Education Platforms and Agent-Driven Course Recommendations
"Find me an online course in data analysis that I can finish in under 8 weeks, costs less than a set budget, and gives me a recognised certificate." Requests like this are increasingly handled by AI agents. For most online learning platforms, the question is whether agents can find the right answer on the site.
When you audit a typical course catalogue, the answer is mostly no.
What Agents Cannot Do
A course catalogue is usually searchable by humans through a filtered browse page. You can select a subject area, difficulty level, and delivery format. The results appear as cards showing the course title, a short description, duration, and price.
The problems for agents are specific.
Course duration is ambiguous. The cards display duration as "8 weeks" or "12 weeks," but this does not indicate hours per week. The actual time commitment varies enormously: an "8-week" course might require 3 hours per week or 15 hours per week. The total learning hours are often listed only on the individual course detail page, inside an infographic rendered as an SVG image.
Pricing is incomplete. The card shows the base price, but many courses have early-bird discounts, bundle pricing, or instalment options. Some courses have a free trial period. None of this is available in the search results or structured data. An agent comparing courses on price is working with incomplete numbers.
Prerequisites are buried. Many courses have prerequisites: prior knowledge, software requirements, or completion of earlier courses in a sequence. This information typically appears two-thirds of the way down the course detail page, in a paragraph of prose. Agents frequently recommend courses to users who do not meet the prerequisites.
Certification details are vague. Courses describe their certificate as "industry-recognised" or "accredited," but the certifying body and specific accreditation are not consistently stated. An agent cannot tell a user whether a course leads to a recognised industry certification or an in-house badge.
The Changes
Add structured course metadata. Give every course Schema.org Course markup including name, description, provider, duration in total hours (not just weeks), price with currency, prerequisites, skill level, language, delivery format, and certification details including the certifying organisation. This gives agents a machine-readable specification for every course.
Add a "Course at a Glance" panel. Give each course detail page a standardised sidebar panel with key facts in a definition list: total hours, hours per week, start dates, price (with any discounts shown separately), prerequisites (as a list of linked course names where applicable), and certification (including the certifying body). Keep this panel visible without scrolling and use semantic HTML.
Add an enhanced search API. Recognising that search must work for agents, supplement a client-side filtered browse page with a public API endpoint that accepts structured queries: subject, max price, max duration in hours, minimum skill level, and certification type. Return JSON with all course metadata. Agents that can make HTTP requests then get precise, filterable results without scraping.
Express prerequisite chains. For courses in a sequence, include explicit hasCoursePrerequisite links in the structured data. An agent recommending a course can then check whether the user meets the prerequisites and, if not, suggest the prerequisite course first. This creates a logical pathway that agents can follow.
The Results You Can Expect
When a catalogue becomes machine-readable, enrolments attributed to AI agent referrals can rise. The increase tends to concentrate in professional development courses (data analysis, project management, digital marketing) rather than hobby courses. This fits the pattern: agents are used most when the user has specific requirements and wants a match, not when they are browsing for inspiration.
Completion rates for agent-referred enrolments also tend to be higher, because better matching at the recommendation stage means fewer students dropping out when a course turns out to be too hard, too easy, or not what they expected. The same structured prerequisite and skill-level data tends to reduce "wrong course" support complaints, because agents make better recommendations in the first place.
What Other Education Platforms Should Know
Course comparison is a structured problem. Users have constraints (budget, time, level, certification requirements) and agents are well-suited to matching against those constraints. But they can only match accurately if the data is available and machine-readable.
Three specific things matter most. First, express duration in total hours, not just weeks. "8 weeks" is meaningless without knowing the weekly commitment. Second, make prerequisites explicit and linked, not buried in prose. Third, specify certification details precisely: which organisation accredits the certificate and what it qualifies the holder to do.
The search API tends to be the highest-impact single change. Agents that can query the API directly bypass all HTML parsing entirely and get clean, structured results. If you can only do one thing, build an API for your course catalogue.