How an insurance agency appears in AI recommendations, and the one study an engine quoted exactly right

An insurance agency is the one vertical in our October measurement where an engine pointed at a real third party dataset, and the dataset held up. Asked how an insurance broker appears in AI recommendations on 21 October 2026, ChatGPT cited a 2026 Conductor study of US insurance searches and reported Google AI Mode at 43.9 per cent of citations, Perplexity at 33.6 per cent and ChatGPT itself at 11.9 per cent. We opened conductor.com/academy/2026-insurance-ai-search-benchmarks/ on 22 October 2026 and found all three figures, in those words. Gemini, asked the same question in the same session, produced a four step purchasing plan with price tags, and our log records that it ran no search for that question and returned zero citations.

The question, the surfaces and the dates

"How does an insurance broker appear in AI recommendations?" went to the ChatGPT product surface at 22:22 UTC and to Gemini at 22:39 UTC on 21 October 2026, through a data provider, with no agency named. ChatGPT's reading records live search true and 73 citations, led by developers.google.com with 12, help.openai.com with three, developers.google.cn with three, conductor.com with two, BrightLocal's help centre with two, capterra.com with two and perplexity.ai with two, with the Wall Street Journal, Reuters, Business Insider, the Washington Post and arxiv.org each appearing once. Gemini's reading records live search false and 0 citations.

Two confident answers, one built from 73 pages and one built from none, and no visible difference on screen. That is the practical reason an agency principal should treat any price in an AI answer as a lead rather than a figure.

What the searching engine said, and what the study behind it says

• The measurement ChatGPT reported, verified word for word. The Conductor page, read 22 October 2026, states "AI Mode drove 43.9% of the Insurance citations we analyzed, with Perplexity close behind at 33.6% and ChatGPT further back", and separately "far ahead of ChatGPT's 11.9%". It also states that "An AI Overview appeared on 40.7% of Insurance searches" and that no insurer owns the citation share, naming Insureon at 2.9 per cent and Progressive at 2.8 per cent. ChatGPT's own framing of this was careful and correct, that these are shares of citations in that study and not market shares or probabilities of being recommended.
• Why that matters more than any tool in either answer. If AI Mode and Perplexity together account for 77.5 per cent of the insurance citations in that dataset, then an agency measuring only ChatGPT is watching the smallest of the three. The study is Conductor's, not ours, and we are quoting it from the page we read rather than reproducing its method.
• ChatGPT then ordered the work and put one paid product fourth. Google Business Profile free, Search Console free, a ChatGPT test of real buyer questions such as best independent commercial insurance broker for contractors in Maryland, then BrightLocal. After that, broker specific authority pages per target industry and coverage type, independent credibility through reviews and association profiles, and a monthly test of 20 to 50 realistic prompts across ChatGPT, Google AI Mode and Perplexity.
• Its closing instruction was explicit. Start with the profile, Search Console and your website before paying an agency or buying AI visibility software, add BrightLocal if local listing management becomes a real workload, and consider specialist support only after measuring a baseline.

Every price in the two answers, opened on 22 October 2026

• BrightLocal, which ChatGPT gave as from 39 dollars a month advertised starting price. brightlocal.com/pricing publishes, for one location, Track at 41 dollars a month billed monthly or 369 billed annually, which the page itself shows as 31 a month, then Manage at 54 monthly or 485 annually and Grow at 65 monthly or 584 annually. The Track feature list opens with a Local AI Visibility Tracker, flagged new. So the engine's 39 is close to the 41 on the page without matching it, and the page's own cheapest reading is 31. The plan cards also print "Price on request" immediately beside the published figure, so brightlocal.com/pricing shows both a price and a request for a price in the same block. We report both readings.
• Schema App, which Gemini priced at around 300 dollars a month and up. schemaapp.com/pricing/ returned 200 and renders 2,159 words with no plan price in them. The only dollar figure is an overage line, 100 dollars per 100,000 external chatbot requests beyond an included 100,000 a month.
• The Merkle schema generator, which Gemini correctly called free. It is, at technicalseo.com/tools/schema-generator, and InsuranceAgency and FinancialProduct are real Schema.org types, so this step is genuinely zero cost.
• Adobe Brand Visibility, which Gemini put at enterprise pricing via demo. That description is consistent with how Adobe sells, and the engine gave no figure, which was the right call.
• YCP and First Page Sage, both given as custom quote. firstpagesage.com returned 200 and renders 766 words with no dollar figure, and the two pricing URLs we tried on that domain returned HTTP 404. So custom quote is accurate.
• Nationwide's agency locator, which Gemini listed as a source data platform with pricing "varies". This one is a category error worth flagging rather than pricing. A carrier's agent locator is a carrier distribution tool for its own appointed agents, not a listings service an independent brokerage buys. An independent agency cannot subscribe to it.
• Profound, named as an AI search auditing platform in a neighbouring reading in this audit. tryprofound.com/pricing returned 200 on 22 October 2026 and renders 1,302 words with no dollar figure anywhere in them. Any specific dollar price circulating for Profound did not come from that page.

