Skip to content

AI search visibility

Being the source the answer comes from

Answer engine optimisation · AEO · GEO

A growing share of the people evaluating you will never see your website. They will ask an assistant, read the answer, and form a view from whatever it cited. Being the source of that answer is a different job from ranking first, and it is increasingly the one that decides whether you make the shortlist at all.

01

Why nobody clicks any more, and what that costs you

Search has been quietly turning into an answering machine. The traffic that arrives is more qualified and there is much less of it, which breaks a lot of reporting and most content strategies.

The answer replaced the click
Someone asks the question, gets a synthesised answer, and never visits a site. If you were not in the answer, you were not in the consideration set.
Your content is unreadable to a model
Long, unstructured pages that bury the claim, hedge the specifics, and give an assistant nothing clean to quote. Search engines forgave that. Answer engines do not.
You cannot see it happening
None of this shows up in Search Console. Most teams have no idea whether they are being cited, misdescribed, or confused with a competitor.
Being wrong is worse than being absent
Assistants will describe your product either way. Without clear, current, structured sources to draw on, they will confidently draw on something out of date.
02

Who this is for

Companies in categories where buyers research before they ever speak to anyone.

  • Considered-purchase B2B

    Long evaluations that start with a question to an assistant rather than a search box.

  • Technical products

    Where the explanation is complicated enough that people ask for it to be summarised, and the summary needs to be yours.

  • Regulated categories

    Where being described inaccurately by an assistant is a compliance exposure, not just a marketing one.

  • Teams watching clicks fall

    Impressions holding up, clicks declining. That is usually the answer layer, not a ranking loss.

  • Companies with strong content

    You have already done the work; it is just structured in a way that is hard to cite.

  • New entrants

    No brand recognition yet, so what an assistant says about your category is doing your positioning for you.

03

What you get

Each workstream is designed to move a specific metric, not to pad a report.

01

Citation audit

What the major assistants currently say about you, your category and your competitors — where you are cited, where you are absent, and where you are described wrongly. This is the baseline, and it is usually uncomfortable reading.

02

Content architecture

Restructuring what you already have so it can be quoted: claims stated plainly, questions answered directly, specifics rather than hedges, and a structure a model can lift from cleanly without losing the meaning.

03

Machine-readable foundations

Structured data, llms.txt, clean semantic markup, and the technical work that makes a site legible to something that is not a browser. Overlaps with technical SEO and is not the same job.

04

Monitoring

A repeatable way to check what assistants say about you over time, so this is a thing you manage rather than a project you did once.

04

How it runs

Four phases with real durations. Nothing here is a proprietary framework — it is the order the work has to happen in.

01

Baseline

Week 1–2

Establish what assistants currently say about you and your category, which sources they draw on, and where the gaps and inaccuracies are. Without this you cannot tell whether anything you do afterwards worked.

02

Foundations

Week 2–4

The technical layer: structured data, llms.txt, semantic markup, and fixing anything that makes the site hard to parse. Fastest to implement and the prerequisite for everything else.

03

Restructure

Month 2–3

Rework the content that matters most so it answers questions directly and can be quoted without distortion. Usually a rewrite of a small number of high-value pages rather than a large volume of new ones.

04

Monitor

Ongoing

Re-run the citation checks, watch what changes, and adjust. This field is moving quickly and anyone claiming a settled playbook is guessing.

05

What this has produced

I currently run a monthly AI visibility programme for a regulated financial firm. I cannot publish their citation data, but I can describe the method, because the method is the part you are buying.

The programme monitors a fixed set of priority queries across six answer engines every month. Each query is paired with an explicit description of what a correct answer looks like — the entity named accurately, the regulatory status attributed to the right legal entity, the product category not collapsed into a wrong one — and drift from that is the tracked signal. Where an engine gets it wrong, the fix is traced back to the source it drew on. In a regulated category a confident, fluent, wrong AI summary is a compliance exposure rather than a marketing annoyance, which is why the work is structured as monitoring rather than a one-off audit.

06

Questions people actually ask

Including the ones about money.

The other four

Most engagements use two or three together. A strategy sets the direction; the rest do the work.