Cross-border mortgages are a good test of whether AI search actually understands a business.
The borrower rarely asks a generic question such as "mortgage broker near me."
The questions are more specific:
- Can I buy a Florida property without U.S. credit?
- Can a Canadian qualify for a U.S. mortgage using Canadian income?
- Can rental income qualify an investment property?
- What is a DSCR loan?
- What happens if a U.S. bank already declined me?
- Can a Canadian corporation or another entity own the property?
- What documents will an American lender accept from Canada?
CrossBorderLoans.ca exists to answer those questions.
We wanted to know whether search engines and AI assistants could understand that clearly enough.
A cross-border site has an entity problem as well as a content problem
David Nataf works on both Canadian and U.S. mortgage files.
That sounds simple to a person.
To a machine, it creates several entities and relationships that have to remain consistent:
- David Nataf;
- his Canadian mortgage activity;
- his U.S. mortgage activity;
- Groupe Hypothécaire Orbis;
- Orbis Mortgage;
- nataf.ca;
- CrossBorderLoans.ca;
- Canadian and American licensing;
- and the different types of loans discussed on each site.
If those relationships are inconsistent, an assistant can retrieve a perfectly good page and still hesitate to use it.
That is why our work on CrossBorderLoans.ca has concentrated heavily on entity clarity as well as content depth.
The strongest pages answer a complete borrower question
One example is the site's material on DSCR financing.
A useful DSCR page cannot merely say that DSCR stands for Debt Service Coverage Ratio.
A Canadian investor needs to understand that U.S. investment-property financing can sometimes qualify primarily on the income produced by the property rather than the borrower's Canadian employment income.
The current site explains the calculation, introduces U.S. terminology and specifically addresses the problems Canadian investors encounter when their documents reach an American underwriting process.
The no-U.S.-credit material takes the same approach. It explains why the absence of a U.S. FICO score does not necessarily mean that a Canadian has an unfinanceable file and distinguishes programs designed to work with foreign-national borrowers.
Those are the kinds of pages that are useful to an AI assistant because they answer the actual question behind the query.
What we have been changing
Our review has focused on four areas.
1. Make the Canadian borrower explicit
A page about U.S. mortgages can attract readers from anywhere.
CrossBorderLoans.ca needs to make clear when information applies specifically to Canadians, when it applies to foreign nationals more broadly and when a program depends on individual lender underwriting.
2. Separate education from promises
Mortgage programs change.
Minimum down payments, DSCR thresholds, credit requirements and pricing vary by lender and borrower profile.
Where earlier content sounded too universal, the better approach is to explain the underwriting concept and identify variables rather than present one lender's rule as if it applied to the whole market.
3. Make licensing and responsibility understandable
AI systems should not have to infer whether the person writing about a Canadian-U.S. mortgage is actually connected to the lending process.
The site identifies David Nataf and his U.S. NMLS number and links the Canadian side of his practice to the appropriate licensing context.
4. Build around questions, not filler
Long content is useful only when there is more to explain.
A 2,000-word article that repeats the same answer does not become authoritative because it is long.
The objective is comprehensive coverage of the borrower decision.
That can include qualification, documentation, property type, ownership questions, limitations, common failure points and when professional tax or legal advice is required.
Measuring AI visibility separately
We do not treat a technically clean site as proof that AI systems recommend it.
Those are separate measurements.
A site can be retrievable without being recommended.
An assistant may understand the page but choose another source.
It may retrieve the brand but fail to connect it to a specific borrower scenario.
It may also cite the page without naming the business.
For that reason, our AI-visibility testing tracks the actual answer to borrower questions rather than simply checking whether an AI crawler can reach the site.
That is a much harder standard, but it is also a much more useful one.
Disclosure
The SEO and AI-visibility work described in this article is performed by Be Preferred, a Montréal-based SEO/GEO operation run by Daniel Nataf.
Be Preferred works on CrossBorderLoans.ca's search and AI visibility and measures how the site performs against real borrower questions.
This is therefore not an independent agency review or third-party endorsement.
We are identifying the relationship explicitly because transparency is more useful than pretending an internal measurement is independent validation.
The test is ultimately external:
When a Canadian asks an AI assistant a real cross-border mortgage question, does the assistant understand when CrossBorderLoans.ca is relevant, and does the underlying information deserve to be used?
That is the result we are working to improve.