The Enrollment Marketing Field Guide.

Download the full guide 24 pages · No cost · PDF

First edition · Q4 2026 and 2027 outlook Updated quarterly · Eight minute read

The argument in four sentences

Enrollment is redistributing rather than shrinking, and the consideration set that decides where a student applies is now assembled inside AI answers twelve to twenty-four months before an inquiry form ever fires. Institutions respond by cutting the awareness spending they cannot measure, which is the spending that was building that consideration set, and then buy harder at the bottom of a thinner and more expensive funnel. The way out is to measure cost per enrolled student rather than cost per inquiry, and to make program pages readable by the machines now assembling the list. Visibility is won at the intersection of a specific program and a specific place, which is the question a regional university can actually win.

Diagram A · The compounding sequence

The loop closes, one cycle tighter.

1. 1The pool of eighteen-year-olds shrinks
2. 2The institution cuts what it cannot measure, which is awareness
3. 3The consideration set stops forming, twelve to twenty-four months out
4. 4Bottom-funnel buying rises in a thinner, costlier auction
5. 5Cost per inquiry climbs, and the CFO concludes marketing is not working
6. 6Next year's cut is deeper

Most private and regional universities run a marketing program built around four things: paid search, retargeting, purchased name lists, and email. That stack harvests existing demand rather than creating it. While the pool was growing, harvesting was enough. It is not enough now, because the students who would have found you on their own are fewer every year.

Connected television is the most underused channel in enrollment marketing, and it has become a performance channel rather than a brand channel. It targets households rather than programs inside the counties your admissions team already works, it measures campus visits through location data partners with enough confidence to survive a CFO conversation, and it produces a usable creative signal in two to three weeks. The practical floor is roughly $60,000 a year; below that you cannot generate enough impressions in any single county to support visit measurement. Digital out-of-home now trades programmatically and pairs well with CTV for frequency, though its measurement stays directional. Streaming audio reaches the parent during a commute at $15 to $25 per thousand impressions, and the message to a parent is a different message: outcomes, cost, safety, and whether the institution will still be open in four years.

Paid search still belongs in the plan, and it should stop being asked to do a job it can no longer do. At $6.23 per click and rising 40% year over year, with AI Overviews absorbing the informational queries that used to produce top-of-funnel clicks, search is now where you capture demand you created elsewhere and defend your brand terms. Your brand campaign is a defensive necessity rather than a performance win. An institution that increases non-branded search budget without increasing awareness investment is bidding harder for the same shrinking set of in-market students, and the cost per inquiry will tell that story within a quarter.

The one we recommend against scaling

Buying undifferentiated name lists at volume does not work, and it has not worked for several years. We have watched institutions load six-figure name buys into Slate, run them through a nurture sequence, and produce inquiry counts that look excellent in a board deck and application yields that are close to noise. A purchased name carries no signal about intent; the student checked a box on a test registration form two years ago. Meanwhile the cost of working those names, in counselor hours and email deliverability and CRM hygiene, is real and rarely counted against the channel. Name buys have a legitimate role when they are narrow, recent, and behaviorally qualified. They are a poor substitute for demand creation, and most institutions are using them as exactly that.

Across the institutions we work with, traditional organic search is compressing as AI Overviews answer more queries directly on the results page. At one private university in the Northeast, organic search sessions fell 8% year over year across the 2025-26 cycle. Read on its own, that number starts a painful conversation about whether the SEO investment is working.

The number sitting next to it tells a different story. Direct sessions at the same institution rose 28% over the same period. AI assistants frequently strip the referrer when they send a user to a site, which means a real and growing share of what your analytics platform reports as direct traffic is people arriving from an AI answer. If you are reading organic and direct as separate stories, you are misreading both.

