Why are my ads bringing the wrong people, and how do I fix it?

What this looks like from your side

Six of my last 58 sales calls asked some version of this, and it never arrives as a calm question. It sounds like "why am I getting completely wrong leads, 70-year-olds?" or "I'm calling them garbage calls" or "it was just a waste of our time." The ads are running, the lead count looks fine on paper, and the owner is spending lunch breaks dialing numbers that don't pick up, people who want their insurance billed, or someone selling them a website.

Here is the thing I want you to hear first: that is not an ads-don't-work problem. A wrong lead is proof the ad reached a human and the human raised a hand. The machine did its job. It just did it for the wrong person, because something in the ad told it that was who you wanted.

Why the platform sends you those people

Meta and Google don't know what a cash-pay patient is. They learn who to show your ad to from two things: who clicks and converts in the first days, and what the creative itself signals. A stock photo of a grey-haired couple walking on the beach tells Meta "older, retired, Medicare." A headline about "pain relief" with no mention of who it's for pulls everyone with a sore anything, including people whose next question is "do you take my insurance?" A free offer with a one-field form is the easiest thing on the internet for a bot to fill out.

What I see in ad accounts, over and over, is that the cheapest leads were the wrong ones and they trained the campaign. The platform optimizes for whatever converted cheapest, and the early garbage became the target. Your first week set the pattern, and every week after it got better at finding more of the same.

One number from the Campfire Benchmark shows how much the counting method changes this picture. Across its nine spending clinic accounts (Oct 2025 to Sep 2026), the pooled cost per lead was $39.85 when every lead counts, and $61.34 for a patient lead once lead-magnet downloads are removed. Same accounts, same spend; a third of the "leads" were people who wanted a free PDF, not an appointment. If you are judging your ads on raw lead count, the number is lying to you in the same direction.

The fix, in the order I run it

Change one variable at a time. On the calls where this went wrong, the owner had been changing the audience, the offer and the image in the same week, so nobody could tell what fixed or broke anything. One audience, one offer, one channel, one change per week.

• Creative first. Keep the ad that produced your best lead, not your cheapest one, and kill the rest. Put the person you actually treat in the image: if your patients are active 35-to-55-year-olds, the picture is someone deadlifting or running, not someone in a waiting room. On two recent calls we replaced a confusing video ad with simple image ads showing the right demographic doing the activity, built on the call, and that was the fastest fix available.
• Headline second. Name who it's for and what it costs them to be wrong. "Runners in [city] with knee pain who want to keep running" filters harder than any audience setting, and it does it before the click, which is where filtering is free.
• Form third. Add two qualifying questions. "Are you looking to pay out of pocket, or use insurance?" and "How soon do you want to start?" Yes, this makes each lead more expensive. It also makes it a lead. On lead forms, ask them to confirm by text; a bot won't and a tire-kicker often won't either.
• Negatives on Google. If the garbage is coming from search, it's usually keywords. Negative terms for "free," "Medicare," "jobs," "near me cheap" and the names of insurers cut the garbage calls on one account without touching the budget.
• Offer last. If the leads are the right people but still don't book, the problem moved from targeting to the front-end offer, and that is a different page.

What about Meta's health data limits?

Nearly every clinic ads call this month asked this. Meta restricts what health-related businesses can send back from a website, so the pixel often can't tell the campaign who became a patient. That makes the creative and the form questions matter more, not less: they are the filter the platform can no longer learn on its own. Lead forms with qualifying questions, and a CRM that tags the source, give you lead quality by hand when the platform won't give it to you automatically.

How you'll know it worked

Don't look at cost per lead. Look at how many leads you were willing to call back, and how many of those booked. In our benchmark the three accounts that track a booked appointment paid $131.96 per booking, pooled. That's the number that tells you whether the leads are real. If your cost per lead goes up and your cost per booked patient goes down, the fix is working, even though the dashboard looks worse for a week.