We audited two real mortgage companies through our Site Audit & Strategy process. One scored 18 out of 100. One scored 75 — and was still nearly invisible where it mattered. Here's exactly what we found, and what to do about it.
"Who is a good mortgage broker in Mission Viejo for a first-time buyer?"
A first-time homebuyer types that into ChatGPT. Four seconds later, they have three names — confident recommendations, brief explanations of specialties, links to follow up. The whole decision takes under ten minutes.
Your name isn't there. Not because you're less qualified. Not because your rates aren't competitive. Because AI doesn't know you exist in any meaningful, machine-readable way.
We've audited dozens of small business websites through our Site Audit & Strategy process — analyzing how sites appear across ChatGPT, Perplexity, Gemini, and Google AI Overviews. The mortgage industry keeps surfacing the same problem. We saw it twice, in real time, in the last two weeks.
We Audited Two
Mortgage Companies.
Here's What We Found.
Both companies are real. Both have professional teams, real loan programs, and legitimate businesses behind their websites. We're keeping their names confidential — we'll call them Company A and Company B.
112 issues across 10 scanned pages. 38 critical. Every page failed. Zero schema. Zero meta descriptions. The scanner timed out from error volume before reaching the full site.
Only 6 issues. Technically clean. Schema 100%. But no content for AI to cite — meaning this site barely appears in AI-generated mortgage recommendations despite a respectable score.
Company A: What a
Score of 18 Actually Means
One hundred and twelve issues. Thirty-eight critical. Every page failed. Here's the actual summary from the audit findings:
Every category failed — all seven. Here's the scorecard:
The four most damaging issues — pulled directly from the audit:
All 10 pages. AI engines have no structured context to identify, summarize, or cite this business in response to any mortgage-related query.
Google auto-generates snippets from random page text. For a YMYL financial services site, this is the first impression a prospective borrower sees in search results.
The Loan Programs page, About, Contact, and all three team pages have no primary heading. Crawlers have no topical anchor for any of these pages.
Every page has untagged images. The contact page has 43 images with empty alt attributes — each a missed keyword opportunity and AI citation surface.
"This is not a site with isolated technical debt — it is a site with zero foundational visibility infrastructure in place across the board. For a financial services company, this means AI engines like ChatGPT, Perplexity, and Google AI Overviews have no structured context to cite this business."
This is a YMYL site — Your Money, Your Life — how Google classifies financial services. Every missing signal doesn't just leave a gap. It actively erodes credibility with the systems doing the evaluating.
Company B: What a
Score of 75 Actually Means
Company B looked very different on paper. Six total issues. Schema at 100%. Crawlability at 100%. A clean, professionally built Squarespace site with good bones.
"A 75 is a solid technical baseline. But fixing the issues gets Company B close to full technical readiness — and still potentially invisible where mortgage searches actually happen in 2026. This site is a launchpad — clean, professional, and thin. There is no content for AI engines to cite."
You can have a perfect technical foundation and still lose to a messier site that has more to say. AI doesn't recommend websites. It cites sources. And to be cited, you need something worth citing.
The Brutal Truth AI
Doesn't Tell You
When ChatGPT recommends a mortgage broker, it isn't checking your production volume, reading your reviews, or weighing your years of experience. It pulls from whatever structured, consistent, verifiable signals exist across the public web — and responds based on confidence, not quality.
The brokers being surfaced right now are not the most skilled. They are the most structured, the most consistently cited, and the most machine-readable. A broker with 20 years of experience and a wall of five-star reviews can be completely invisible if their digital footprint is thin, inconsistent, or unstructured.
AI answers two questions in rapid succession: Who is relevant to this query? Then: Which of these can I confidently recommend? Company A fails the first. Company B passes it — and fails the second. Together they cover the two failure modes that account for virtually every invisible mortgage broker today.
The 4 Signals That
Determine Whether AI
Recommends You
Your business name, location, and specialty must appear identically across every platform. Company A's name appeared differently across their Google Business Profile, NMLS entry, and website. To AI, that looks like three different entities — and AI defaults away from ambiguity.
