AI search coverage

How we get you answered first in AI search.

Someone asks an assistant for the best supplier in your field, in your town. They get one answer. Maybe three names inside it. There is no page two. If you are not in that answer you do not exist for that person, and you will never see them in your analytics, because they never clicked anything.

That is the problem. Here is exactly what we do about it, in the order we do it.

Three numbers

We measure before we write a word.

Presence rate

Out of every hundred answers, how many mention you at all.

Share of voice

Of every brand named across those answers, how much of the naming is yours.

Visibility score

A 0 to 100 read that weighs where in the answer you appear, the context, and the sentiment around you.

An example from a live account. A charity we work with sits at 28.7 percent presence, so roughly three answers in ten name them. Their share of voice is 47.6 percent against 108 other organisations the models treat as relevant, which puts them first by a distance. Two stories from one account. Very strong when named. Absent seven times out of ten.

Stage one

Where you actually stand

We measure before we write a word. First we build a set of real questions your buyers ask, in the words they use. Not keywords. Questions.

There are nine engines that matter right now. ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity, Microsoft Copilot, Grok, DeepSeek and Claude. Three carry most of the volume for most UK businesses, so a standard baseline tracks ChatGPT, Google AI Overviews and Perplexity every day, and we widen the set when your audience sits somewhere else.

We sample around 150 answers a day, per client. Ask an assistant the same thing twice and you can get two different lists, so a single check tells you nothing. A fortnight of daily sampling tells you the truth.

Stage two

The gap almost nobody looks at

There is a difference between being named in an answer and being the source that answer came from.

One live account showed 52 brand mentions in a fortnight and zero citations for the first several days. The assistants were talking about them using other people's pages. A directory, some reviews, a news piece. Their own site was contributing nothing.

That gap is usually the fastest thing to close, and it is almost never measured.

Stage three

The technical pass

An assistant has to be able to read you before it can quote you. We audit the whole site and score it in three weighted parts. Technical health at 30 percent, content at 35 percent, and answer readiness at 35 percent.

Answer readiness is the one most sites have never been scored on. Structured data, answer format, citation worthiness, semantic richness and machine accessibility.

A real finding. A 358 page site where a run of URLs had been encoded with emoji characters lifted from the page titles. Human browsers coped. Machine readers did not. It had sat there for years.

Stage four

Rewriting to be quotable

Assistants do not lift paragraphs. They lift claims. A claim gets lifted when it is specific, self contained, attributed, and able to survive being pulled out of the page it came from.

So we put the answer in the first two sentences, replace vague copy with figures, dates, places and named people, add structured data, build question shaped pages for the gaps the baseline exposed, and put a named author and review date on anything that makes a claim.

We do not rewrite everything. We rewrite the pages that sit behind the questions where you are losing.

Stage five

The work that happens off your site

Assistants quote sources they already trust, so part of this job is never on your own domain. We look at the answers where a competitor is named and you are not, and we find what those answers are citing. Directories, trade titles, review platforms, industry roundups, local press.

Then we get you into those same places, accurately. Consistency matters more than volume. If your name, location and description differ across five listings, a model holds five weak facts about you instead of one strong one.

Stage six

The monthly loop

The sampling never stops. Every month you get presence, share of voice and visibility tracked against your baseline, split by platform, because the engines disagree with each other.

On one charity account the same brand sits at 39.7 percent on ChatGPT, 29.9 percent on Google AI Overviews and 23.6 percent on Perplexity. Sixteen points between the best engine and the worst, same brand, same fortnight. An average would hide that. We report it split out and work on the weakest engine deliberately.

You also get citation movement, the questions you newly won and the ones you newly lost, and what we are doing next month with the reason for it.

Honest timelines

We are not going to tell you this is quick.

One to two weeks

Changes to pages that are already indexed start to show up.

Two to four weeks

Genuinely new pages reach the models.

Three to six months

Shifting how a brand is understood, not only whether it is found.

We aim for five to ten percent improvement in visibility a month. Compounded across a year that is a serious change in position. It is not a switch you flip. And if you are starting from nearly nothing, we will say so plainly. One account we track opened at 8.7 percent and slid to 5.3 percent inside a fortnight, before any work began. That is a real starting point, and a better place to begin from than a flattering guess.

What you actually receive

We do not sell reports. We do the work, then show you what the work did.

A baseline report inside the first two weeks, with your real numbers, split by engine.
A prioritised list, ordered by what will move soonest.
The rewrites and the technical fixes delivered, not just recommended.
A monthly report you can put in front of a board without translating it first.

Start with the baseline