Published October 2026.
Direct Answer: In September 2026, AI answer engines cited Risk Coverage Hub 7,683 times across 34 articles — roughly sixty citations for every one web click. The most-cited page, our guide to reinsurance treaty structures, was cited 960 times in a single month. This report publishes the receipts: which pages the machines reached for, which questions drove them, and what it means when a website is read far more than it is visited.
The headline numbers
Every number below comes from first-party data: Bing Webmaster Tools (web search and AI citation reports), Google Search Console, and Google Analytics, all for September 2026.
- 7,683 AI citations. That is the count of times Microsoft’s AI answer engines grounded an answer in a Risk Coverage Hub page during September 2026. Thirty-four pages earned at least one citation.
- ~124 web clicks. Bing web search delivered 7,174 impressions at an average position of about 5, with a 1.7% click-through rate. Ninety-six percent of those impressions came from desktop.
- 685 active users, 773 sessions. Google Analytics for the month, with a 22.64% engagement rate — respectable for a reference site, and beside the point, as the next section explains.
- 4 clicks from Google in three months. Google Search Console shows 2,650 impressions and essentially zero traffic; Google sent almost no traffic — 2,650 impressions at an average position near 50 over three months — while Bing became the citation channel.
The ratio that matters: about 60 AI citations for every 1 web click.
The citation leaders
Citations concentrate. The top ten pages account for the large majority of all 7,683 citations, and four of them are the site’s deep reference guides:
| Page | AI citations (Sep 2026) |
|---|---|
| Reinsurance Treaty Structures: Quota Share, Excess of Loss, Aggregate | 960 |
| Insurance Underwriting Cycles: Hard and Soft Markets | 895 |
| Reinsurance Treaty: Complete Guide 2026 | 724 |
| Catastrophe Modeling: Complete Guide 2026 | 664 |
| Umbrella and Excess Liability: Primary Limits | 480 |
| Climate Risk Pricing: Catastrophe Model Updates 2026 | 361 |
| Business Owner’s Policy: Coverage Structure and Eligibility | 353 |
| Cyber Insurance Market Evolution: AI Threats and Deepfakes 2026 | 320 |
| Personal Liability Coverage: HO-3 Homeowner’s Guide | 317 |
| Claims Management: Complete Professional Guide | 292 |
The grounding queries tell the same story from the demand side. “Quota share reinsurance” drove 232 citations on its own; research-intent queries around digital risk, cyber theft, and deepfake liability combined for roughly 266; “reinsurance treaty” added 112; “what is treaty reinsurance” added 84; and “cat modeling insurance” added 81. The machines are not browsing. They are asking definitional questions and reaching for the most structured answer available.
The 60:1 ratio — and what it means
For twenty years, the funnel was simple: rank, get the click, monetize the visit. The 60:1 ratio inverts it. When Copilot answers a user’s question with your sentence, you won the query — and the user never had to click anything.
This changes what a page is for. A page that earns 960 citations and 13 clicks is not a traffic failure; it is a distribution success in a channel that has no click-through rate. The citation is the impression. The old metrics — sessions, bounce rate, pages per visit — measure the waiting room. The citation count measures the consultation.
It also changes who the reader is. Ninety-six percent of Bing impressions arrive on desktop, and the grounding queries skew informational and research-intent: professionals and models doing work, not consumers browsing. The audience is small, expert, and machine-mediated — which is exactly the audience that moves commercial insurance decisions.
Why citations are the metric that matters
Risk Coverage Hub is written to be retrieved, not browsed. Every article opens with a Direct Answer, carries FAQ and defined-term schema, and defines its vocabulary in quotable sentences. That structure is not decoration; it is the reason a machine can lift a paragraph intact and stand behind it.
Call it a book for bots: a reference work whose primary readers are answer engines, and whose human readers arrive through the answers. The 7,683 citations are the proof that the format works — that being the most structured, most definitive answer in a niche earns machine readership at a scale human readership cannot match.
