What Counts as an AI Application? A Classification Lab
“Built with AI” can describe three very different things: code written with an assistant, an application that calls an AI service, or a program that implements a model. This short discussion asks students to separate those meanings before they evaluate a project.
Goal
Classify an application’s relationship to AI and identify the evidence needed to explain, test, and use it responsibly.
Activity
Give students these three scenarios:
- A student uses an AI assistant to help write a calculator, but the finished calculator contains no AI feature.
- A study app sends a user’s question to a language-model API and displays the response.
- A program trains a small classifier from labeled examples and uses the model to make a prediction.
For each scenario, students must:
- classify it as AI-assisted development, an AI-powered application, or a student-built model;
- identify the user, purpose, input, output, and decision made by either the program or model;
- name one test for correctness and one risk involving privacy, bias, unsafe output, or inappropriate automation;
- explain what should be documented or attributed.
Deliverable
Submit a one-page comparison table plus a 100-word answer: Which scenario most clearly counts as an AI application, and what evidence would prove it?
Discussion and safety
An application does not become an AI application merely because AI helped write its code. Students should not enter personal, protected, or confidential information into an AI service. Any external tool must follow school policy, age requirements, and account rules.
Source material
First spotted in the November 20, 2026 PTIR Daily Briefing. College Board’s preview of the redesigned AP Computer Science Principles course distinguishes writing and evaluating code with AI tools from designing, programming, and testing an AI-powered application for real users. Its emphasis on purpose, audience, testing, attribution, safety, and human judgment generated this lab.