The comparison stage of the college funnel has quietly moved into AI assistants. Prospective students used to open ten tabs across rankings sites, forums, and .edu pages to shortlist schools. Now they type one prompt into ChatGPT, Gemini, or Perplexity and get a named list of five institutions in a single answer. That answer is the new first impression, and most universities are not in it.
Where AI now sits in the enrollment funnel
Students still use Google, but for a narrower job: navigational lookups once they already know a school's name. The open-ended, exploratory work is where AI is taking over.
- Discovery happens in AI. "Best schools for marine biology on the East Coast." "Undergraduate programs for students interested in game design and computer science." The engine returns a shortlist and the student takes that shortlist forward.
- Comparison happens in AI. "How does Northeastern compare to BU for co-op programs." "Is Fordham or Villanova better for finance." The engine synthesizes program strengths, outcomes, and vibes from whatever sources it can quote.
- Logistics happen in a mix of AI and .edu. "What is Berklee's application deadline." "How much is room and board at Michigan State." The engine answers if the info is easy to extract; otherwise the student clicks through to the site.
The first two stages used to drive most of your top-of-funnel traffic. That traffic is not dropping by half in one quarter, but it is eroding steadily while branded search stays flat. That hides the problem in the numbers.
The query patterns that matter
If you group the questions prospective students actually ask AI assistants, four patterns show up over and over.
- Region plus program. "Best liberal arts colleges in the Pacific Northwest." "Affordable computer science programs in the Southeast." These are the queries where a shortlist gets built.
- Named comparisons. "X university vs Y university for engineering." The engine writes a paragraph on each and often names a third option the student had not considered.
- Constraint stacks. "Nursing programs under $25,000 a year with no SAT requirement." AI is exceptionally good at these because it can filter across many dimensions at once.
- Deep logistics. "What GPA do I need for Cal Poly SLO business," "how does the FAFSA work for graduate school." These pull citations from the .edu when the .edu content is structured cleanly.
The first three are where you either appear as a named school or lose the student before they ever hear of you. The fourth is where a well-structured .edu earns a direct citation.
Who gets cited today, and why it isn't you
Ask any AI assistant a shortlist question about your region and program area and watch what it cites. It will name a rankings site (US News, Niche, Princeton Review), a couple of aggregators, one or two Reddit threads, and maybe one institutional website — often not yours.
There is a reason for that, and it isn't domain authority. AI engines cite content they can extract, verify, and attribute. Rankings sites publish structured lists with schema. Reddit publishes plain-text answers to specific questions. .edu program pages publish prose-heavy marketing copy, PDFs, and information split across dozens of templates. The engines skip the .edu because they cannot cleanly extract a citation from it.
What enrollment teams should do first
Start with a baseline. Pick 20 to 30 questions your prospective students actually ask, pick three to five competitor schools, and run the query set across ChatGPT, Gemini, Perplexity, and Google's AI Overviews. Record which schools each engine names, which sources it cites, and how many of those answers include your institution at all.
Nine times out of ten that baseline is worse than leadership expects, and better than the panic reaction would suggest. It gives you a starting number to move, which is what makes this problem tractable instead of anxiety-provoking.
From there the work is structural: restructure your program pages into the Q&A patterns AI engines extract, ship the schema and llms.txt layer, and track citations monthly against the same competitor set. First citations typically appear inside 30 to 60 days. Meaningful presence takes one to two quarters.
If you want us to run the baseline for you, that is the first step of our higher education AI search visibility service. We can also just point you at the tools and let your team do it. Either way, do not skip the baseline. Also worth reading: our broader primer on how to get cited by ChatGPT and Perplexity.
Seattle, WA
A full-service digital agency working in WordPress, Drupal, Shopify, Webflow, React, and React Native. We partner with universities, governments, and growing brands to ship sites and products that hold up after launch.
