AI will decide which businesses exist.

The temptation, when something new arrives, is to treat it as a tactic. SEO 2.0. A new channel. A box to tick.

That instinct will be expensive.

AI selection is not a channel. It is a layer of judgement that sits above every channel. It mediates discovery, comparison, and recommendation across every category, in every market, for every buyer who reaches for an assistant before reaching for a search engine.

And those buyers are the ones with the highest intent. They are the ones already past the browsing stage. They are the ones asking, in plain language, who should I use? The answer they receive will frequently be the only answer they consider.

You can spend a fortune on the channels underneath and still lose the moment of decision, because the decision no longer happens inside your channels. It happens inside the model, before your prospect ever sees your website, your advert or your carefully written homepage.

AI Visibility Optimisation. We chose the word with care. Not marketing. Not optimisation in the SEO sense. Visibility in the literal sense: the engineering of how an AI system perceives, classifies, verifies and recommends your business.

It rests on four pillars. Interpretability - making your business legible. Coherence - making your signals agree. Authority - making your claims independently verifiable. Intent alignment - making your structure match the questions buyers ask.

None of these pillars are speculative. Each is measurable. Each can be moved. The businesses that move them deliberately, now, will define the shape of their categories for the decade ahead. The ones that wait will inherit whatever the AI's classification of them happens to be on the day someone asks.

It helps to demystify the picker, because fear of the unknown is a poor foundation for strategy. When an assistant recommends a business, it is assembling an answer from what it can retrieve and trust. That assembly follows a handful of deeply unglamorous rules.

First, retrieval. If your business is not described clearly enough - on your own site, in directories, in trade press, in reviews, in the places machines learn from - it never enters the candidate set. You cannot be recommended from a document that was never written.

Second, consensus. The model looks for agreement. Does the trade body list you? Do the case studies match the claims? Does what your clients say resemble what you say? Where signals agree, confidence rises. Where they conflict, the model hedges - and hedging means someone else gets named.

Third, specificity. Generic excellence is invisible. "Leading provider of innovative solutions" gives the machine nothing to hold. "Bristol firm that handles probate for farming estates" gives it everything: a category, a place, a buyer, a moment.

Fourth, recency. Models weight the living web. Signals that stopped being refreshed years ago read as a business that stopped being relevant years ago.

Notice what is absent from that list: tricks. There is no tag to inject, no hack to run. There is only the patient, deliberate construction of a business that is easy to read, easy to verify, and easy to believe in. Which, it turns out, is also what human buyers want.

The cruelest feature of this shift is that exclusion leaves no trace. In search, you could watch your position slip and react. There is no equivalent dashboard for the answer you were not named in.

A prospect asks an assistant for a recommendation in your category, in your city, this afternoon. Your competitor is named. Work is won. You were never told the conversation happened. There is no lost-lead email, no bounce rate, no warning. The loss is recorded nowhere, because as far as anyone can measure, nothing occurred.

This is why most businesses will not act in time. Not out of scepticism, but because the damage does not announce itself. By the time a decline becomes visible in revenue, the selection layer has been forming its picture of the category for quarters, and the businesses named early have accumulated the corroboration that keeps them named.

Selection compounds. Early presence generates mentions; mentions generate corroboration; corroboration generates presence. The gap between the named and the unnamed widens on its own. Every month of inaction is not a neutral month. It is a month in which your competitor's case got stronger and yours did not.

AVO is the practice. We measure your visibility. We diagnose your gaps. We restructure your entity. We construct the corroborated authority signals AI systems require. We verify the result against live AI responses, by name, by query, week after week.

We work in fourteen-day cycles. We report in plain numbers. We do not promise ranks; we engineer probabilities. We do not optimise for clicks; we optimise for being chosen.

And we publish what we learn. The audits. The methodology. The insights. Because the entire UK economy is going to have to learn this in the next thirty-six months, and it should not have to learn it the hard way.

You do not have to take any of this on faith. Ask the question yourself. Open ChatGPT, Gemini, Perplexity or Claude and type what your best customer would type: the thing they need, in the place you serve, in the words they would actually use.

Read the names that come back. If yours is there, ask how secure it is, and what happens when a better-documented competitor arrives. If it is not there, understand that this is precisely the moment your most informed buyers are forming their shortlists - and that they are doing it without you.

Either way, you will have learned more in thirty seconds than most businesses learn this year. And if you want help changing the answer, that is the work we do, every day, for UK businesses that intend to be named.