The client
A Texas-based DSCR lender serving real estate investors, operating in the Austin metro area. Established business, real track record, real testimonials on their own site. The kind of company a referral would send you to without hesitation.
What we measured
Ten standardized questions, asked once on each of three AI platforms, in a clean session with no personalization and no prompting designed to produce a mention. Questions covered direct queries such as "tell me about [company]" and "are they legit," recommendation queries such as "best DSCR lenders in Texas" and "who do you recommend for investment property financing in Austin," and service queries specific to their offerings.
Every answer was captured verbatim and scored against a fixed rubric: Visibility 40 points, Accuracy 25 points, Competitive standing 20 points, Foundations 15 points.
The score
19/100
Baseline score. Absent.
0
Times named across 21 buying-intent answers
133
Lender recommendations in those answers, all to competitors
Across ChatGPT, Gemini, and Perplexity, the client was named zero times. The recommendations went most often to the same two firms, appearing in nearly every platform and query type.
On the client's own home-market question, asked with their city in the prompt, six competing firms were named. One competitor's full street address was quoted inside the answer. The client's was not.
What AI actually said
Asked directly whether the company was legitimate, one platform concluded there was not enough independent evidence to confirm it, and listed what it had looked for and not found: a licensing record, a business profile, and a review history on an independent platform.
A second platform did not recognize the company name as a business at all. It read the name as a generic industry term, recommended four other lenders instead, then advised the reader to search by state or city to find the correct entity.
Where the client was described accurately, it was because a platform was reading their own website back to them, verbatim, and repeating it as fact.
Why this happens
AI systems do not rank pages. They assemble an answer from whatever evidence they can find and corroborate about a business. In this case, that evidence had four holes.
- No licensing record surfaced anywhere in open search, and none appeared on the client's own site. Every platform independently told the reader to go verify one before proceeding.
- No Google Business Profile surfaced. Every competitor named on the home-market question had a locatable local record to match against. The client had nothing to match against.
- At least four name and address variants were in public circulation, across the website, FAQ copy, and press materials. Every additional variant splits the evidence into a smaller, less confident pile.
- The only review evidence that existed publicly lived on the client's own domain. No independent platform carried a single review. One AI platform named this gap directly as its reason for withholding confirmation.
What we are doing about it
Three moves, in the order the measurement supports: publish the licensing record where every platform is already looking for it, build a review record that lives somewhere other than the client's own website, and collapse the name and address variants into one consistent identity across every public surface.
This is measured monthly, using the same ten questions and the same scoring, so what changes is visible rather than asserted.
Where it stands
Baseline measured. Remediation underway. Re-measurement scheduled at 30 days.
Published anonymously at the client's request. AI Visibility Benchmark, Methodology v1.0. Figures are readings from a fixed question set on the dates shown, not a guarantee of future results. AI platforms change how they read and weight sources without notice.