The cheapest path for an independent agency, in order

• Zero. Google Business Profile complete, with real service areas and the correct business category, and Search Console verified. The searching engine put both ahead of every purchase.
• Zero, and this is where the agency actually wins. A page for each target industry and coverage type, in the buyer's words. Cyber insurance for small professional firms. Workers compensation for residential construction. Commercial auto for a local fleet. Each page naming the eligibility, the quote process, the licensed states, the carriers you can place with where you are allowed to say so, and the named producer with a licence number a stranger can check. ChatGPT asked for licensed expertise, actual service areas and original case studies by name.
• Zero. InsuranceAgency and FinancialProduct schema from the free Merkle generator.
• Zero. The monthly test the engine described, 20 to 50 realistic prompts across ChatGPT, Google AI Mode and Perplexity, recording recommendation frequency, citations and competitor mentions. Given the Conductor shares, AI Mode and Perplexity belong in that test and not just ChatGPT.
• 31 to 41 dollars a month. BrightLocal Track for one location, 369 billed annually or 41 billed monthly, with a Local AI Visibility Tracker in the plan. The only verified subscription figure in either answer.
• Ask. Schema App, Adobe Brand Visibility, YCP, First Page Sage and Profound. Five paid names, no published figure between them on the pages we read.

The compliance point neither engine made

Both answers treated an insurance brokerage like any other local business, and in one important way it is not. Everything the engines asked you to publish, the coverage pages, the case studies, the carrier names, the testimonials, passes through a review process that varies by state and by carrier appointment. Gemini at least named the category, calling insurance a Your Money or Your Life subject, which is a search industry term rather than a regulatory one. The practical consequence is that the page an engine can quote is the page your compliance reviewer has already approved, so the bottleneck on this work is the review queue and not the writing.

That makes the measurement half cheaper than the publishing half, which is unusual. Running the prompt test costs nothing and tells you which coverage lines and which states you are invisible in, which is exactly the list worth taking into a compliance review rather than a general request to approve more content.

Where AI Knows Us sits, and where a rival beats us

We sell AI visibility measurement and nothing an insurance agency would call marketing. We do not buy leads, run ad accounts, write compliant copy, manage carrier relationships or request reviews, and nothing above is a pitch for any of that. What we do is ask the buyer's question with your agency's name absent from the question, on ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, and report which agencies were named and which URLs produced each answer. Our ladder, read from aiknowsus.com/llms.txt on 22 October 2026, is SEO Start at 149 dollars a month in the US and UK or 99 elsewhere, which watches 0 buyer questions on 0 engines, Growth at 299 in the US and UK or 199 elsewhere for 100 questions on all five surfaces weekly, and Scale at 599 and 399 for 300 questions. Growth is the cheapest plan of ours that reads all five surfaces, so 149 is not an AI visibility entry price.

Where a rival beats us, by name and with the figure. A single office agency should buy BrightLocal before it buys us. Its one location Track plan is 41 dollars a month billed monthly or 369 billed annually, shown on the page as 31 a month, with a Local AI Visibility Tracker inside the plan, which is roughly a ninth of our cheapest engine watching tier for an agency whose market is one county. The engine that actually searched put BrightLocal in its answer and not us, and on price for that buyer it was right. SE Ranking beats our arithmetic a step up, publishing Core at 129 dollars a month billed monthly with 100 prompts tracked daily on seranking.com/subscription.html, read 22 October 2026, against our 299 for 100 questions watched weekly. We did not verify its engine list, so read that line yourself. Profound publishes single sign on, SAML, SOC 2 compliance and up to nine answer engines on its enterprise tier, and a word boundary search of our own pricing page, 119,203 bytes read on 22 October 2026, returns zero hits for sso, saml, soc 2, iso, dpa and white label. For a brokerage with a vendor security questionnaire, those lines will be blank for us. Our pricing page also contradicts itself, rendering "Businesses covered 1 1 1 1 1 1 10 1" above "More than one business No" in all eight columns, which matters to an agency group with several named entities. Both readings render and we publish both.

What this page establishes. One question per engine on one night with the search flag recorded for each, one fetch per URL on 22 October 2026, and the Conductor figures quoted from the study's own page rather than from the engine that reported them. Every price is attached to the page it was read on.

What agency principals ask about AI recommendations

Which engine should we measure first

On the Conductor data ChatGPT cited and we verified, Google AI Mode drove 43.9 per cent of insurance citations and Perplexity 33.6 per cent, with ChatGPT at 11.9 per cent. An agency testing only ChatGPT is watching the smallest of the three in that dataset.

Do we need software to start

The searching engine said to start with the profile, Search Console and the website before paying anybody, and to add a tool when listing management becomes a real workload. The 20 to 50 prompt monthly test it described costs nothing.

What should a coverage page contain

The buyer's industry and the coverage line in the heading, the eligibility rules, the licensed states, the quote process and timeline, the named producer with a checkable licence number, and an original case study if compliance will clear one. An answer about a commercial broker for contractors in a named state can only be assembled from a page that says contractors and names the state.

Why did an answer quote prices for products that publish none

Because that answer had not looked anything up. Our log records zero citations and no live search for it, and five of the paid names in it publish no figure on the pages we read. Open the vendor page and note the date before you budget.

Can we use a carrier's agent locator for this

Not as a listings service. One answer listed a carrier agency locator as a data platform to leverage, and that is a carrier distribution tool for its own appointed agents rather than something an independent brokerage subscribes to. The listings that matter here are the ones you control, starting with Google Business Profile.