AI Search Visibility
Audit UK.

For twenty years, the question was “are we ranking on Google?”. The answer determined budget, agency choice, and quarterly board reporting. Two things broke that.

The first is that buyers no longer click ten blue links. They ask a question and read a synthesised answer. The synthesis happens inside an AI engine. The engine cites sources or it doesn’t. If your business isn’t in the cited set, you don’t exist for that conversation, regardless of where you rank on Google.

The second is that there is no longer one engine. ChatGPT, Perplexity, Claude, Gemini, and Brave each draw from a different index or stack of indexes. Ranking on one tells you almost nothing about whether you surface on the others. We covered the structural reason for that in The Multi-Index Problem.

An AI search visibility audit measures both: what each engine knows about you, and what structural signals you’re missing that would change that.

A credible audit covers five dimensions. The order matters.

1. AI crawler access. Are GPTBot, ClaudeBot, PerplexityBot, GoogleOther, OAI-SearchBot, Bingbot, Bravebot, and CCBot allowed in your robots.txt? A blocked crawler is an automatic Unlikely verdict for every AI citation query, regardless of content quality.
2. Multi-engine indexation. Are you verified in Google Search Console and Bing Webmaster Tools? Does your robots.txt allow Bravebot? Do you implement IndexNow (which notifies Bing, Yandex, Seznam, and Naver, not Brave)? Bing’s index powers ChatGPT Search. Brave’s independent index feeds Kagi, Mojeek, and several privacy-first AI tools. Google alone leaves roughly 60% of the substrate dark.
3. AI Share of Voice. A live test of 6 to 12 sector-relevant queries against the public AI answer layer. For each query, score against five citation factors: direct-answer format, named authority, fact density, schema markup, external citation signals. Output: a verdict per query (Likely Cited, Possible, Unlikely) and a count out of 6.
4. Citation graph. The third-party hosts AI engines pull from when answering category queries in your sector. Press, directories, comparison sites, Reddit, trade publications. You don’t need to be on every host. You need to be on the high-frequency ones.
5. Vendor Bake-Off Readiness. Can an AI compare you head-to-head against a competitor when asked? Pricing must be extractable as HTML text. Features must be in semantic lists or tables. Comparison claims must be supported by specific numbers.

A UK audit in May 2026 tests against three canonical engine families. The choice is not arbitrary; it is the smallest set that covers the three distinct indexes underneath the major AI assistants. We covered the structural reason for that in The Multi-Index Problem.

Perplexity sits beside these three as a hybrid: it runs its own crawler, with third-party search providers layered on top, and tends to weight academic and trade-press sources more heavily than the others. We treat it as a fourth test in higher-tier audits, but the three-engine canonical grid is what holds the floor. Three indexes, three verdicts, one structural answer for whether your business shows up in any of them.

A credible audit is a diagnostic, not a dashboard. The deliverable should answer four questions in plain language:

• Where do we sit today, on each engine, against named queries that matter to our business?
• What single change would move the most queries from Unlikely to Possible or Cited?
• What is this costing us in pounds, modelled against a basket of commercial-intent keywords?
• What is the 90-day order of operations to close the gap?

The format we use at F&T is a five-dimension scored dashboard with a single signature finding at the top, followed by a Top 5 Actions list ordered by Critical / High / Medium and time-to-implement. The point is that a board can navigate it in a meeting and a junior marketer can implement it on Monday morning.

The market has settled into three clear price points:

• £250 entry-level AI search report: AI dimension only, 18-cell SoV test (6 queries across Google AI Overviews, Bing-powered ChatGPT, and Brave), one recommended action. Fast, cheap, useful as a sanity check before a bigger commitment.
• £500 to £1,500 full website intelligence report: five dimensions, named competitors, £-quantified Cost of Inaction, 90-day plan. Where most UK SMEs land.
• £2,500/year annual intelligence plan: quarterly re-scans, drift monitoring, one new audit topic per quarter. For businesses where AI search visibility is now a board-level conversation.

Anything below £250 is usually a free tool with a sales pitch attached. Anything above £15k is usually a multi-entity engagement (parent group plus subsidiaries) or a custom research report.

You need an audit if any of the following are true:

• You sell into a category where buyers research before they call (B2B services, considered consumer purchases, hospitality, weddings, care, professional services, premium retail).
• You’ve noticed organic traffic flatlining or declining in 2025 to 2026 despite no obvious cause.
• Your last SEO audit was before 2024 and didn’t mention crawler access, schema markup, or AI citation factors.
• Your competitors are surfacing on AI answers for queries you used to rank for, and you can’t explain why.

You probably don’t need one yet if you’re pre-revenue, your category is fully offline (most of what surfaces in AI is desk research first), or you already have an internal AI visibility scorecard you trust.

From the F&T audit base, the same five findings appear in nine out of ten reports:

1. Robots.txt blocks at least one major AI crawler, usually inadvertently, often via a Cloudflare WAF default.
2. Pricing is hidden behind “contact us”, which makes Vendor Bake-Off impossible.
3. Schema is partial or missing. Organization, Service, FAQPage, BreadcrumbList rarely all present. Article schema on the blog often missing entirely.
4. The site is verified only in Google Search Console. Bing Webmaster Tools unverified, and Bravebot status not declared in robots.txt (Brave has no webmaster tools panel; the only on-site lever is the robots.txt allow). Bing powers ChatGPT Search. Brave is the independent index. We covered this in The Multi-Index Problem.
5. The brand exists in the third-party citation graph at <30% of competitor frequency, usually because no one has been pitching trade press or directories with sector-relevant data.

The five fixes for these are mostly cheap and mostly fast. The compounding effect is what matters: each one moves multiple queries from Unlikely to Possible. We modelled the maths in The Cost of Not Knowing.