What do prospective students actually type when they research a university, and can they complete the task those words describe? We audited 18,000 queries directed at a university and tested whether searchers could finish what they came to do. The top questions concerned cost after aid, job outcomes, and transfer credits, yet task success stood at just 38 percent, meaning most searchers left without the answer they arrived for.
I Discrimen
Students asked about money, work, and credits, while the site answered different questions
The gap in this study is between the questions in the search box and the pages behind them.
The three leading question families in our audit were cost after aid, job outcomes, and transfer credits. Each is a decision question, asked by someone comparing options with real consequences attached. None of them is answered well by a page written to celebrate the institution, however accurate that page may be.
Independent usability research keeps finding the same hierarchy. Nielsen Norman Group, in its guidance for university web professionals, lists whether a prospective student can afford to attend, considering tuition costs and financial aid opportunities, among the top four questions students bring to a university site. Our audit is that finding measured in one institution's own search logs: affordability was not one concern among many, it was the front of the queue.
Readers who work on learning products will recognise the pattern from the other side of the desk, and the selection discipline in how to choose a PR agency for EdTech starts from the same premise.
A search box is a confession booth for unmet needs.
II Machina
Eighteen thousand queries, clustered by the job each searcher was trying to finish
We sorted demand by intent first, and only then judged the content meant to serve it.
The audit analysed 18,000 queries and grouped them into question clusters, families of wording that share one underlying task. A cluster for cost after aid, for example, gathers every phrasing of the same anxiety: what will this cost me once grants and scholarships are counted? Task success was then tested against those clusters, asking whether a searcher following that demand could reach a complete answer.
Sector research supports building around program level demand. Echo Delta, in its Designing for Decisions study of how prospective students rank information on university websites, found that program specific tuition and aid emerged as a top concern across student segments, with tuition and fees for the major ranked as the number one priority for most students.
The analytical habits behind choosing a PR agency for data analytics apply directly here. A query log is behavioural evidence, not opinion, and it punishes guesswork. We let the clusters, not the site map, define what the content owed its readers.
Cluster by the question, and the rebuild list writes itself.
III Exemplum
The study in full: a task success rate far below the benchmark, then level with its top
The figures below are the audit record, including the yardsticks it uses.
Task success before the rebuild stood at 38 percent. The study record sets a Benchmark Task Success median of 65 percent and a top quartile of 81 percent, so the university began 27 points below the median position for comparable work. The three leading clusters, cost after aid, job outcomes, and transfer credits, show where that failure concentrated: the highest stakes questions were the ones the content served worst.
| Measure | University audit | Benchmark median | Benchmark top quartile |
|---|---|---|---|
| Task success before the rebuild (percent) | 38 | 65 | 81 |
| Task success after the rebuild (percent) | 81 | 65 | 81 |
| Queries analysed | 18,000 | Not benchmarked in the record | Not benchmarked in the record |
| Leading question cluster | Cost after aid | Not benchmarked in the record | Not benchmarked in the record |
| Second question cluster | Job outcomes | Not benchmarked in the record | Not benchmarked in the record |
| Third question cluster | Transfer credits | Not benchmarked in the record | Not benchmarked in the record |
fons: Aerovector research record, query log audit, for the figures and benchmarks in this table.
Read the benchmark honestly. A median of 65 percent means a typical result still fails a third of searchers, and the top quartile at 81 percent still fails nearly one in five. Our study describes a site that climbed from the bottom of the range to its recorded top edge.
Thirty eight percent is not a content problem; it is a broken promise at scale.
IV Damnum
Every failed task taxes the application decision that follows it
A student who cannot price a degree will price your silence instead.
Task failure is not a neutral event, because the searcher still reaches a conclusion: this university is expensive, evasive, or not for people like me. Pew Research Center reported in 2024 that only 22 percent of United States adults said the cost of a four year degree is worth it even if someone has to take out loans. Against that scepticism, a cost page that cannot answer cost after aid confirms the suspicion the student arrived with.
The commercial stakes are documented as plainly. Modern Campus, reporting its research with Ruffalo Noel Levitz, found that more than nine in ten high school students use an institution's website to guide their college research, and separate Modern Campus findings reported that 40 percent of high school students lose interest in a college after a poor website experience. The website is therefore not a brochure supporting the decision. For most searchers, it is the decision environment itself.
Multiply a 38 percent success rate by thousands of searchers and the loss stops looking like friction. It is a recruitment programme quietly arguing against itself.
An unanswered price question reads as an answer.
V Genus
Searchers arrive in kinds, and each kind carries a different first question
One homepage cannot serve a transfer student and a first time applicant in the same breath.
Our clusters separated naturally by searcher kind. First time applicants asked about cost after aid and the shape of student life. Transfer students asked about credits, a question with a documented cost when answered badly: a peer reviewed analysis published through PubMed Central in 2019, drawing on federal data, reported that transfer students lost an average of 43 percent of their credits, roughly a full semester of coursework.