Engagement rate by source, one institution

ChatGPT		82%
Organic		69%
Paid search		49%
Paid social		28%

This is the most actionable thing in this guide. At the same institution, ChatGPT referrals landed on 694 distinct pages over twelve months, and the homepage accounted for roughly 6% of that traffic. The pages doing the work were individual program pages, including a graduate certificate in school building leadership, an applied behavior analysis master's, the nursing programs at two separate campuses, and a newly funded MSW. AI assistants do not recommend universities in the abstract. They answer questions about specific programs, in specific places, with specific attributes, and they cite the page that answers the question directly.

The assets here are your program pages, and at most institutions those pages were written by the department, last revised three years ago, and say nothing concrete about cost, length, format, licensure, outcomes, accreditation, or how to apply. Rewrite program pages so each one answers, in plain text a machine can lift, what the program is, who it is for, what it costs, how long it takes, whether it runs online or on campus, what license or credential it leads to, and what graduates do. Add question-and-answer blocks phrased the way students actually ask. Mark the pages up with structured data. Confirm your robots configuration allows the AI crawlers you want citing you, because a surprising number of institutions have blocked them by accident.

Diagram B · The 694 pages finding

Nearly all of the visibility sits in the 693 pages beyond the homepage.

Share of ChatGPT-referred sessions across twelve months at one private university in the Northeast.

Figure 2 · Where AI visibility is won, and where it is lost

Prompt type	Example	Typical visibility	What it tells you
Branded	Tell me about [institution]'s nursing program	Near universal	Confirms the model has you. Says nothing about discovery.
Program plus regionThe winnable band	What are the best nursing programs on Long Island	Strong where program pages are specific	This is the winnable question. Fund it first.
Program plus state	Best nursing schools in New York	Weak for most regional institutions	Contested by flagships and national brands. Expensive to win.
Institution category	Best private universities in New York	Very weak	Reputation-driven and slow to move. Do not lead here.
Attribute-qualified	Affordable online MSW programs near me	Strong where cost and format are stated plainly	Cheapest visibility available. Most institutions omit the attributes.

Visibility patterns reflect Happy Cog prompt tracking across client institutions in 2026. Individual results vary by program mix, geography, and competitive density.

Every enrollment marketing program needs one primary metric, and it should be cost per enrolled student, measured by program and attributed by channel, rolled up to the institution. Published benchmarks put this at roughly $1,505 for undergraduate programs and $3,804 for graduate programs. It is the number the CFO already thinks in, because it sits next to net tuition revenue. It is the number a dean can act on, because it is reported at the program level. And it is the number that finally makes awareness defensible, because a channel that produces no inquiries but measurably raises application volume in the counties where it ran will show up here and nowhere else.

Most institutions send data one direction. A student clicks an ad, fills out a form, and enters Slate, and that is where the connection ends. The ad platforms never learn which of those inquiries became an application, an admit, or a deposit, so their bidding algorithms optimize toward form fills rather than enrolled students. Closing that loop with offline conversion imports, feeding deposit and application events back with the original click identifier attached, typically makes cost per inquiry rise and cost per enrolled student fall. That is the trade you want, and it is also the chart you need to explain carefully the first time you present it.

Connected TV, out-of-home, and audio do not produce clicks in any useful volume, and no amount of attribution modeling will make them. The honest method is a geographic holdout test: run the channel in a matched set of counties, withhold it in another, and compare inquiry and application volume across a full cycle. A holdout is less precise than a conversion pixel, but it measures what the channel changed, which a pixel cannot. Design the test before the campaign launches, because a holdout you define retroactively is not a holdout.

Benchmarks: $140 average cost per inquiry, $1,505 cost per enrolled undergraduate, $3,804 cost per enrolled graduate student. Cost per application and cost per enrolled student are different measures; an institution's cost per enrolled student will always be higher.

Central marketing builds awareness of the institution. Programs convert that awareness into applications. The handoff between them is where most universities lose a quarter of their marketing program's value, and the loss is political rather than technical. Central marketing wants consistency, scale, and activities it can measure, while deans want control, responsiveness, and activities they can see. Neither side is wrong, and the standoff costs both of them.

