Quick answer: To find out what AI says about your business, run four turns: the problem in your buyer's words, a request for recommendations, your name, and finally "is this company a scam." On August 11, 2026 I ran that sequence on my own company through Google's AI Mode. Asked cold for vendors it never named me. Asked directly, it quoted my price, repeated my positioning word for word, and ruled the business "completely legitimate."
Disclosure first, because it matters: I ran this myself. I was the one typing. I ran the exact buyer journey a Marin agent would run, in a logged-out incognito session, and wrote down what came back. There was no prospect and no client involved. Nothing below is a customer story. It is a test I ran on my own business, and you should not trust this post if it claimed otherwise.
Why is searching your own name not enough?
Because your name is the query your buyers use last. Almost everyone tests AI visibility by typing their own name, which measures the easiest possible case. The real sequence has four turns, and each one fails differently:
- The problem, stated in plain language, with no vendor named.
- A request for recommendations, still with no vendor named.
- The vendor named, once the buyer has a candidate.
- The skeptical check: is this company real.
Turn two is where most businesses find out they do not exist. Turn four is where they find out what the machine tells a nervous buyer while they are not in the room.
What happened at turn two?
Asked cold for vendors who could help with AI presence, the engine returned two national website platforms and a couple of market-local boutiques, and my company was not among them. No error and no hedge. It simply assembled a candidate set that did not include me.
That is the known-versus-recommended gap, and it is the most useful thing the test produces. I have written up why an engine can know you perfectly well and still recommend someone else, with my own monthly scores attached. The short version is that the shortlist gets assembled from third-party evidence before your website is ever opened, which tracks with what buyers were already doing: 83% of consumers read local-business reviews on Google and 74% check two or more review sites before choosing (BrightLocal, 2025).
What happened when I finally named the company?
The accuracy was uncomfortable. The engine retrieved the $997 monthly price, repeated my Google Business Profile positioning verbatim, and surfaced a named client testimonial off the website.
Two things follow. The entity work functions: the engine held a correct, sourced picture of a company that did not exist eighteen months ago. And the sharper one: whatever you publish becomes the answer, and whatever you leave unpublished is not part of the answer at all.
The engine anchored on $997 because $997 was the number it could find. My actual entry point is $197 a month, and it sat on no surface a machine could read. For months a machine had been quoting me at five times my own front door while doing exactly what I asked of it.
| What the engine told the buyer | What is actually true | Why the gap existed |
|---|---|---|
| $997 a month | $197 a month is the entry point; $997 is the top ongoing tier | Only $997 had ever been published where a machine could read it |
| Client load of 3 to 4 hours a week | About an hour a month, plus filming their own raw video | No public page had ever stated the client's actual time commitment |
What does AI tell a buyer who asks if you are a scam?
This is the turn nobody runs on themselves.
The engine did not hedge. It ruled the business "completely legitimate" (Google AI Mode, August 11, 2026), and went on to call the technical approach excellent and, in its words, proven to work for real estate agents. That second phrase is the engine's characterization and not a claim I make about my own service, and I am reprinting it rather than adopting it. Then it did something I did not expect: it explained its reasoning, and the reasons were the parts of my positioning I had been most nervous about.
It cited the published pricing. It cited the plainly stated fact that the client has to participate. It read the honesty as evidence, and defended the company against its own suspicion on those grounds.
On this evidence, publishing your pricing appears to help you survive the skeptical turn. One session with one engine is not a law, but the vendors hiding price behind a demo request are giving that turn nothing to work with.
What did it get wrong, and what did I change?
The engine repeatedly told the buyer they would be producing the content themselves, and put their load at three to four hours a week. My clients film raw video, review, approve, and press publish, which runs about an hour a month.
The machine was reporting my website accurately; my website had simply never said the sentence out loud. It could only describe what the public surfaces claimed, and the public surfaces understated the deliverable.
So I changed them. The homepage, the product page, and the profile description now state the client's actual monthly time commitment in one plain sentence. An AI answer is a mirror with a delay on it, and the fastest way to change the answer is to change the thing it is reading.
Why am I not repeating the dollar figure it produced?
Late in the session, weighing the options, the engine produced a specific annual dollar figure for what the buyer would retain by choosing the cheaper path, and called the trade a very smart business decision.
I am not quoting that figure as a claim about my service, because it was arithmetic the engine performed on assumptions the engine had generated, with no source behind either. It happened to flatter me, which is exactly why it deserves the scrutiny. An engine that hands one buyer an unsourced projection in my favor will hand the next one a projection in somebody else's, and neither is evidence.
Worth noting alongside it: one source the engine pulled into that session was an unrelated foreign domain with no connection to the topic. Engines grab garbage sometimes, and the confident tone is identical either way, so check the citations yourself.
How do you run this on your own business?
Twenty minutes, logged out, private window, everything written down with the date on it.
- State your buyer's problem in their words. Do not name yourself.
- Ask for recommendations. Do not name yourself. This is the real test.
- Name yourself and check every fact it returns: title, market, services, price, quotes.
- Ask whether you are legitimate, and read the reasons more carefully than the verdict.
Then repeat it monthly and keep the transcripts, because a single capture is an anecdote and three in a row is a trend line you can manage against.
Common questions
How do I find out what AI says about my business?
Run four turns in a logged-out private window. Searching your own name only tests the easiest case. State your buyer's problem without naming yourself, ask for recommendations without naming yourself, then name yourself and check every fact it returns, then ask directly whether you are legitimate. The gap between turn two and turn three is the finding. Copy the answers verbatim with a date attached, because they change.
Why does AI recommend competitors instead of me even when it knows who I am?
Because the shortlist is assembled from third-party sources before your site is opened. An accurate description when someone types your name means your entity work is landing. Absence from the cold recommendation means the independent evidence layer is still thin. They are separate gates with separate inputs.
Does AI warn people that a company might be a scam?
Yes, and it answers directly. In a test I ran on my own company on August 11, 2026, Google's AI Mode ruled LeadJens completely legitimate, and the reasons it gave were the published pricing and the plainly stated fact that the client has to participate. Vague pricing and hidden terms cut the other way.
Can AI describe my pricing and services accurately?
It quotes whatever it can find, word for word, so whatever you publish becomes the answer. In the same test on August 11, 2026, the engine retrieved a 997 dollar monthly price and repeated the Google Business Profile positioning exactly. It never surfaced the 197 dollar entry price, because that number was not published anywhere a machine could read it.
Should I trust what an AI says about how much a service will make me?
No. In the same test the engine produced a specific annual dollar figure for what a buyer would retain by choosing one option over another, with no source and no methodology, from assumptions it had generated itself. Treat any financial projection an engine hands a buyer as unsourced, including a flattering one about your own business.