Job outcome questions cut across every kind, because they convert an abstract degree into rent, debt payments, and a first salary. Postgraduate searchers asked them most forensically, comparing programmes rather than institutions. Each kind deserves content built for its question order, not a shared page rearranged.
Most of these tasks now begin on a phone, in fragments, between other obligations, which is why the mobile first thinking in how to choose a PR agency for mobile applications matters to a university team as much as to an app team. A cluster page that collapses on a small screen has failed the task before its first sentence is read.
Segment by the searcher, then write to the question they actually typed.
VI Regio
Search is the front door in every market a university recruits from
Your campus may be local; your search results compete nationally by breakfast.
A prospective student in another region meets the university first as a list of results, wedged between rival institutions, government tools, and AI generated summaries. The United States Department of Education publishes College Scorecard so families can compare costs, graduation, and earnings across institutions, and the National Center for Education Statistics offers College Navigator with tuition and outcomes data for the same purpose.
Google's own guidance points the same way. The Google Search Central SEO Starter Guide advises organising content around the topics users look for and writing pages that serve the reader's need first. That is a plain description of what a question cluster rebuild does: it rebuilds the site around recorded demand instead of internal structure.
The comparison set now includes machine written answers, and the visibility questions raised in how to choose a PR agency for artificial intelligence (AI) are therefore admissions questions too. If an AI summary prices your degree using a third party estimate because your own cost after aid page is vague, you have lost the framing of your most important number.
Whoever answers the question owns the first impression.
VII Occasio
A question cluster content rebuild lifted task success from 38 to 81 percent
The fix rebuilt pages around recorded questions rather than polishing the old ones.
The PR action our audit led to was a question cluster content rebuild. Content was reorganised so that each leading question family owned a complete, findable answer: cost after aid worked through with aid applied, job outcomes stated at programme level, and transfer credits explained in the terms a transferring student would use. The writing followed the clusters, and the clusters followed the 18,000 queries.
The recorded outcome was task success rising from 38 percent to 81 percent. Against the study benchmarks, that is a move from well below the median of 65 percent to exactly the top quartile figure of 81 percent. We report the movement as the audit recorded it, for one university and one rebuild, without claiming the same lift waits inside every site.
The scale of the jump invites a fair question, and the honest answer is in the starting point. When success begins at 38 percent, the largest gains come from simply answering, completely and in the open, questions the institution already had the facts to answer. The rebuild was less an act of creativity than an act of alignment.
Answer the question that was asked, and the metric moves.
VIII Structura
A cluster only works when each page carries the whole answer
Structure is the difference between a topic page and a task completed.
Each rebuilt cluster page followed a discipline: the question in the searcher's words at the top, the direct answer early, the supporting detail beneath, and the next step stated plainly. Cost after aid pages showed the arithmetic of aid rather than a promise to discuss it. Job outcome pages named outcomes at the level the query asked about. Transfer pages explained how credits are assessed before asking for an application.
Student behaviour now rewards that completeness in a second arena. Ruffalo Noel Levitz reported in its 2025 E-Expectations research that 68 percent of college bound students had already interacted with AI assistants on college websites. A fragmented answer is now failed twice: once for the human reader, and once for the machine summarising you.
The rebuild also has to survive inside the platforms that host and measure it, which is where the platform discipline in how to choose a PR agency for SaaS companies becomes unexpectedly relevant to an admissions team. Structure, metadata, and measurement are one system, and the cluster is its unit.
A page that answers half the question has answered a different question.
IX Consilium
Audit the queries, rebuild by cluster, and test the tasks again
The method is a loop, and the second measurement is the point of it.
Begin with your own query data, internal site search and search console exports together, and cluster by intent before anyone proposes new pages. Rank the clusters by volume and consequence, because cost after aid at the top of the list is a different emergency from a long tail of timetable queries. Then rebuild the leading clusters completely, and resist spreading the same effort thinly across the whole site.
Test task success again after the rebuild, with searchers given the original tasks, and publish the comparison internally against your benchmark. The full scope of an audit and rebuild programme, from Media Map research to Coverage Reporting, is set out in our public relations services.
| What your audit shows | First rebuild move | Measure to watch |
|---|---|---|
| Task success below the 65 percent benchmark median | Rebuild the top three question clusters completely | Task success on the original tasks |
| Cost questions dominate the clusters | Publish cost after aid answers at programme level | Task success within the cost cluster |
| Transfer questions dominate the clusters | Write the credit assessment process in searcher language | Task success within the transfer cluster |
fons: Aerovector research record for the benchmark figure; decision rows are this post's prescription built on the audit findings.
Keep the audit running after the rebuild, because demand drifts with fees, policy, and the labour market. The clusters you fix this year will not be the clusters that fail next year. A university that reaudits annually treats search as a listening post, not a launch event.
The query log is the cheapest honest research a university owns.
X Quaesita
Questions
What did the search demand audit measure?
