AI Visibility

Your buyers now ask an AI before they ask you. This is whether you're in the answer — and what it takes to get there.

1

The problem

• The answer is a shortlist. "Who are the best X consultancies in the UK?" returns three or four names. That is the new first cut, and it happens before anyone fills in a contact form.
• You don't own the sources. The engine cites whoever it finds authoritative — directories, aggregators, review sites, a competitor's blog. Your own site is one candidate among many, and often not the one it picks.
• Most of it is zero-click. The buyer reads the answer and never visits anybody. Nothing lands in GA4. The loss is invisible by default.
• It is machine-readable or it doesn't exist. Being cited depends on whether the crawlers can fetch your pages and parse them — a structural problem, not a brand problem.
• Nobody in the mid-market has claimed this ground. Which is the opportunity: it is cheap to do, compounding, and almost uncontested.

2

The evidence — from a live client programme

GIG Gulf, a UAE insurer (rebranded AXA Gulf). Measured across ChatGPT, Claude, Google, Perplexity and Gemini between June and July 2026. Everything below is our own instrumentation, not a vendor's dashboard.

Proof point 1 — AI traffic is already real and it converts attention

• This is the visible tip only. It counts people who clicked through from an AI answer — the majority who read the answer and moved on leave no trace at all.
• An engagement rate near 90% says the traffic is qualified: people arriving from an AI answer already know why they're there.

Proof point 2 — you can measure share of answer, precisely

• We built the measurement rather than buying it: 20 real buyer prompts × 3 engines, scored for citation, then extended to five engines and repeat sampling.
• Result: cited in 83% of AI answers (50 of 60). By engine — Claude 100%, Google 80%, ChatGPT 70%. By phase — Competitive 100%, Trust 100%, Category 92%, Attribute 83%, Utility 33%.
• The loss was concentrated, not broad: three prompts out of twenty, all in the biggest line of business, all "how do I…" utility questions — and absent on two engines each, so a real content gap rather than model noise.

• The pattern, not the sector: the brand won every question about itself — who it is, whether it's any good, how it compares. It lost every question about how to do the thing that precedes a purchase.
• All three losses were practical "how do I…" questions asked by someone already in-market. A rival and a directory took those answers instead. The brand had no page worth citing.
• That is the general rule. Brand and comparison queries are the ones a company already invests in. The pre-purchase practical question is the one nobody writes for — and it's the one the buyer asks first.

Left two columns: measured. Right column: the equivalent questions in a B2B buying process — not yet measured for any client, because nobody is tracking them. That is the opening.

Proof point 3 — the competitive picture is brutal, and quantified

Live pull from the Ahrefs AI-citation index — how often each domain is cited by each engine.

• Aggregators out-cite the brand 10–20× on the browsing engines. Not a marginal gap — a different order of magnitude, on the surfaces buyers actually use.
• The lesson transfers directly: whoever publishes the structured, authoritative, comparison-grade content gets cited. Brand strength alone does not buy you in.

Proof point 4 — we caught our own measurement error and published the correction

Proof point 5 — off-the-shelf tools get the entity wrong

• The client is a rebranded business — a quarter of its demand still rides on the legacy name. Any tool string-matching only the new brand under-reports visibility and invents a category gap the legacy equity already fills.
• We resolve legacy → current in the measurement. No off-the-shelf tool does. For any client who has rebranded, merged or been acquired, that single correction changes the entire picture.

3

What the work actually is

• Baseline. Build the real buyer prompt set — the questions their customers ask, not keywords. Run them across the engines, with repeat sampling. Score citation, by engine, by phase of the buying journey, by line of business.
• Diagnose. Separate the three causes: content that doesn't exist, content that exists but isn't machine-readable, and content that exists and is simply beaten by a better-structured source.
• Fix the sharpest gap first. On this programme: a utility hub answering the exact losing questions, properly structured with FAQ and HowTo schema, factual, current, with a soft route into the commercial page.
• Protect what's working. Legacy-brand pages stay live and crawlable; comparison and claims pages get reinforced where a rival co-appears.
• Confirm the machines can read it. Check the AI crawlers can actually fetch the new pages — a citation problem is usually a crawlability problem wearing a marketing costume.
• Report a trend, not a snapshot. Weekly re-run, share-of-answer over time, and the named prompts moving from lost to won.

The targets we set — and how they're judged

• The three losing prompts move from 1-of-3 to 3-of-3 within four weeks of the hub going live.
• The weakest line of business lifts from 73% → 85%+; the weakest journey phase from 33% → 70%+.
• Overall share of answer holds at 83%+, with rival co-citation on category queries reduced.

4

Why it transfers to a B2B business

• The mechanic is identical, the stakes are higher. A B2B buying committee researching a supplier is doing exactly what a consumer does — asking for a shortlist. The difference is that one B2B answer can be worth a six-figure contract.
• The prompts that matter are obvious once you look: "who are the leading suppliers of X in the UK", "best alternatives to [incumbent]", "is [company] any good", "what should I ask a [category] supplier". None of those are keywords anybody is currently tracking.
• It suits a business that doesn't run paid media. No ad spend, no performance budget, no B2C mechanics. It is editorial, structural and technical work — precisely what an embedded specialist does well.
• It compounds and it's cheap. Structured content keeps earning citations. And in the mid-market almost nobody has started.
• Be straight about the evidence: the proof above comes from a regulated consumer-facing brand at national scale. The method, the instrumentation and the failure modes carry over intact; the B2B citation benchmarks have to be built client by client, because nobody has them yet. First mover advantage cuts both ways.

5

How we sell it

• As a diagnostic inside the Strategy & Planning Model: an AI-visibility baseline is one of the audit lenses. Cheap to run, and it produces a slide no competitor has shown the client — their own name missing from an answer their buyer just read.
• Then as an ABMS module: the content and structure work, delivered by our specialists, reported as share of answer alongside pipeline contribution.
• The opener that lands: run three of their real buyer questions live, in the room, on a laptop. Either they're in the answer or they aren't. That takes ninety seconds and it sells itself.