling. Mostly it is not. The systems answering these questions are reading the open web, and the qualities that make a page useful to them are the qualities that made it useful to a search engine: clear structure, real information, consistent facts, machine-readable markup, pages that load.
The differences are in emphasis. Structured data matters more, because it removes the need to infer. Consistency matters more, because a synthesiser reconciling contradictory sources will sometimes reconcile them badly. And directly answering questions matters more, because the output is an answer rather than a list of candidates.
Is this worth doing now, for a small business?
The reasonable answer is: the work is worth doing, and it would have been worth doing before AI assistants existed. Almost none of it is speculative. Accurate structured data, consistent business details, pages that answer real questions and a site that machines can read are foundations, not a bet on a trend.
What has changed is the cost of neglecting it. Previously an inconsistent set of facts cost you a little relevance. Now it can produce a confident, wrong sentence about your business, delivered to somebody who was ready to hire you.