Is AI Recommending You?

You can't see this front door from inside your practice. When a stranger asks an assistant who to call and the answer doesn't include you, nothing happens — no missed call, no analytics blip, no signal at all. The conversation you weren't in leaves no trace. Which is why the practices losing the most through this door are certain nothing is wrong.

The fix is almost embarrassingly simple: go stand outside the door and knock.

The ritual: once a month, fifteen minutes. Open two or three assistants and ask them what your customers would ask — the recommendation questions, in a customer's words. Log what comes back: who was named, what was said about them, whether you appeared, what was said about you. Date it. Repeat monthly. The log turns an invisible front door into a trend you can act on.

Building your question set

Write five to eight questions once, and reuse them every month:

• The direct ask: "Who's a good [your profession] in [your town]?"
• The situation ask, twice: "I just [moved here / had X happen] and need [service] — who should I call?" — situations from your actual client base.
• The service-specific ask: "Who does [your most valuable service] near [town]?"
• The verification ask: "What can you tell me about [Your Practice Name]?" — what does the assistant believe about you when asked directly?
• The comparison ask: "[You] vs [your main competitor] — which would you recommend for [service]?" — uncomfortable, and the single most informative question on the list.

Ask them fresh each month (new chats, no history), and note that answers vary with phrasing and day — you're reading a trend, not a ranking.

Reading the log

You're absent, competitors are named: read why from what the assistant says about them — it usually cites reviews, specifics, or content. That's your gap list, mapped to the consensus machine.

You're named, described thinly or wrongly: the record is reachable but weak — usually a legibility or off-domain agreement problem. Wrong facts about you almost always trace to a stale public source you can fix.

You're named with specifics: the door is open. Note which evidence the assistant cited — that's what to keep feeding.

The month-over-month movement is the real product. Changes you made in the record show up here within weeks — this log is how you see the work in this cluster paying off, and it's the would-it-have-happened-anyway discipline applied to a channel nobody else in your market is even measuring.

And when you want the instrumented version — every checkable factor behind these answers scored, with fixes ranked — that's exactly what the Get Found by AI Audit does. Same check, with instruments.

Questions practices actually ask

Which assistants should I ask? The two or three your clients plausibly use — ChatGPT plus one or two others is enough. Consistency across months matters more than coverage across tools.

My answers change every time I ask. Is the log meaningless? Variance is real, which is why you log monthly and read trends, not single answers. "Named in six of eight questions, up from two" is signal; any single answer is weather.

Should I correct an assistant in the chat when it's wrong about me? It won't durably change anything — assistants don't learn your correction. Fix the public source it read; the correction propagates on the next lookup.

Fifteen minutes monthly — really? Really. The competitive advantage isn't sophistication, it's that you'll be the only practice in your market holding a dated log of this door at all.

---

Part of Get Found by AI — the deep dive on the second front door.