A forensic analysis of Admiral Business Insurance across 1,036 LLM responses — and why being cited 271 times by ChatGPT, Claude, Gemini and Perplexity has not translated into a single ranked recommendation outside the brand's home turf.
Across 1,036 tasks (259 queries × 4 providers — Anthropic, OpenAI, Gemini, Perplexity), Admiral is named in just 24. Simply Business is named in 278. Hiscox in 258. AXA in 163.
The Bloxx competition pipeline pulled 17,104 citation URLs from the four leading LLM providers' answers to 259 small-business-insurance queries. admiralbusiness.com is the 6th most-cited domain in the entire dataset — more than Hiscox UK, more than Markel UK, more than Marsh Commercial. And yet the LLMs ranked Admiral on only 2.3% of tasks. The data tells us with unusual clarity what's wrong, where it's wrong, and which pages would convert if rewritten.
| Provider | Tasks | Hits | Vis % |
|---|
| Location | Tasks | Hits | Vis % |
|---|
| Niche | Tasks | Admiral hits | Visibility | Avg rank | Status |
|---|
Of 180 tasks where the LLM crawler pulled an admiralbusiness.com URL into its evidence pack, Admiral was ultimately ranked on just 30. The other 150 are content that's good enough to be cited as a source — but not good enough to be picked as the answer.
This is a content-conversion problem, not a discovery problem. The LLMs find Admiral. They read its pages. They cite the URL in their source list. Then they rank Simply Business, Hiscox, or a hyperlocal broker instead. The mechanism is the page text: it informs without persuading.
When a competitor wins a ranking despite Admiral being in the source list, the LLM is making a specific decision: "I read this page. The information is correct. But the reasoning a buyer needs to pick this provider isn't here." The reasoning that's missing follows a precise grammar — covered in §07.
A counter-intuitive read: when Admiral is cited, it converts to a ranking at 15.7% — higher than Simply Business (13.1%) or Hiscox (12.9%). Admiral's problem is not page quality on the pages that exist, but page coverage. There are not enough Admiral pages addressing the right queries.
| Company | Tasks citing URL | Tasks ranking | Cite ∩ Rank | Cite → Rank |
|---|
| Provider | Query | Niche | Location |
|---|
Trade + price-tell in £/month + South Wales / Cardiff anchor. Click into any row to read the exact reason the LLM gave for choosing Admiral — including the verbatim phrases lifted from admiralbusiness.com.
Click any row to see what the LLM said.
| Query | Niche | Location | Provider | Rank |
|---|
Decorator, courier, consultant, landscaper, plumber — total absence across all 56 tasks each. Manchester and Birmingham — total absence across 144 tasks combined.
Read the matrix left-to-right: rows are niches Admiral sells products for; columns are the 5 UK cities tracked. Each cell is "Admiral hits / Total tasks". Cells in dark navy are zero. The Cardiff–designer cell is the only consistent green strip.
Read top-to-bottom: the only column with multiple non-zero rows is "(unspecified)" — meaning when there is no geographic constraint, Admiral has a fighting chance. The moment a city is named, Admiral loses except in Cardiff.
17,104 citation URLs were harvested from the LLM evidence pack. After excluding Google's Vertex redirects, here is the source landscape that determines who gets ranked in UK SMB insurance.
Admiral sits 6th by raw citation count — above Hiscox UK, above Markel UK, above Direct Line for Business. The citation funnel is working; the conversion funnel after the citation is not. Note the prominence of aggregator and review sites (Trustpilot, Feefo, NerdWallet, MoneyGeek, Yell, Houzz, Yelp) — these are functioning as quasi-publishers and a complete LLM strategy must include managing presence on them.
| # | Domain | Citations | Type | Share |
|---|
28 distinct admiralbusiness.com URLs were cited. Their conversion rates — citation → ranking — range from 50% to 0%. The pages that convert do something specific; the pages that don't are missing the same specific thing.
| admiralbusiness.com path | Times cited | Converted to ranking | Cite→Rank % | Performance |
|---|
High converters (40%+) like /professional-indemnity, /employers-liability, /articles/what-is-public-liability-insurance-guide-costs are doing one of three things well: they name a specific price, they cite an explicit FAQ-style buyer question, or they reproduce a statutory anchor. These pages give the LLM something quotable.
Low converters (under 15%) like /who-we-cover/decorators-insurance, /who-we-cover/design-and-creative-insurance, /who-we-cover/designers-insurance read as generic product descriptions. The LLM cites them as evidence that decorator insurance is a real thing, then recommends a provider whose page has the proof.
Analysing the "reason" text from every top-3 ranking across 1,036 tasks reveals a tight grammar of phrases LLMs lift directly from competitor pages. These are the patterns Admiral needs on its own pages.
Admiral wins when its page says "from £7.08/month" and "South Wales-based team". It loses when those phrases aren't there. — Anthropic, Perplexity, Gemini reason fields, 24 ranking events
Specific £/month figures lifted into reason text
Trustpilot/Feefo scores LLMs quote
Customer/review counts referenced
"Public liability from £6.07/month for tradespeople" — must appear as a sentence, not just on a button.
"4.8/5 on Trustpilot from 222 reviews" — LLMs cite the score and the reviewer count together.
"30,000+ customers" — Admiral already uses this and it works. Extend per trade.
"South Wales-based team" / "Cardiff headquarters" — this is the single most-cited Admiral asset.
"Meets NICEIC £2m PL minimum" / "Compliant with EL Act 1969" — gives the LLM legal authority to quote.
AXA is the top_competitor on 104 of 253 open gaps. AXA UK on a further 45. Together they account for nearly six in ten gap events — but the long tail tells a different story.
| # | Brand | Gaps |
|---|
| Category | Open gaps | Distribution |
|---|
The long tail is more important than the AXA hat. Below AXA the blockers are hyperlocal brokers: Coversure Leeds (9), B&M Insurance Brokers, 538 Underwriting, Focus Insurance Services, Diamond Assurance Group (Birmingham), Cardiff Trade & Contractor Insurance Ltd, Creative Risk Solutions, Birmingham Commercial Risk Solutions, Ravenhall (Leeds). These are companies with a tiny fraction of Admiral's underwriting capacity that win local prompts because their pages say "Birmingham-based" or "Cardiff broker" verbatim.
| # | Company | Tasks appearing | Visibility | Avg rank | Best rank |
|---|
Even when the buyer types "Admiral Business" into ChatGPT verbatim, the LLM does not return Admiral as a recommendation. The brand entity is not legible to large language models.
Ten queries in the dataset contain "Admiral Business" or "Admiral" in the prompt text — questions like "Is Admiral Business any good for small business insurance in the UK?" and "How much does Admiral Business insurance cost for freelancers in the UK?". Each runs across 4 LLM providers, generating 40 branded tasks. The Admiral rank-row count across all 40: zero.
This is a brand-entity problem: the LLMs cannot reliably recognise "Admiral Business" as a distinct insurer when asked about it directly. The fix is an Organization-schema'd entity hub, Wikidata entry, and consistent sameAs linkage from a structured About page to every social and review property.
| Branded query | Provider | Result |
|---|
We use cookies to understand how visitors use our site and improve your experience. Learn more.