What being named means for an insurance broker, line by line

Named for which line? Why one score hides the answer

Three grades apply to any answer: the brokerage is mentioned, it is recommended to the buyer who asked, or it is named first. Our methodology does not weight them; its score counts whether the brokerage was named, and first place is reported beside it. A brokerage needs a fourth dimension, the line. Its result is several results, one for each line it writes, and an overall visibility score averages a strong line with a weak one until neither shows.

So read the panel by line. A broker named on every home question and on none of the commercial ones has a finding the average would hide: whatever the commercial pages say, it is not reaching those answers. Every answer is kept with its question, so the answers can be read line by line, and for a brokerage that is the first thing to read.

Keep a mention rate per line beside the score as well: the share of that line's runs in which the brokerage appears at all. Every recommendation is also a mention, so the mention rate is where a new line page would show first, before there is any sign of whether the page persuades.

Who else can turn up in a broker's answer

The names beside yours are the other half of the reading. For an insurance question they can be other brokerages, insurers that sell direct, comparison sites or directories, and each kind means something different for what you do next. Record every name in order, with a note of what kind of business it is. That note turns a list of names into a picture of your share of voice among brokers in particular.

A comparison site or an insurer in the answer is not a defect to fix. It is a different kind of answer, possibly to a buyer who wants something different. What a brokerage's pages have to state for an answer to say what the broker does instead is the subject of broker or comparison site.

A panel by line of insurance and situation

Write the panel from the lines the brokerage wants more buyers for, weighted toward the ones it places best, and give each line the situation that changes the question. Twelve questions is the size we run for a paying client; six is a fair first pass of your own. How to keep them comparable from month to month is in how to measure AI visibility, and the record format is in setting up prompt monitoring.

Place needs care in a broker's panel. A buyer names a town or says "near me", while your pages name the states, provinces or countries you work in. Put a town into most questions, the way buyers do, and keep at least one line question for each jurisdiction you want buyers from, so a reading can tell a missing place apart from a missing line.

Leave out questions about cover rather than about a broker. "Does home insurance cover a sewer backup" asks how policies work. Its answer may help a buyer, but it is not a buyer choosing a broker, and that choice is what the panel is for.

What to note in a broker's answers, beyond named or not

The standard fields, from named through to the verbatim answer, stay as they are. A brokerage adds five of its own, because an answer about a broker attaches facts that a buyer may act on:

• The line the answer named you for, in its own words.
• What it called you: a broker, an agent, an independent agency, an insurer. If it says you sell one company's policies and you place with several, your pages are not saying so.
• The insurers it said you work with, checked against your current agreements.
• The license, jurisdiction or area it stated, checked against the licensing body's lookup.
• Any premium or price it attached to you. A figure you never published gets traced to its source, and an answer's figure is never repeated to a buyer as a quote.

Grade each fact right, wrong or stale. A wrong one has a source somewhere, an old page or a listing, and the source is what gets corrected; nothing is fixed inside the assistant. How to follow a wrong fact back to where it started is set out in when AI gets facts about your business wrong.

If your book renews in clusters, note those months beside the readings as well. Buyers who are shopping at renewal may ask differently from buyers with a first policy, and a month in which the line questions move is easier to explain when you can see what else was happening in it.

Reading a brokerage's first month

A first month has one job: to be what the second month is compared with. The only naming figures we have published come from a probe inside our audit of 9,483 small business websites (40 trades, 90 metros in the US, Canada, Europe and the UAE, late August to early September 2026). Asked one buyer question per business with web retrieval on, Claude named 14.9% of 9,240 businesses; in the 5,928 answers that took the form of a clearly differentiated list, the share was 17.7%.

Every trade is pooled in those two numbers, taken on one day from one assistant, and none of it is broken out for brokers. It is context for a first reading, not a benchmark for it. The baseline earns its keep at the second reading, when the same questions show which lines moved and which did not, and the patterns below say where to look.

What the reading cannot tell a brokerage

The panel counts answers. Policies bound and premiums written happen elsewhere, and a buyer who heard your name from an assistant and then called appears nowhere in the readings; ask on the first call where they heard of you, and note the answer. The quality of your advice and your claims help is outside it too. And each question is asked fresh, while a real buyer may have described the house, the claim history and the budget before asking for a broker.

Keep readings out of your marketing. An answer varies from run to run, and whether a line such as "recommended by ChatGPT" counts as an endorsement or a claim under the advertising rules where you are licensed is a question for your compliance function or counsel. The reading is for deciding which pages to fix, starting with the facts set out on the insurance brokers overview.