Can we use AI-generated students in our marketing?

For illustrative imagery, yes, as long as nothing suggests they are real students. A generated face attached to a name, a quote or an outcome is a fake testimonial. For stories and results, use real students or graduates who have given written consent, and check what their data rights allow.

What can we claim about graduate outcomes?

Only what your data supports, stated with its basis: which cohort, which period, how it was measured and what share of graduates responded. Regulators in the US and UK treat job and salary claims as objective claims needing evidence. Keep approved outcome statements centrally and block generated variations of them.

Can AI make our courses accessible?

It can do much of the work: draft captions and transcripts, alt text for images and diagrams, simplified summaries and audio versions. People still need to check captions and describe complex visuals properly. Build accessibility into generation from the start rather than auditing a finished course against WCAG afterwards.

AI Content for Education and Edtech

AI content creation for education and edtech.

Education providers and edtech companies use AI to produce course marketing, learning materials and translated courses at a pace manual production struggles to match. The constraints are specific: outcome claims about jobs and salaries need current evidence, student records and images are protected by laws such as FERPA and the GDPR, and learning content has to meet accessibility standards from the start.

Outcome claims: jobs, salaries and completion

The most sensitive words in education marketing are about outcomes: graduates hired, salary uplift, completion rates, 'career-ready in 12 weeks'. The US FTC has pursued for-profit colleges over misleading job and earnings claims for years, and in 2021 it sent about 70 institutions a Notice of Penalty Offenses, which exposes those that make such claims deceptively to civil penalties. In the UK, earnings and employment claims for courses are objective claims that need evidence under the CAP Code. Generated copy produces confident outcome language by default. Keep outcome claims as approved statements tied to a dated data source, with the cohort, time frame and method stated, and ban unqualified promises.

Students' records, faces and work

Education content is full of students, and students are protected. In the US, FERPA restricts disclosure of personal information from education records without consent, and institutions must honour opt-outs even for directory information. Online services for under-13s also fall under COPPA. In Europe, a student's photo, voice or coursework is personal data under the GDPR. Two rules follow for AI work. Never paste student records, grades or submissions into consumer tools to generate case studies or personalised messages. And never present a generated student as a real one: a synthetic graduate with a made-up success story is a fake testimonial. Use real students with written consent, or clearly illustrative imagery without names or claims.

Accessible from the first draft

Learning content that is not accessible excludes students and, increasingly, breaks the law. WCAG 2.1 at level AA is the common benchmark. In the US, the Justice Department's 2024 ADA Title II rule makes it the standard for public schools, colleges and universities, with compliance deadlines extended in 2026 to April 2027 for public entities with a total population of 50,000 or more and April 2028 for smaller ones. In the EU, public institutions fall under the Web Accessibility Directive. For AI-made materials that means captions checked by a person, transcripts, meaningful alt text for diagrams and charts, sufficient colour contrast, readable fonts and equations that screen readers can handle. Generating these at the same time as the content costs far less than retrofitting a finished course.

One course, many languages

Multilingual courses are one of the clearest wins for AI in education: the same curriculum in the languages of every student group or workforce, updated in all of them when the content changes. Quality depends on terminology. Subject vocabulary, from anatomy to accounting standards, needs a glossary per language, and assessments must stay equivalent after translation, so a question does not become easier or ambiguous. The studio localises text into 30 languages with terminology enforced, and dubs lecture and explainer video with timing preserved, so on-screen actions still match the narration. Have a subject expert fluent in each language review assessments and key explanations before release.

Updated 25 September 2026 · General information, not legal advice. Rules change, so check the current text with your legal team before relying on it.