AI for Facebook Ads: How Top Brands Produce Better Creative Without Bigger Budgets
November 6, 2025
Google Ads works for medical training programs when you judge it on cost per enrolled student, not cost per lead. Phlebotomy, CNA and medical assistant schools sell seats in a class with a start date, so the account has to learn which clicks became students and in 2026 that means feeding enrollment data back to Google.
The most useful number I can give you on this comes from Allied Medical Training. They're a medical training school whose paid media ran mostly on Google with Meta alongside. Their founder and CEO Sean Ewen described their first six months working with us like this:
"In our first six months with Unicorn, year-over-year revenue grew 70%. One product line went from a 35% decline to a 25% increase: our best year in 15 years."
What I'd point you at in that quote is not the 70% on its own. It's that Sean measured the result in revenue. For a training school that means students who signed up and paid, not leads sitting in a CRM. That is the lens for everything below, because it's very easy to build a Google Ads account around the opposite number.
You're buying filled seats in a specific class by a specific date. That makes this category different from almost every other lead gen business running search ads.
Most of these programs are short. The Bureau of Labor Statistics says phlebotomists typically enter the field with a postsecondary nondegree award and that those programs usually take less than a year (BLS Occupational Outlook Handbook, phlebotomists). CNA and medical assistant programs tend to run on a similar rhythm of frequent start dates and small classes. So your calendar is basically a string of cohorts that each have to fill.
Those are perishable seats in a cohort. A seat in the March phlebotomy class that's empty on day one cannot be sold in April. The person who reaches out the week after the class starts either waits for the next one or signs up somewhere closer. So a month where Google sent you a pile of leads and the class still started half full was not a good month, even if the cost per lead looked great.
This is a lead-to-close business, so two numbers matter and they show up at different times. The banked number is students who start each class. It always lags because people need time to sort out schedules, financing and paperwork. The leading number is qualified leads this week and what each one cost you. It is useful as long as everyone remembers it's a leading indicator and not revenue.
Three numbers decide it: cost per qualified lead, the share of qualified leads that become students and cost per enrolled student. The first one is easy to see in Google Ads. The other two live in your student system and that is why most accounts optimize for the first one by default.
Here's where it goes wrong in a way that's specific to training schools. Google's default click-through conversion window is 30 days. You can set it anywhere from 1 to 90 days depending on the conversion source (Google Ads Help, about conversion windows). A lot of students reach out, go silent, then sign up for the next start date six or eight weeks later. When that happens the class starts after the conversion window closes. Google never connects the new student back to the click that produced them.
So set the window to at least your real lead-to-student time. If you don't know that number yet, pull your last few classes and measure it before you touch bids. A window shorter than your sales cycle doesn't make the account look worse, it makes it blind.
The second fix is getting your students into Google at all. Google's enhanced conversions for leads is built for this. It measures offline sales that came from a website lead, using hashed first-party data from your lead forms to attribute them back to the campaign (Google Ads Help, about enhanced conversions). Once those come back in, you can tell Google which leads were worth paying for. Otherwise it chases whoever fills out forms most cheaply.
There is not a magic spend number here. I'd be suspicious of anyone who quotes you one before they've seen your numbers, because the right budget comes out of your own seats and your own conversion rates.
Work it backward from the class. Take the seats you need to fill for the next start date. Divide by the share of qualified leads that usually become students and you get the qualified leads you need before that date. Multiply that by what a qualified lead costs you today and you have a budget tied to something real.
Here's the math with made-up numbers, not a benchmark. If you need 12 seats and one in four qualified leads signs up, you need 48 qualified leads before the class starts. The budget is 48 times whatever your account pays per qualified lead. If most of those leads come in the three weeks before a start date, your spend should be heavier in those weeks too.
Signal is the other half of this. If you only sign up a handful of students a month, that is not much for Google's bidding to learn from. So it often makes sense to optimize toward an earlier step that happens more often, like a qualified lead, a booked campus tour or an attended info session. You keep uploading new students in the background to check that the earlier step actually predicts them.
The first leak isn't in the bidding at all. It's in the ad copy and most schools never notice it.
Google's personalized advertising policy restricts targeting for employment ads in the United States and Canada. It describes them as employment opportunities or hiring a person for a job (Google Advertising Policies, employment in personalized advertising). Those ads can't target by gender, age, parental status, marital status or ZIP code, though radius and city targeting are still allowed (Google Advertising Policies, restricted targeting in personalized advertising).
Google's page does not mention schools or training programs, so I am not telling you a phlebotomy class is an employment ad. I'm telling you an ad leading with "now hiring" or "get paid while you train" reads a lot like a job offer. ZIP targeting is also how plenty of schools draw their commuting area. Write the ad as a program, not a job offer. You keep the targeting you planned on.
