Key Takeaways
- AI assistants use two separate mechanisms for local businesses: a proximity-based places component that lists nearby businesses, and a content-driven prose component that names specific ones.
- Showing up in a local map pack and being directly cited by an AI are not the same outcome, and they do not require the same strategy.
- Specialised, problem-specific content is what produces a direct AI citation. Proximity alone did not produce one anywhere in this study.
- Some AI assistants have no places component at all. In those systems, proximity provides no advantage, and only citable content earns a mention.
- Research documented at Autonomous Growth shows that a single well-written page addressing a borrower’s specific situation was cited in a market where a much larger local broker presence produced no individual citation at all.
- For mortgage brokers, real estate agents, lawyers and other local service providers, the difference between being listed and being named by an AI comes down to what is written on the website, and whether it matches the exact problem a potential client is already asking about.
Proximity and Content Play Entirely Different Games
When someone types a question into an AI assistant, two different things can happen behind the scenes. One system surfaces businesses that are geographically close and match a broad service category. Another reads web content looking for a source that actually answers the question being asked.
Those two systems do not talk to each other. A business can appear reliably in the first and be entirely absent from the second. Understanding that split is the starting point for any serious local content strategy in 2026.
What AI Actually Does With “Near Me”
Proximity-based queries trigger what is essentially a map pull. The assistant surfaces businesses that are nearby, in the right category, open, and reasonably well reviewed. The result looks familiar: a handful of names, star ratings, and maybe a one-line description.
The Places Component vs. The Prose
Some AI assistants visually separate these two outputs. A panel appears above the answer with business cards carrying names, ratings and categories. Below it, a written answer appears. The names in the panel and the names in the prose sometimes overlap, but they do not come from the same logic.
The panel finds businesses. The prose cites sources. Those are different jobs, and a business that optimises only for the panel leaves the prose entirely to chance.
Listed Without Verification of What You Actually Do
When a proximity-based component surfaces a local broker for an ITIN mortgage question, it does not verify whether that broker offers ITIN programs. The language gives it away. Businesses are described as ones that “may discuss” ITIN programs, or “could review” bank statements, or “may explore” non-QM options. Modal verbs signal that the system is working from category, not from confirmed content.
In one documented case from Autonomous Growth’s two-city AI study, an assistant asked about ITIN borrowers listed five local brokers and then stated, in plain English, that it had found them as active mortgage brokers in the area and would not assume any of them actually offered an ITIN program. That disclaimer is proximity-based listing working exactly as designed. It is not a recommendation.
Proximity Is Necessary, Not Sufficient
None of this means proximity does not matter. For local service businesses, it is the entry ticket. Without a properly configured Google Business Profile, with accurate categories, a correct service area and consistent name, address and phone data, a business will not appear in proximity-based results at all. That visibility has value, particularly for high-intent, low-specificity queries such as “mortgage broker near me”.
Relevance, Prominence and Reviews Fill the Gaps
Google’s local ranking system weighs three factors: proximity, relevance and prominence. Proximity gets a business into the room. Relevance, meaning how well the category and services match the query, narrows the field. Prominence, meaning reviews, authority signals and overall online presence, influences the order. Businesses that treat a Google Business Profile as a checkbox exercise and ignore the content side tend to plateau here: visible, but never cited.
How AI Decides to Name a Specific Business
When an AI assistant names a specific person or business in its prose response, it is because a web page made that citation possible. The page answered the question being asked, in enough detail that the assistant could point to it as a source.
Specific Problem Pages, Specific Citations
A documented study comparing AI responses in Cleveland and Phoenix illustrates this. In Cleveland, a self-employed borrower question returned a places panel showing three businesses, and a prose section that named one individual: David Goldberg, a loan officer at Mutual of Omaha Mortgage’s Seven Hills office, cited directly to goldbergmortgage.com, which hosts a page specifically about bank-statement loans for self-employed borrowers in Northeast Ohio.
He was not in the panel. His employer’s office was. The panel found the business by proximity. The prose found the person because of the page.
The URL patterns of the cited businesses in that study tell the same story: /bank-statement-loans, /mortgage/self-employed, /dscr-loans-cleveland, /itin-home-loan-arizona, /dscr-rental-phoenix-az. Every cited address was the borrower’s problem, turned into a page.
National Sources Win When Local Content Is Absent
In Phoenix, a market with significantly more independent brokers and greater local visibility than Cleveland, the same self-employed borrower question returned five local brokerages in the panel and no individual name in the prose. The citations went national instead: state-level product pages, lenders in other time zones, bank-statement content from large sites. More brokers, more proximity, more local presence, and still no local citation. Nobody had written the page.
AI Assistants That Operate Without a Places Panel
Two of the three assistants tested in that study have no places component whatsoever. No panel, no proximity pull, no business cards with star ratings. In those systems, the only path into a response is to have been cited. Across both cities and all six borrower situations, those two named no individual mortgage professional at all, including for the easiest question in the set: a first-time buyer with good credit and stable employment.
A business that has directed its entire optimisation effort toward its Google Business Profile is, in those assistants, invisible.
Map Pack vs. AI Citation: Different Intent, Different Outcome
Near-me listings: high intent, low specificity. Someone searching “mortgage broker near me” has purchase intent but has not defined a problem. Panel visibility suits that audience and remains worth pursuing, though the enquiries reflect that lack of specificity: more comparison shopping, more price sensitivity, more friction before anything is agreed.
Direct citations: expertise matched to an exact need. Someone asking an assistant “I filed Chapter 7 two years ago and want to buy a house, who can actually help me?” has already defined the problem in detail. The business named in that answer is not sitting in a list of five near-identical options. It has been identified as the one that handles this situation. Whether that converts better is not something this study measured, but the borrower arrives having stated their circumstances rather than their postcode.
Write the Page That Is the Borrower’s Problem
Every niche situation a client could walk in with deserves its own page. Bankruptcy two years out. ITIN with no social security number. Self-employed with income a tax return understates. DSCR rental financing without personal income qualification. A divorce buyout refinance. Not a paragraph buried in a general services list, but a dedicated page, written in the language the borrower uses, answering the question they are already asking an assistant at eleven at night.
That is what earned a citation in every case this study documented. Not a better star rating. Not a tighter service-area radius. A page that states the problem clearly enough that an assistant can point to it.
What the Study Cannot Tell You
The research covers two metropolitan areas on a single day, with one run per question and no repetition, so it cannot report how much answers vary between sessions. Anyone repeating it next month will get different names.
It also contradicted its own starting hypothesis. The expectation was that borrowers in difficult circumstances would find no named professional. That held in Cleveland and reversed in Phoenix, and the hypothesis was dropped as a result.
The gap between a business that is listed and one that is named is, on this evidence, a content gap. Close it one page at a time.
The full study, including all six questions and the method, is at https://www.autonomousgrowth.io/being-listed-is-not-being-recommended
Autonomous Growth ( part of RReputatioNN )
109 Sint-Lenaartsesteenweg #1
1
Rijkevorsel
Antwerpen
2310
Belgium
