From invisible to cited: six months with a US specialty insurance broker
A California-licensed specialty brokerage went from 8,032 monthly search impressions to 311,102, and from absent in every AI answer about its categories to named in seven out of ten, including as a broker to get a quote from. This is what we built, in what order, and what it produced.
First full month of the engagement against the last. Google Search Console unless marked GA4.
About the broker
A young, California-licensed specialty brokerage placing excess and surplus lines for businesses and high-value homeowners, with a modern stack, a fast-quote proposition, and carrier appointments that kept widening through the engagement.
The brief was not "more traffic." It was zero customer acquisition cost. The broker was buying leads through a paid channel that worked but did not scale economically, and wanted an owned channel that produced qualified quote requests without a per-lead price attached. That shaped everything: the categories chosen, the depth of each page, and the decision to keep pivoting when a category proved slower than the one next to it.
The Challenge
Insurance is one of the most expensive categories in paid search, so organic positions displace real money. And insurance buyers increasingly ask an AI assistant before they open a comparison site, so the work had to earn citations, not just rankings.
Where the site started: 60 pages earning any search impression, concentrated in two categories from an earlier sprint (med spa and restaurant). Average position 30.9. Twenty-three clicks in the first full month. A GEO audit on 12 April ran ten questions across ChatGPT, Google AI Mode and AI Overviews, and the broker appeared in none of the seven non-branded answers. Competitors held every slot.
The finding that set the strategy: on "best insurance broker for medical spas", ChatGPT's web search retrieved one of the broker's pages and then declined to cite it. The page was in the consideration set and lost on authority signals, not relevance. The gap was citation density and topical depth, not indexing, and every later decision followed from that.
Our approach
Category research and planning
We analysed roughly 33,000 keywords across six insurance categories before a page was written. Two patterns came out of it. Question-format queries were two to four times easier to rank for than their plain equivalents, and state-specific terms carried CPCs of $30 to $130 at single-digit difficulty. Those became the content template. Keyword tools were only the starting point: what people were discussing on Reddit, and the searches AI assistants ran on their behalf, decided what got built next.
Content planning, writing and publishing
The site went from 60 pages earning impressions to 408 live URLs. Content shipped as clusters rather than one-off posts: a pillar page for each product, sub-pages for cost, coverage, alternatives and requirements, then supporting blog depth underneath. In total, 166 blog posts and roughly 240 product and location pages, written and published by Geology.
Technical and entity work
FAQPage, Article, BreadcrumbList and InsuranceAgency schema across the page set. Internal linking rebuilt at each cluster launch. Index coverage audited weekly by URL inspection, with dead sitemap URLs, uncrawled clusters and orphan pages kept as a standing action list.
Off-site placement
Reddit seeding across insurance and regional subreddits, then a structured programme of threads and comments. Sponsored placements went to the priority clusters rather than spread evenly, because links speed up indexing where content depth already exists.
Weekly measurement and course correction
Every week we read Search Console, analytics, index coverage and AI visibility against the plan, and changed the plan when the data said so: which cluster to build next, which pages to retitle, where internal links and placements should point, and when to stop investing in a category. Both category pivots were decided inside this loop, within a week of the evidence.
Three pivots, and why each one happened
The account did not run in a straight line. It changed category twice, and both changes came from reading the data early rather than defending the original plan. Each was decided within a week of the evidence, not at a quarterly review.
- Feb – Mar 2026
Proof in a narrow category
A small first sprint on med spa, restaurant and startup cyber insurance. Traffic was modest, but the broker started appearing in ChatGPT answers next to MedPro, ProAssurance, CM&F and The Doctors Company, in a category where it had no authority six weeks earlier. That result is what earned a larger mandate.
- 9 April 2026
First pivot: small commercial
New carrier appointments let the broker write commercial risks, so the work moved to hotel, bed and breakfast, and short-term rental insurance. The hotel category went from one page to 22 in six weeks. The head term for short-term rental insurance sat at keyword difficulty 10, where every comparable head term sat at 60 to 90.
- 4 May 2026
Second pivot: California and Florida property
The client redirected to wildfire-zone and high-value home insurance, where California and Florida held roughly 80% of the opportunity. Hotel work was finished and deprioritised rather than abandoned, and link investment moved with the content. The California FAIR Plan cluster shipped seven days later.
