In the 15 years Risepoint has surveyed online learners, few behaviors have shifted as fast as how they find a program. In the education technology company's 2026 Voice of the Online Learner report, the share of prospective students using an AI website or app to research schools nearly tripled in a single year, from 6% to 17%.
Discovery now starts with a prompt
Respondents to the Risepoint survey are mostly working adults, and among the youngest students the change is further along. A national survey of more than 5,000 high school students, reported by Inside Higher Ed, found 46% using AI in their college search, up from 26% the previous spring. A separate figure the outlet cited from software company Adobe put the share of Gen Z who begin a search by prompting a chatbot at 28%. AI tools are becoming a first stop for gathering information rather than a novelty.
When Abby Isles set out to help her daughter build a college list, Inside Higher Ed reported, she described her priorities to ChatGPT and asked it to name schools that fit; her daughter later enrolled at Northwestern, in part at the tool's encouragement.
The Risepoint data points the same way for adult learners: 75% said they prefer self-service research, and AI is the most capable self-service tool yet built. "We're seeing a clear shift toward more self-directed, digital-first behavior, with greater reliance on AI and rising expectations for how institutions prepare students for a changing workforce," said Fernando Bleichmar, CEO of Risepoint.
What changes when the chatbot answers first
The mechanics of AI-mediated search differ from a search engine in a way that matters to institutions. A Google search ends with the user clicking through to a website, where the school controls what it says about itself. A chatbot conversation often ends inside the chatbot. "There are no blue links. They're not clicking onto websites and finding discovery elsewhere," said Alexa Poulin, chief digital officer at Carnegie Higher Education, a higher-education marketing firm, told Inside Higher Ed. The model summarizes, recommends, and moves on, and the institution no longer writes the first impression a student forms.
That hands the opening pitch to a system prone to error. Research last fall from the European Broadcasting Union and the BBC, cited in the same reporting, found AI assistants made significant mistakes in 45% of their responses about news. College information is no more reliable by default. When Inside Higher Ed posed as a prospective student and asked ChatGPT to suggest Christian colleges in the South, the tool did not surface Belmont University, even after the "student" specified an interest in music business, Belmont's largest program. A program can be strong and still be missing from the answer a student actually reads.
Google itself is accelerating the same shift. In January 2026, Google confirmed that Gemini 3 had become the default model behind AI Overviews, the AI-generated summaries that now sit atop most search results and answer many questions directly on the page rather than sending the searcher onward to a website. A SparkToro analysis of Similarweb clickstream data found that 68% of U.S. Google searches ended without a click to any site in early 2026, up from about 60% two years earlier, and that on the more than one in five searches where an AI Overview appears, click-through rates fall by nearly 60%. Google's older "People Also Ask" feature, which surfaces short answers directly on the results page, had already been narrowing that path for years before AI Overviews arrived.
The scramble for AI visibility
Institutions have started to respond with what marketers call answer engine optimization (AEO) or generative engine optimization (GEO), a cousin of the search engine optimization (SEO) they have practiced for years. The goal is to increase accurate visibility in relevant AI searches. One of the first tasks is unglamorous: keep the public-facing website accurate and current, because AI tools tend to invent details when they have to fill a gap, and they can surface outdated pages a human searcher would have skipped.
Universities are calibrating expectations accordingly. Rebecca Shineman, executive director of marketing at York College in Pennsylvania, told Inside Higher Ed that her institution does not expect to appear in every search but hopes to surface for students seeking the affordability and job outcomes it emphasizes.
That work takes time, and increasingly, money. Inside Higher Ed has reported that consultancies and ed-tech vendors are now marketing AI-visibility audits and fixes to universities, turning AEO into a service rather than a task an existing marketing team simply absorbs. A regional institution without a dedicated digital-marketing staff, or the budget to bring in outside help, risks falling further behind, not only the flagships with name recognition to spare but any better-funded competitor that gets to AEO first. The divide that shaped a decade of traditional search engine optimization is reappearing in AI search, and moving faster.
Why regional and online programs have the most riding on it
A flagship with a national reputation can surface in a chatbot's answer on name recognition alone. A regional university whose online programs depend on being discovered by working adults, and which carries no such household name, is far more sensitive to whether a model happens to mention it. For those institutions, and for the online programs the Risepoint report tracks, accurate representation in AI answers is edging from a marketing concern toward an enrollment one.
It is early, and the report data supports caution as much as alarm. Inside Higher Ed noted that only about 7% of students said they turn to a chatbot first when researching a specific institution. For now, AI mostly supplements campus visits, review sites, and conversations with real people rather than replacing them. But the direction is not ambiguous. Adult-learner AI research went from 6 to 17% in a year, and teenage use from 26 to 46%, and there is no sign either curve has, or will, flatten.
