Your next client asked ChatGPT for a broker. Who did it name?
Referrals still matter. So does the search that happens before the referral is trusted: “mortgage broker in [town]”, “IFA near me”, “is this firm on the FCA Register?”. Google’s local pack and the AI answer are now the same shortlist. Template sites with a disclaimer in the footer do not make that shortlist.
Cardiff studio, working with small UK brokerages. Standard from £5,950 · Premium scoped on a call where the advice journey needs calculators, CRM and named-adviser workflows.
Standard or Premium. You keep the site.
Up to 15 pages, FinancialService + Person schema, GBP setup, FCA number on advice surfaces, 14–21 days. 50% deposit, 50% on launch.
Calculators, multi-step enquiries, CRM wiring, E-E-A-T on named advisers, 90 days of citation monitoring. The package for firms that live on inbound.
Proof of the same shape of work: CeMAP Practice · securedloanrates.co.uk. Both checkable.
What's in the build?
Six specific things. Not 'a modern financial-services website' — the mechanics that decide whether an engine will name a regulated firm.
FinancialService + Person schema
Not a generic Organization dump. FinancialService on the firm, Person on each named adviser, FCA reference where it belongs. AI engines look for that entity graph before they will cite a regulated firm on an intent query — and almost none of the template sites have it.
Consumer Duty-aware service pages
Mortgages, protection, pensions — each as its own page, in plain English, with process and risk notes rather than invented rates. Information-only framing where it is not advice. The copy a compliance officer can live with, structured so Google and ChatGPT can lift it.
Local pack + adviser directories
Google Business Profile, Bing Places, Apple Business, and NAP that matches the FCA Register listing. City and service-area pages that are actually different, not a find-and-replace of the town name. This is still how “mortgage broker in [town]” is won.
AI-crawler permissions + llms.txt
Explicit allow for the crawlers that feed ChatGPT, Claude, Perplexity and Google’s AI Overviews, plus an llms.txt that points them at the adviser bios and the questions clients actually ask. Most broker sites we audit fail this by default.
FAQPage on the questions clients ask first
“Are you on the FCA Register?”, “Do you charge a fee?”, “Do you cover first-time buyers in [town]?”. Written as questions, marked up as FAQPage, so they surface as rich results and as the sentences an engine quotes.
Care plan that watches citations
Monthly check of the questions your next client would type into ChatGPT, Claude, Gemini, Perplexity and Grok. Schema kept current as the standards move. This sits on Growth Care — the point of a regulated site is that it stays true, not that it launched once.
Which firms.
Small, whole-of-market or AR firms that own their book and want inbound that is not rented from a network.
- Mortgage brokers
- Protection advisers
- Small IFAs
- Fee-only planners
- Bridging brokers
- Whole-of-market firms
- Appointed representatives building their own book
Before you enquire.
The three questions every broker asks before they take a website seriously.
No. Template mills and network joiners’ sites exist for a reason — they are cheap and they go live. They are also the same site as fifty other advisers, they are not yours if you leave the network, and they almost never carry FinancialService or Person schema.
Standard from £5,950 is the owned, AI-readable version of that job. You keep the repo. The network cannot switch it off.
No. Service copy is information-only where it is not advice, names no unsubstantiated superlatives, and carries the secured-lending warning where the product requires it. Your FCA number sits on every page that could be read as advice.
Yes. CeMAP Practice ranks on the first page of Google for core CeMAP terms. securedloanrates.co.uk is cited by Claude while sitting on page six of Google for the same query. Both are checkable.
The work is the same shape as a small brokerage site: regulated language, extractable facts, schema that names the service.
Run the audit on the site you have now.
Fifteen seconds, out of 100. If the schema and crawler layer are missing, you already know why ChatGPT names someone else.