Commercial and specialty insurance
Independent agencies, MGAs and wholesale brokers with 5 to 50 producers. Right now the AI answer for most commercial coverage questions points at aggregators and carriers.
AI SEO and GEO for B2B firms
We show you what AI says about your firm.
The buying problem
Your buyer shortlists before they ever contact you.
A commercial buyer researching a vendor now asks an assistant first. They describe their situation in a sentence, get three or four named firms back, and start their shortlist there. By the time anyone contacts you, the comparison has already happened somewhere you cannot see.
That is a different problem from ranking. A search result gives the buyer ten options and lets them choose. An assistant gives them three and explains why. If you are not one of the three, there is no page two to be on.
The mechanics are not mysterious. An assistant assembles an answer from sources it retrieves, and most of those sources are not your website. They are directories, association listings, review platforms, trade press, forums and comparison pages. Your site tells the model who you are. Everyone else’s content decides whether you get named.
You cannot sell to a buyer who never reached your website.
The four layersFour surfaces,
Four surfaces,
four different jobs.
Most of this industry uses these four words interchangeably. They are four different places, run by different systems, and they respond to different work.
| Layer | The surface | What actually moves it |
|---|---|---|
| AI SEO | Google AI Overviews and AI Mode | Passage-level clarity, entity definition, content Google can extract and attribute |
| GEO | ChatGPT, Perplexity, Gemini, Claude, Copilot | What third-party sources say about you, and whether your entity is unambiguous |
| SEO | Google search results | Technical health, page structure, topical depth, links |
| Local SEO | Google Maps, the local pack, Bing Places, Business Profile, reviews, listings | Profile completeness, categories, review velocity, listing consistency |
All four are in every retainer. We do not sell them separately, because the last two do not work without the first two. Tiers differ by scope: locations covered, prompts tracked, assets produced.
Who we build this for
We sell to three kinds of firm.
Firms doing $3M to $50M in revenue, with 20 to 200 staff, selling a considered purchase to another business. We are US-first, with Australian clients served in USD.
B2B professional services
CPA and tax advisory firms, corporate law boutiques, management advisory practices.
IT services and security
MSPs, MSSPs, cybersecurity consultancies and cloud partners.
How the programme works
What the first 90 days look like.
One retainer, one team, a fixed rhythm.
Baseline
A crawl of your site, your Search Console and Business Profile data, your listing and review footprint, and a prompt set built from the questions your buyers actually ask. We run that set across ChatGPT, Perplexity, Gemini, Copilot and Google AI Mode and record what came back, with dates.
Entity and foundations
Assistants need to be certain who you are before they will recommend you. One consistent description of your firm across your site, your schema, your Business Profile, your listings, your association entries and your team profiles. Where those disagree, we fix the disagreement.
The evidence layer
Getting named in an AI answer usually depends on sources you do not own, so the work moves off your site: association profiles, industry directories, review platforms, trade publications, expert commentary under the names of your actual people, and comparison content a model can retrieve and attribute.
Rerun, compare, report
The prompt set runs again on a fixed schedule. You see the same prompts, the same platforms, a new date, and what changed.
We do not control ChatGPT, Google or Perplexity, and we will not tell you when you will appear in them. What we commit to is the work, the measurement, and the dated record of both.
Evidence
What we can prove today, and what we cannot.
Nobody in this category has a long results record in insurance or professional services, including us. The category is roughly two years old. Rather than fill that gap with numbers, here is the evidence that does exist.
Dated observations, not a state
An AI answer is not a ranking. Ask the same question twice and you can get two answers. So we never tell a client they are invisible in AI search. We tell them what we saw, on which platform, from which prompts, on which date, and we run the same set again on a schedule so the comparison means something.
Every number we put in front of a client carries the platform, the prompt set and the date it was observed.
How we handle evidenceThe free snapshot
Start with the snapshot.
It is a small version of the first two weeks of the engagement. We build a prompt set from your buyers’ questions, run a sample of it across the assistants, crawl your site for the technical and entity signals that feed those answers, and send you the result.
How it is produced
- You tell us the site and the buyers you sell to.
- We build a prompt set from the questions those buyers actually ask.
- We run a sample of it across the assistants and crawl your site for the entity and technical signals that feed those answers.
- You get the prompts, what each platform returned, the date we ran it, and the two or three things we would do first.
What it is not
A sample of prompts is a sample. Different prompts, a different day, or a different account can produce a different answer.
The snapshot tells you what we observed on one date. It does not tell you your permanent position, and any agency that says otherwise is guessing.
Fit
We would rather say no at the snapshot stage.
Than bill you for six months of the wrong thing.
Worth a conversation if
- You sell a considered purchase with a sales cycle measured in weeks, not minutes
- Referrals have carried your growth and you want a second channel
- Someone at your firm can approve and publish website changes
- You have people who are willing to be named and quoted publicly
- You can commit to three months before judging the work
Not worth either of our time if
- You are pre-revenue, or selling something low-ticket and transactional
- You need leads this week
- You want a one-off project rather than a programme
- Nobody at your firm can edit the website
- Leads arrive and nobody follows up
Before we talk
Six honest answers.
Can you guarantee we appear in ChatGPT?
No, and no one can. We do not own or control these systems, their retrieval changes without notice, and the same prompt can return different answers on different days. What we commit to is the work, a measurement instrument you can inspect, and a dated record of what changed.
How is this different from the SEO we already pay for?
Classic SEO targets Google’s blue links. That work still matters, and it is inside our retainer, because assistants retrieve from sources that rank. What is new is the entity work and the off-site evidence layer that decides whether a model names you when it assembles an answer. Most SEO retainers do none of that, and most do not measure any AI surface at all.
Can you show me results from a firm like mine?
Usually yes, with one limit we state up front. In the B2B areas closest to most of our buyers, managed IT and cloud, cybersecurity, and B2B SaaS in part, every engagement was delivered white-label under signed NDAs. We can walk you through the scope, the method and what we changed, but we cannot name the client or arrange a reference call on that work. Where we have not worked in your exact vertical we say so, rather than stretching an adjacent case to fit, because the standards vocabulary and the buying committee change between verticals and a case that does not transfer is worse than no case. Either way you get a snapshot of your own situation, with your prompts and your competitors, before you pay anything.
How long before anything moves?
The technical and entity work shows up in weeks. The off-site evidence layer, which is the part that drives assistant citations, is slower, and we ask for three months before you judge it. Anyone quoting you a date for an AI citation is making it up.
Do we have to write the content?
No, we write it. We do need access to your people for roughly an hour a month, because the material that gets retrieved and cited is specific, and specific means it came from someone who does the work. Generic content written without you will not be retrieved.
What do you need from us to start?
Access to Google Search Console, Google Analytics, your Google Business Profile, and the ability to publish changes to your website. Plus one person who can make decisions and one hour a month from your subject-matter people.
See what the assistants say about you.
We will run a sample of your buyers’ prompts, crawl your site, and send you what we found with the date on it. If the answer is that you are already well positioned, we will tell you that too.