The institutions that resolve this well share three habits. They publish a clear list of what central marketing funds and what a dean's discretionary budget covers, so nobody negotiates from scratch every spring. They provide program marketing as a service, so a dean receives campaign support without hiring an agency or a work-study student to run Instagram. And they report program-level results back to the deans with the same rigor they report institutional results to the board, which earns central marketing the standing to be more directive the following year. One thing does not work at scale: centrally managed social accounts for individual departments. Give departments the tooling, the brand assets, and a quarterly training session, and let them run their own.

Figure 4 · What to centralize, what to give the programs

Activity	Centralize	Program	Why
Institutional brand campaigns (CTV, out-of-home, audio)	Yes	No	Scale and consistency exist only at the institution level
Program search and social campaigns	Yes	Contributes	Central buying power, program-level targeting and budget
New program launch packages	Yes	No	Predictable trigger, standard package, institutional efficiency
Program page content	Template and standards	Subject matter	Marketing owns structure and AI readability, faculty own accuracy
Department social accounts	Tooling only	Yes	Faculty voice and community scale do not centralize
Application deadline campaigns	Yes	No	Institution-wide timing against the existing funnel
Open house and visit events	Yes	Co-hosts	Central promotion, program-level participation
Measurement and reporting	Yes	Receives	Single system, single dataset, reported back to programs

Treat this matrix as the starting point for a conversation with your deans rather than a verdict.

Is college enrollment falling? +

Total postsecondary enrollment grew 1.0% in fall 2025, with community colleges up 3.0% and public four-year institutions up 1.4%, while private nonprofit four-year institutions fell 1.6%. Enrollment is redistributing rather than simply shrinking. Underneath that sits a real contraction: high school graduates peaked at roughly 3.9 million in 2025 and are projected to fall 13% through 2041, unevenly by region.

How many high school students use AI in their college search? +

Forty-six percent of high school students now use AI tools during their college search, up from 26% in spring 2025. Eighteen percent have already removed a college from consideration based on what AI-generated results told them. Among students using AI, 62% use it to find colleges that fit them and 34% say their interest in an institution grew because of what they found.

Why is our organic search traffic falling? +

Traditional organic search is compressing as AI Overviews answer more queries directly on the results page. At one private university in the Northeast, organic sessions fell 8% year over year while direct sessions rose 28% in the same period. AI assistants frequently strip the referrer, so a growing share of what analytics reports as direct traffic is people arriving from an AI answer.

Which pages drive AI visibility for a university? +

Program pages, not the homepage. At one private university in the Northeast, ChatGPT referrals landed on 694 distinct pages over twelve months and the homepage accounted for roughly 6% of that traffic. AI assistants answer questions about specific programs, in specific places, with specific attributes, and they cite the page that answers the question directly.

What is the benchmark cost per enrolled student? +

Published benchmarks put cost per enrolled student at roughly $1,505 for undergraduate programs and $3,804 for graduate programs, with average cost per inquiry near $140. Cost per enrolled student, measured by program and attributed by channel, is the metric that connects marketing spend to net tuition revenue, which is why it survives a board meeting when cost per inquiry does not.

How do you measure connected TV or out-of-home for enrollment? +

With a geographic holdout test. Run the channel in a matched set of counties, withhold it in another, and compare inquiry and application volume across a full cycle. A holdout is less precise than a conversion pixel, but it measures what the channel changed, which a pixel cannot. Design the holdout before launch, because a holdout defined retroactively is not a holdout.

Do purchased name lists still work? +

Not at volume. A purchased name carries no signal about intent, and the cost of working those names in counselor hours, email deliverability, and CRM hygiene is real and rarely counted against the channel. Name buys have a legitimate role when they are narrow, recent, and behaviorally qualified, and they are a poor substitute for demand creation.

What should a university do in the next 90 days? +

Audit the chain from ad click to deposit in the first 30 days. Baseline AI visibility and rewrite five program pages between days 30 and 60. Put one awareness channel under a real geographic holdout test between days 60 and 90. The sequence builds evidence before it asks for money, which is the order that works when budgets are frozen.