The second leak is the outcome claim. People tend to worry about Google's healthcare and medicines policy. As of October 2026 its sections cover prescription drugs, pharmacies, unapproved substances, recruiting for drug trials, addiction services and a few related categories (Google Advertising Policies, healthcare and medicines). Training programs aren't one of them.
The misrepresentation policy is the one that bites. It bars "claims that entice the user with an improbable result" and failing to clearly disclose the full expense a user will bear (Google Advertising Policies, misrepresentation). So the guaranteed-job line is the riskier sentence, along with a salary promise or a price that leaves out fees, books and exam costs. It also brings in the wrong students, who drop out when reality doesn't match the ad.
The third leak is optimizing for form fills. If the only conversion Google sees is a lead, it gets very good at finding people who like filling out forms. That is not the same group as people who'll show up to a 7am lab. You see it as a cost per lead that keeps falling while the number of new students doesn't move.
The fourth leak sits between Google and your admissions team. A lead that waits two days for a call is worth less than one that gets a call the same hour. Google still gets blamed for a gap it can't control. Before you change anything in the account, check how fast your team actually follows up.
Google captures demand that already exists. Someone searching for CNA classes near them has already decided they want the credential. That's why search tends to be the first channel that works for a training school and why it's usually the right place to start in 2026.
What Google can't do is create demand in people who haven't started searching yet. When you're buying most of the relevant searches in your radius and new students flatten as you add budget, that's the signal to add a channel. Allied's mix is a good example of the shape: mostly Google, with Meta alongside.
Meta can work for education too, though it's a different job. Dr. Z Functional Medicine Academy sells training to practitioners. They saw a 316% increase in ROAS, a 77% reduction in CPA and an 18% increase in AOV on a Meta-only account, not Google. I would never present that as a Google result or as proof a CNA program will see the same thing. A high-ticket practitioner program and a short certification course are very different purchases. What it does show is that education buyers respond to ads they weren't searching for when the offer and creative are built for them.
Whoever runs this should be able to show you cost per enrolled student by class. They should also be able to explain how new students get back into Google. An ad agency can staff search, social and landing pages at once, which is a real advantage when you need all three. The catch is attention, since the agency media buyer on your account is usually managing 10+ other businesses (here's why that number matters).
The 2026 UM Vetting Benchmark shows fewer than 1% of marketer applicants pass our 79-point vetting process across 5 stages. Each Unicorn Marketer runs 3 to 5 accounts at a time. If you're evaluating anyone for this, I wrote up what to look for in a paid media partner. If your current setup already feels off, these are the signs it isn't working.
Apply for a Unicorn Assessment. We dig into your Google Ads account, your tracking, your lead flow and your creative. You will walk away knowing what a new student costs you today and where the leaks are, whether or not we ever work together.
If it's a fit, you meet and approve the specific Unicorn Marketer before any work begins under our Perfect Match Guarantee. You can also switch or walk for any reason inside 60 days.
Yes, when the account is judged on cost per enrolled student instead of cost per lead. People searching for these programs have usually already decided they want the credential, so search intent is strong. The catch is that Google only learns from what you report back. If new students never reach the account, it optimizes toward cheap leads that don't show up on day one.
Not on the policy's own text as of October 2026. Google's healthcare and medicines policy covers prescription drugs, pharmacies, unapproved substances, recruiting for drug trials, addiction services and related categories. Training programs are not listed. The bigger risk for a school is the misrepresentation policy, which bars improbable-result claims like a guaranteed job and requires the full cost to be disclosed clearly.
Google's employment policy page doesn't name schools, so a program ad isn't automatically treated as a job ad. The risk is copy that reads like a job offer. Ads about employment opportunities in the US and Canada can't target by gender, age, parental status, marital status or ZIP code. Keep the ad about the program and use radius or city targeting if you're unsure.
Track new students as the outcome that matters. Then pick an earlier step Google can optimize toward more often, like a qualified lead, a booked tour or an attended info session. Use enhanced conversions for leads to send new students back to Google from your lead form data. Set the conversion window to at least your typical lead-to-student time, since the default is 30 days.
Plan on at least one full class cycle, from the first leads through the start date, before judging it. Cost per qualified lead shows up within days but it's only a leading indicator. Students per class is the banked number. It lags by however long your students take to decide, which for a lot of schools is several weeks.
Usually after Google, not instead of it. Search catches people already looking for the program. Once you're buying most of those searches in your area and new students stop growing with budget, Meta can reach people who haven't started searching. Dr. Z Functional Medicine Academy's education results came from Meta alone, but that was a high-ticket practitioner program, so treat it as proof the channel can work rather than a forecast.
%20(1).webp)