- Jun – Aug 2026
Scale the template that worked
The FAIR Plan structure was reused for California and Florida homeowners, then high-value homes, then Texas and Colorado, then New York commercial and apartment habitational. Impressions tripled across those ten weeks. The state template carried to Texas without changes, which is the clearest sign that the structure, not the topic, was doing the work.
The California FAIR Plan cluster
California's insurer of last resort became a mass-market topic when carriers withdrew from wildfire-exposed zip codes. Homeowners who had never heard of the FAIR Plan were suddenly researching it, and the existing content was thin: the plan's own site, a few news write-ups, and almost nothing on what it costs or what to do instead.
Eight pages shipped on 11 May 2026: a pillar explaining the plan, then cost, alternatives, brush score, a comparison against admitted carriers, and supporting depth. Cost got its own page because that is where the volume sat. The comparison page exists because “what do I do instead” is the question a broker can actually answer.
Before deepening the cluster we read the conversation directly: the 80 Reddit threads that ChatGPT and Gemini were citing for homeowner, wildfire and HOA questions, and roughly 200 searches the AI assistants ran on users' behalf when answering them. That research produced pages a keyword export would not have suggested: leaving the FAIR Plan through the state clearinghouse, State Farm's California pullout, wildfire-zone mortgages, and city pages for the metros where the threads clustered. The State Farm pullout post alone has earned 217 clicks at an average position of 5.1.
| Page | Clicks | Impressions | Avg position |
|---|---|---|---|
| /california-fair-plan/cost | 605 | 45,115 | 6.3 |
| /california-fair-plan/alternatives | 141 | 7,502 | 7.0 |
| /california-fair-plan/brush-score | 135 | 4,078 | 7.4 |
| /blog/california-fair-plan-vs-admitted-carrier | 59 | 6,171 | 7.2 |
| /california-fair-plan/clearinghouse | 44 | 2,097 | 5.8 |
| Cluster total (13 California FAIR Plan pages) | 1,102 | 90,283 | – |
Google Search Console, page dimension, 22 Feb to 31 Aug 2026. The cost page became the largest single source of clicks on the site, more than three times the homepage.
Depth, not a single spike
Every figure compares the first full month of the engagement (March 2026) with the last (August 2026), measured in Google Search Console on the verified domain property. The growth compounded month on month and held after the period closed.
Impressions grew 38.7× and clicks 103× from March to August
Growth held after the engagement period: September reached 336,644 impressions and 2,473 clicks at an average position of 8.4.
| Metric | March | August | Change |
|---|---|---|---|
| Search impressions / month | 8,032 | 311,102 | 38.7× |
| Search clicks / month | 23 | 2,359 | 103× |
| Distinct queries ranked | 358 | 10,976 | 30.7× |
| Queries in the top 10 | 37 | 7,068 | 191× |
| Queries in the top 3 | 5 | 3,612 | 722× |
| Average position | 30.9 | 10.1 | 20.8 places up |
| Organic sessions / week (GA4) | 34 | 731 | 21× |
| Quote-form clicks / week (GA4) | 2 | 20 | 10× |
| Non-branded AI answers naming the broker | 0% | 70% | 0 → 30 of 43 |
| Month | Impressions | Clicks | Avg position | CTR | Queries ranked |
|---|---|---|---|---|---|
| February (from 22nd) | 886 | 2 | 14.5 | 0.23% | 57 |
| March | 8,032 | 23 | 30.9 | 0.29% | 358 |
| April | 18,046 | 36 | 13.4 | 0.20% | 671 |
| May | 53,187 | 168 | 18.2 | 0.32% | 2,414 |
| June | 93,871 | 709 | 15.0 | 0.76% | 3,524 |
| July | 167,731 | 1,582 | 12.5 | 0.94% | 5,860 |
| August | 311,102 | 2,359 | 10.1 | 0.76% | 10,976 |
| September (follow-up) | 336,644 | 2,473 | 8.4 | 0.73% | 10,626 |
- Rankings deepened as coverage widened. Queries in the top three went from 5 to 3,612 while the query base grew thirty-fold. Average position improved from 30.9 to 10.1, where adding thousands of new terms usually drags it down.
- The growth is not brand demand. 99.88% of impressions came from queries that never mention the broker. Of 16,166 distinct queries, 16 were branded.
