AI Visibility for Universities: Programs, Tuition and Deadlines in AI Answers

Students and parents ask AI assistants the questions they used to type into a search box or ask an admissions counselor. The questions are specific, and the right answer changes every year. Here are realistic examples, with a placeholder name:

• "What is tuition for the online MBA at Example State University for 2027?"
• "When is the early action deadline at Example State this year?"
• "Is Example State's mechanical engineering program ABET accredited?"
• "Does Example State offer a part-time master's in data science?"
• "Is Example State test-optional for fall 2027 admission?"
• "Best affordable nursing programs in Ohio with evening classes"
• "Compare Example State and a rival school for computer science"

The first five name you, so the risk is a wrong fact. The last two are discovery questions, so the risk is not being named at all. Treat them as two separate problems. Write 10 to 20 of these questions for your own institution, sorted by program and by stage (discovery, comparison, application).

An AI answer comes from one of two places. Either the model answers from its training data, or it runs a live search and reads pages first. Retrieval is per question and discretionary: the engine decides, and you can't see which path it took unless it shows citations.

Training data is frozen. As of October 2026, OpenAI's model documentation lists a February 16, 2026 knowledge cutoff for the GPT-5.6 family and August 31, 2025 for the faster Instant path. An answer from that memory can only know the tuition, deadlines and program list that existed before the cutoff. Nothing you publish this month reaches it.

Live retrieval is where you have leverage. If the engine can fetch your current tuition page, it can quote the current figure. If it can't, it falls back on memory or on a third-party page that copied last year's number.

So diagnose before you fix. An outdated figure that was once true points to memory or a stale source. A figure that was never true is a different failure. This guide to hallucination versus outdated answers shows how to tell them apart.

Students use several assistants, and each one gets its pages from a different place. Measure each one on its own, because a single blended score hides the differences.

Google states that a page must be indexed and eligible to show with a snippet to appear as a supporting link in AI Overviews or AI Mode, and that no extra optimization is required. Answers in the consumer apps can also differ from what an API returns, which matters if your team tests through developer tools. See API answers versus consumer app answers.

These are the facts worth checking first, because each one changes on a schedule and each one drives a decision:

• Tuition and fees. Last year's rate, the in-state figure quoted as out-of-state, or a per-credit price shown as a per-term price.
• Deadlines. Early action, regular decision, scholarship and financial aid dates from a prior cycle.
• Programs. Discontinued degrees still described as open, new programs missing, online and on-campus versions merged.
• Accreditation. Program-level accreditation confused with institutional accreditation, or a status that has changed.
• Admissions policy. Test-optional status, application fees and required materials.

The fix that has evidence behind it is simple. State each fact as a plain sentence with its year on a page that can be fetched: "Undergraduate tuition for the 2027 to 2028 academic year is $X per semester for state residents." That gives an engine a sentence it can quote. Then make every page that states the figure say the same thing. Contradictions between the bursar's page, a program page and an old PDF give the engine a choice, and it may pick the wrong one. The reasoning matches whether to publish prices so AI gets them right, and the steps in fixing AI answers that get your pricing wrong apply to tuition.

When an engine cites other sites instead of yours, those sites shape the answer. The evidence shows an association between being named on cited sites and appearing in answers. It doesn't promise an effect. For a university, check:

• College ranking and search directories that list tuition and deadlines.
• Accreditor directories for each accredited program.
• Government education data portals that publish cost and enrollment figures.
• Wikipedia and Wikidata entries for the institution.
• Student forums and discussion threads, where admissions questions get asked and answered.
• Scholarship and program aggregators that copy your figures.

Look at the citations in real answers to your questions, then open each cited page. If a directory shows last year's tuition, request an update through its correction process. Discussion threads are harder to influence. This note on Reddit and AI answers covers what to expect.

A university rarely has one website. Admissions, graduate school, each college, online programs, the catalog and the bursar often sit on separate subdomains run by separate teams. Each subdomain serves its own robots.txt. One blanket rule copied from a template can quietly remove a whole college from ChatGPT search. Audit by property:

robots.txt per subdomain, per bot

Rules are per bot. OpenAI states each setting is independent. This example keeps ChatGPT search eligibility while opting out of training:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

Repeat the check for PerplexityBot, Claude-SearchBot, Googlebot and bingbot on every subdomain.

Firewalls and CDNs

A firewall or CDN can block a crawler that robots.txt allows, often with a 403 response. OpenAI's help documentation says eligibility requires both robots.txt permission and a host that allows its published search crawler IP addresses.

Stray noindex and snippet controls

noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers on Google and Bing. Staging templates often ship with noindex still in place.

JavaScript-rendered catalogs

Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. A program finder or tuition calculator that only fills in after JavaScript runs may be invisible to those crawlers. Compare the raw HTML with the rendered page.

Inherited Search Console controls

Google's generative AI inclusion control is opt-out: inclusion is the default. Child properties inherit the parent's setting, so one choice at the top can affect every subdomain beneath it. Check each property you own. The AI Crawler Checker requests a page as each AI crawler and shows which ones the firewall turns away.

1. Days 1 to 5: inventory. List every subdomain that states a program, tuition figure, deadline or accreditation. Name an owner for each one. Write your buyer questions.
2. Days 6 to 12: access. For each property, check robots.txt per bot, firewall responses, noindex and snippet directives, Search Console settings and Bing Webmaster Tools status. Fix the blocks.
3. Days 13 to 20: facts. Publish one canonical page per fact type with the academic year in the sentence. Make sure it appears in the HTML before JavaScript runs. Remove or update old PDFs and program pages that contradict it.
4. Days 21 to 26: third parties. Submit corrections to directories and aggregators that show outdated figures.
5. Days 27 to 30: re-check. Ask your questions again on each engine and record which answers changed. The free checker on this site gives a snapshot for up to three buyer questions.

Expect memory-based answers to lag until models retrain. You can't change those, so focus on what live retrieval can fetch. Public institutions face similar problems, covered in AI visibility for local government.