There is an incumbency effect here worth naming. Once an answer engine learns that a particular source reliably resolves “quota share reinsurance,” it keeps reaching for that source. Being early and definitive in the AI citation graph compounds the way domain authority compounded in the link graph — except faster, because the graph is rebuilt with every model update rather than every crawl.
The 30-day experiment now running
On October 1, 2026, twelve new depth-cluster pages were published beneath the two biggest citation winners: six treaty-mechanics pages under the 960-citation reinsurance treaty structures guide, and six hard-and-soft-market pages under the 895-citation underwriting cycles guide. Six more catastrophe-modeling pages are in staging beneath the 664-citation cat-modeling guide.
Together the eighteen pages form a clean crawl-economics experiment. The question: do new, deeply interlinked pages published beneath proven citation winners earn AI citations faster than standalone pages? The baseline is zero citations at birth for all eighteen. The variables are controlled — same site, same schema treatment, same internal-linking discipline, same parent authority.
The November edition of this report will publish the 30-day citation velocity results: first-citation latency per page, citations per page at day 30, and whether proximity to a citation-winning parent predicts machine pickup. If the experiment confirms the hypothesis, the playbook is repeatable: find the citation gradient, mint depth beneath it, measure.
What this proves — and what it does not
This report is the receipts page for a specific claim: that structured, schema-marked, definition-first content earns AI citations at scale. The methodology behind these numbers — Direct Answer leads, FAQ parity, defined terms, citation-grade internal linking — is the same methodology sold in AI search visibility packages. The numbers above are what the method produces when it is applied to a full site for a full month.
Honesty requires the limitations, stated plainly:
- Bing-only data. These are Microsoft’s numbers — Copilot and Bing AI grounding. Other answer engines do not publish equivalent reports, so this is a partial view of total machine readership.
- Citations do not pay directly. Nobody writes a check per citation. The value converts through services, lead flow, and enterprise value — the honest economics are indirect.
- Google is a separate story. Four clicks in three months is a distribution channel lost, not a rounding error. The citation thesis holds on Bing; Google recovery is its own project.
With those caveats on the record: 7,683 citations in thirty days, from a standing start, on a site most humans have never visited. That is what being the definitive answer looks like from the machine’s side of the glass.
FAQ
What is an AI citation?
An AI citation is recorded when an AI answer engine — such as Microsoft Copilot — grounds its answer to a user’s question in your page’s content. The engine read your page, used it to compose or verify its response, and logged your page as a source. It is the machine-readable equivalent of being quoted.
How are these citation numbers measured?
Through Bing Webmaster Tools’ AI Page Stats report, which reports grounding citations per page per month. The September 2026 figures in this report come directly from that report’s October 1, 2026 export, cross-checked against the AI Search Queries report for the grounding queries behind them.
Why do citations outnumber clicks sixty to one?
Because the answer is delivered inside the chat. When Copilot answers “what is quota share reinsurance” using your definition, the user’s question is resolved without a click. The citation replaced the visit — which is why optimizing for clicks alone misses most of the value a reference page creates.
Do AI citations drive revenue?
Not directly — nobody pays per citation. The value converts indirectly: through services sold on the strength of demonstrated visibility, through lead flow from buyers who arrive pre-educated, and through the enterprise value of owning a niche’s citation graph. The honest economics are indirect, and any report that implies otherwise is selling something.
What is the 30-day citation experiment?
Eighteen new depth-cluster pages — twelve published October 1, 2026 beneath the site’s two most-cited pages, six more in staging beneath the third — tracked from zero citations at birth. The November Mint Report will publish first-citation latency and 30-day citation velocity per page, testing whether proximity to a citation-winning parent accelerates machine pickup.
When is the next Mint Report?
Monthly. The November 2026 edition will add October’s citation counts, the 30-day experiment results, and a month-over-month leaderboard — including whether any of the eighteen new pages cracked the top ten.