- Attribution is unusually clean. In July and August, pages built during the engagement earned 488,053 impressions and 3,804 clicks. The 60 pages that existed before earned 16,960 and 149.
- The market is the right one. 87.8% of all impressions came from the United States, rising to 89% in September.
| Page | Clicks | Impressions | Avg position |
|---|---|---|---|
| /california-fair-plan/cost | 605 | 45,115 | 6.3 |
| /blog/best-and-worst-florida-homeowners | 508 | 28,561 | 7.2 |
| /blog/state-farm-homeowners-florida | 224 | 13,585 | 5.2 |
| /blog/state-farm-california-pullout | 217 | 15,887 | 5.1 |
| /blog/aaa-homeowners-california | 204 | 23,440 | 9.7 |
| /california-homeowners-insurance/companies | 172 | 14,855 | 6.3 |
From absent to recommended by name
At the 12 April baseline audit, the broker appeared in none of the seven non-branded AI answers about its categories. Within two months ChatGPT was naming it when buyers asked who to get a quote from.
“How much does med spa insurance cost and which brokers provide quotes?”
The broker is listed beside established med spa specialists (CPK Insurance, MedispaCover, Juno Insure) as a firm to request a quote from, linking to its med spa page.
“Best homeowners insurance in California?”
The answer's action plan names the broker's comparison tools as a place to compare quotes, and cites two of its pages, including on pricing a FAIR Plan plus difference-in-conditions policy.
Several pages hold page-one positions and earn almost no clicks, because Google shows an AI Overview above the result and cites the broker inside it. The buyer gets the answer without a click, so Search Console understates the AI channel on those queries. The prompt set grew with the content, so the July figures are a snapshot, not a like-for-like trend.
What the channel is worth
Insurance is among the most expensive keyword sets in Google Ads. Live cost-per-click on the queries the broker actually earns clicks on has a median of $14.54 and a click-weighted average of $23.21. At the conservative rate, buying August's organic traffic would have cost about $34,000, or $411,000 a year at the same pace. Paid clicks stop the day the budget does. These pages keep earning.
| Basis | Rate / click | August (2,359 clicks) | Annualised |
|---|---|---|---|
| Median CPC (conservative) | $14.54 | $34,300 | $411,598 |
| Click-weighted average CPC | $23.21 | $54,752 | $657,028 |
A funnel model at the August run-rate (3,423 earned sessions a month, a measured 4.3% quote-form click rate, and standard US brokerage assumptions downstream) puts the channel at about 8 policies bound and $4,400 in first-year commission a month, or roughly $29,500 in lifetime commission from one month's policies. The broker has not shared bind data, so this is a model, not a record of closed business. It lands within 15% of the paid-search figure once both are counted over the life of the policies.
What this account taught us
Moving topics beat evergreen ones. The best pages covered things that were changing: carriers leaving states, a plan homeowners had never needed before, rate hikes going live. Keyword tools report historical averages, so they show these topics late or not at all.
Indexing is an authority problem, not a sitemap problem. Google is slow to trust a new name on insurance, where bad advice costs people money. New clusters sat discovered but uncrawled until authority reached them through off-site placements and internal links from pages Google already trusted.
How these numbers were produced
- Search data is a direct Search Console API export on the client's verified domain property, 22 Feb to 31 Aug 2026. September figures were pulled from the same property on 13 Oct 2026.
- Query-level totals are lower than daily totals because Search Console withholds low-volume queries.
- GA4 figures stop on 12 Aug 2026, when the site's GA4 tag stopped recording. Search figures are unaffected.
- AI visibility uses a 43-prompt, non-branded set run against ChatGPT, Google AI Mode and AI Overviews on 19 Jul 2026, against a ten-question baseline audit on 12 Apr 2026.
- Cost-per-click is DataForSEO Google Ads data for the United States, pulled 18 Aug 2026.
“We saw our first leads from ChatGPT inside the first month — and by month three we were getting around 20 organic leads a week. The Geology team turned our SEO game completely upside down, in the best way possible.”
This is the playbook behind our insurance SEO program, part of the broader financial services program. For the method in full, read our insurance SEO guide.
Get results like this
See how Geology can build your AI and organic visibility with a free audit and custom strategy tailored to your industry.