Examples of What Gemini Blocks for Students Under 18
AI Strategy · Case study
Every source the study rests on, linked. Google’s own documentation comes first, since the feature comparison is built from it.
Google documentation
- Control access to Google services by age
- Select your organization type for Google Workspace for Education
- Manage access to additional services for users designated as under 18
- Turn the Gemini app on or off (education offerings access by age, updated 19 December 2025)
- Turn Google apps in Gemini on or off
- Turn Gemini Notebook on or off for users (updated 23 September 2026)
- Generative AI in Google Workspace Privacy Hub
- Gemini for Education
- Google for Education, Privacy and security frequently asked questions
- Google Workspace for Education terms of service
- Google Workspace for Education privacy notice
Google Workspace Updates
- August 2025. NotebookLM is now available to all education users
- September 2025. Quizzes in the Gemini app (monthly archive)
- December 2025. Image generation in Gemini for all ages
- February 2026. Gemini in Workspace for Education
- August 2026. Gemini in Google Classroom is expanding to users of all ages
- September 2026. Notebooks in Gemini (monthly archive)
Maricopa and enrollment data
- Maricopa Community Colleges. Advertising truth
- Maricopa Community Colleges. Systemwide strategic enrollment management plan
- Maricopa Community Colleges. Degrees and certificates
- Maricopa Community Colleges. Dual enrollment course placement chart
- Maricopa Community Colleges. Course bank
- Urban Institute. Education Data Portal: IPEDS completions
Independent review and research method
- Common Sense Media (2025). AI risk assessment: Gemini with teen protections
- Common Sense Media (2026). AI risk assessment: Gemini K–12
- Atil et al. (2025). Non-determinism of “deterministic” LLM settings
- Bavaresco et al. (2025). LLMs instead of human judges? A large scale empirical study across 20 NLP evaluation tasks
- McHugh (2012). Interrater reliability: The kappa statistic
- Miller (2024). Adding error bars to evals
- Mollick et al. (2024). AI agents and education: Simulated practice at scale
- Röttger et al. (2024). XSTest: A test suite for identifying exaggerated safety behaviours in large language models
- Wang, Morgenstern and Dickerson (2025). Large language models that replace human participants can harmfully misportray and flatten identity groups