We analysed 18,000 queries directed at a university and grouped them into question clusters by intent. We then tested task success: whether a searcher following that demand could reach a complete answer to the question they had asked.
What were the top questions students asked?
The leading clusters were cost after aid, job outcomes, and transfer credits. All three are decision questions, asked by people comparing the university against alternatives with money and time at stake.
How did task success compare with the benchmarks?
Task success began at 38 percent, against a Benchmark Task Success median of 65 percent and a top quartile of 81 percent in our record. After the rebuild it reached 81 percent, level with that top quartile figure.
What was the question cluster content rebuild?
It was the PR action the audit produced: content reorganised so each leading question family owned a complete answer, written in the searcher's terms and structured so the direct answer came first. Cost after aid, job outcomes, and transfer credits were rebuilt as clusters rather than scattered across departmental pages.
Did the rebuild really lift task success to 81 percent?
The audit recorded task success rising from 38 percent to 81 percent after the rebuild. That is the result for one university and one programme of work, reported as a study finding rather than a promise about other institutions.
Where should a university start its own audit?
Start with the queries you already hold, from internal site search and search console data. Cluster them by the job the searcher was trying to finish, rank by volume and consequence, and test whether each leading cluster can actually be completed today.
How often should the audit be repeated?
Repeat it on a regular cycle, because fees, aid policy, and job market questions shift demand every year. Our prescription is an annual reaudit with the same task tests, so movements in task success are comparable over time.
How this guide was built
This post reports Aerovector's own research record for a search demand audit of a university: 18,000 queries analysed and clustered by intent, the leading question families and task success measures stated in the text, the benchmark median and top quartile from the same record, a question cluster content rebuild as the PR action, and the task success outcome recorded afterwards. The study figures are Aerovector's own research record, and identities are withheld, so the institution is described only as a university. The study covers one institution and one rebuild, and its benchmarks come from the same research record. Third party sources were used only to set context on student questions, website use, and transfer credit evidence, and each is named with its year where it appears.
Sources
- Aerovector, Search demand audit research, 2026: the first-party study reported in this post, figures from the practice research record, identities withheld.
- Nielsen Norman Group, Five Questions for University UX Professionals, listing affordability among prospective students top four questions, 2023. https://www.nngroup.com/articles/university-ux-professionals/?lm=children-on-the-web&pt=report
- Echo Delta and EAB, Designing for Decisions, research on how prospective students rank tuition, aid and program information, 2025. https://6124401.fs1.hubspotusercontent-na1.net/hubfs/6124401/Echo_Delta-Designing_for_Decisions.pdf
- Pew Research Center, Is College Worth It, reporting 22 percent of adults saying a degree is worth the cost even with loans, 2024. https://www.pewresearch.org/social-trends/2024/05/23/is-college-worth-it-2/
- Modern Campus and Ruffalo Noel Levitz, E-Expectations research coverage, reporting over nine in ten students using institution websites in college research, 2023. https://moderncampus.com/newsroom/2023-student-expectations-report-press-release.html
- Community College Review via PubMed Central, The Correlates of Credit Loss, reporting transfer students lost an average of 43 percent of credits in federal data, 2019. https://pmc.ncbi.nlm.nih.gov/articles/PMC6820296/
- United States Department of Education, College Scorecard, the federal tool comparing college costs, graduation and earnings, 2025. https://collegescorecard.ed.gov/
- National Center for Education Statistics, College Navigator, the federal directory of tuition and outcomes data, 2025. https://nces.ed.gov/collegenavigator/
- Google Search Central, SEO Starter Guide, guidance on organising content around user needs, 2025. https://developers.google.com/search/docs/fundamentals/seo-starter-guide
- Ruffalo Noel Levitz, E-Expectations Trend Report, reporting 68 percent of students interacting with AI assistants on college websites, 2025. https://www.ruffalonl.com/wp-content/uploads/2025/07/23E72-The-Importance-of-Personalization-From-Interest-to-Application.pdf
XI Incipe
When students cannot find the price, Aerovector rebuilds the answer around their questions
Our Media Map work starts with your query data, clustering real searches into the questions your institution must answer completely, before a Message House shapes how each answer is expressed. A Newsroom Calendar then sequences the cluster rebuild across admissions cycles, and Coverage Reporting tracks how the rebuilt answers perform in search and in the press. The audit discipline in this study is the front end of that programme.
Ranking in AI search results and guaranteed media coverage are both built into how Aerovector works, and together they are what converts attention into leads. For a university, that chain is concrete: the cost after aid answer a family finds in an AI summary or a news story is the answer that decides whether an application ever starts. If your task success looks anything like the 38 percent in this study, send us the shape of your query clusters through the form on our contact page, with your leading question families and current success rate, and we will show you how the rebuild would be structured.
Aerovector turns your search logs into answers students can finish.
Additional Details
Description: Our Research: Search demand audit for a University: what the study measured, what it found, and what the finding changes for teams facing the same question.
Published: 2026-10-25