Schema markup is how AI knows what you are. Company A scored 0% here — the single highest-impact gap in their entire audit. Without it, AI has no structured context to cite. Company B scored 100% — which is exactly why their 75 was recoverable.
This is Company B's gap. Every FAQ answer, loan guide, and local market page is a citation surface. A broker with 20 FAQ answers has 20 opportunities to appear in AI responses. A thin brochure site has none — regardless of how clean its technical score is.
The NMLS registry is exactly the kind of structured, government-linked data AI systems trust. Your license number prominently on your site — linked to your NMLS entry — gives AI a direct line to credential verification. Most brokers aren't using this at all.
The Platform Problem
Nobody Talks About
Company B's audit surfaced something most brokers never consider: the platform your website runs on is a ceiling on your AI visibility. Squarespace renders most content client-side via JavaScript. AI crawlers often encounter JS-heavy pages and read only a stripped-down version of the content.
"Company B is early enough that migration costs nothing in lost history. There's no legacy to protect. Moving to WordPress now means every article, glossary entry, and FAQ page is built on a platform AI engines can fully read from day one."
Company A — already on WordPress — has the platform advantage. The work is implementing the missing foundation, then building content on top of it. Here's their recovery trajectory:
What Homebuyers Are
Actually Asking AI
These aren't simple "mortgage broker near me" queries. They're nuanced, scenario-based questions — and brokers with rich, specific content dominate every one.
- "Who's the best mortgage broker for self-employed borrowers in [city]?"
- "I have a credit score of 620 — what kind of mortgage should I look at?"
- "What's the difference between FHA and conventional for a first-time buyer?"
- "Is now a good time to buy in [city] with rates where they are?"
- "Can you recommend a VA loan specialist near me?"
A broker with a guide on FHA vs. conventional loans appears in the third query. A broker with a local market analysis appears in the fourth. A broker with a dedicated VA loan page appears in the fifth. Every one of those pages is a citation surface that doesn't currently exist for Company A or Company B.
5 Steps to Start
This Week
Check every platform — Google Business Profile, NMLS, Zillow, LinkedIn, directories — and make your name, address, phone, and license number identical. Not similar. Identical. This is the foundation everything else builds on, and the exact issue that compounded every problem in Company A's audit.
On WordPress, Yoast or RankMath handle most of this. At minimum: LocalBusiness schema with FinancialService classification, address, service area, and NMLS number. Add Person schema on each loan officer's page. This was worth 18 score points in Company A's audit — the highest single-fix impact in the entire report.
Put it in your header or footer: "Licensed Mortgage Broker · NMLS #[number] · [State]." The NMLS registry is government-verified, structured, and machine-readable — exactly what AI uses to corroborate credentials before recommending someone.
Claim and fully complete your profiles on Bankrate, NerdWallet, Zillow, the NAMB directory, and local business directories. Aim for 10–15 completed listings with identical NAP data — each one another verification point AI can cross-reference.
Pick the three questions first-time buyers ask you most. Write a substantive answer to each — 500 to 800 words, structured with headings, direct answer first. That's three new citation surfaces targeting the exact queries your future clients are already asking AI — and exactly the gap Company B needs to close.
Visibility Is No Longer
Optional
Company A didn't know they had 112 issues. Company B didn't realize a technically clean website could still struggle to appear in AI-driven search experiences. Both were operating with blind spots that directly impacted discoverability, trust, and digital visibility.
Our Illuminate Audit identifies the technical, structural, and content-level signals influencing how AI systems interpret your business across ChatGPT, Perplexity, Gemini, and Google AI Overviews.
In many cases, an Illuminate Audit report can be in your hands in as little as 1 business day.
The businesses that win the next era of search won't necessarily be the loudest. They'll be the clearest, the most structured, and the easiest for AI systems to confidently understand and recommend.
The introduction used to begin with a Google search.
Now it begins inside an AI conversation.
Make sure your business is part of it.
Glossary
Terms Used in This Article
Browse the full Firefly Glossary →
