How to choose an AI course for your child: 8 questions to ask before you pay

Almost every kids' coding company in India now has "AI" in its course names. Some rebuilt their curriculum. Many did not. So courses that look identical on the landing page can be completely different in the classroom.

This guide is only about AI courses. For coding classes in general, see how to choose a coding class for your child. Here we focus on one question: is the "AI" in this course real, and is it right for your child's age?

The one-line test: does your child touch a real model?

Scratch and beginner Python are good things to learn. We recommend them. (If you are unsure which comes first, see Scratch vs Python for kids.) The problem is only when they are sold as AI. A useful test: by the end of month one, will your child have done at least one of these?

• Collected examples (photos, sounds, sentences) and trained a model to recognise them
• Written code that sends a request to a real AI model and uses the reply
• Tested an AI tool deliberately to find where it fails, and explained why

If the answer is "that comes later," ask how much later. A course where AI appears only in module 9 is a coding course with an AI chapter. That can be fine, but it should be sold as one.

The 8 questions to ask before you pay

1. Will my child work with data and train a model?

AI learns from examples. A child who has trained a model on 30 photos of their own handwriting, and watched it fail on messy writing, understands more about AI than one who has watched ten videos about neural networks. Free tools make this possible even for young children: Google's Teachable Machine trains image, sound and pose models right in the web browser, and the trained model can be exported and used in code. Ask: "What data will my child collect, and what model will they train with it?"

2. Will they write code that uses real AI, at a level that fits their age?

For teenagers, this means writing Python that calls a real model through an API and does something with the answer: a study helper, a chatbot for a school club. For a 10- or 11-year-old, it may be simpler code that uses a trained model inside a game. Avoid courses where the child only drags blocks labelled "AI" and never sees what goes in or comes out. Ask to see a sample project from a student your child's age.

3. Do they learn to prompt well and to check AI output?

Writing a clear instruction for an AI and judging whether the answer is correct is the AI skill children will use most, in school and later at work. A good course teaches them to test a prompt several times, compare answers, spot made-up facts, and explain why an output is wrong. We cover the parenting side of this in our guide to ChatGPT for kids. Ask: "How do you teach children to catch AI mistakes?"

4. How are ethics, safety and accounts handled?

The big consumer AI tools have age rules. As of September 2026, OpenAI's terms require users to be at least 13, with a parent or guardian's permission for anyone under 18. Anthropic's consumer terms for Claude require users to be 18 or older. Google lets a parent turn on Gemini for a child under 13 through Family Link, with some features restricted. A course that asks your 10-year-old to sign up for adult AI accounts on their own is not taking safety seriously.

Ask whose accounts and keys the class uses, what children are told never to type into an AI, and whether bias, deepfakes, privacy and honest homework use are covered. Ethics should show up inside projects ("who could this tool treat unfairly?"), not as one lecture.

5. What will my child have built after 3 months?

The product of an AI course is a portfolio of working projects, not a certificate. Ask: "After 12 classes, what 2 or 3 things will my child be able to open and show me?" A good answer names projects that run outside the classroom, on a phone or a shared link. For a teenager eyeing college applications, a few real projects with code and a short write-up say far more than a certificate.

6. Who teaches, and how many children share the class?

A mentor who has only taught from slides will struggle when a child asks "why did the model say that?" Ask what the mentor has personally built with AI, and whether the same mentor stays with the batch. Then ask the maximum batch size. AI projects break in individual ways (a missing key, a bad dataset), so every child needs time with the mentor on their own screen. In a batch of 25, most children end up watching.

7. What should the first class show you?

A good first AI class has your child doing something within half an hour, ends with something small that works, and includes a moment where the AI gets something wrong and the class discusses why. Warning signs: a talk about "the future of AI," a sales pitch at the end, or a child who cannot tell you what they made.

8. What are the refund and lock-in terms, in writing?

Many AI courses are sold as 12- or 24-month packages, discounted for paying upfront, sometimes via EMI. Read the actual refund clause. Is it counted from the purchase date or your child's first class? Are there deductions? Can you stop monthly? Our coding class guide lists common refund-policy traps in more detail.

What "AI" can realistically mean at 8, 11 and 15

Be wary of any course that teaches the same "AI" syllabus to a Class 3 child and a Class 10 child. What is realistic at each stage:

For a serious teenager, the work should get harder every month. Our piece on why Indian children should learn AI coding, not just use AI explains why building matters more than using.

Red flags that should make you pause

• The certificate is the product. "AI expert" badges are easy to print. Ask to see work instead.
• "AI" only in the name. The syllabus is Scratch animations, basic Python and one lesson on "what is AI."
• Huge batches. Twenty-plus children per mentor, cameras off.
• Long lock-ins. Pressure to buy 12+ months today, or a refund window that closes before your child has had a real class.
• Guaranteed outcomes. Promises of jobs, internships or "the next ChatGPT."

Printable checklist: take this into the sales call

How we approach this at Plural

We run AI classes ourselves, so briefly: our AI Builders track (ages 13–18) teaches Python, LLMs, machine learning and real APIs through projects. Younger children start with logic, directing AI and block coding (8–10) or Python and games (10–13). Classes are live on Zoom with a maximum of 8 children and one mentor per batch. You pay upfront, and if it is not a fit after the first 2 classes, you get a full refund. Hold us to the same checklist.

Frequently asked questions

What is the best age for an AI course for kids?

It depends on what the course means by AI. Around 8, children can train simple image or sound models and learn to give AI clear instructions, without writing code. Around 10 to 12, they can start Python and use trained models in their own games and apps. From about 13, a teenager can write code that calls real AI models through APIs and do basic machine learning on real data.

How can I tell if an AI course for kids is real or just a coding course with a new name?

Ask what your child will do with AI in the first month. In a real AI course they will collect data and train a model, write code that uses a real AI model, or test an AI tool to find its mistakes. If AI appears only as a late module or a video about the future, it is a coding course with an AI chapter.

Is my child too young to use ChatGPT in an AI class?

As of September 2026, OpenAI's terms require ChatGPT users to be at least 13, with parental permission for anyone under 18, and Anthropic requires Claude consumer users to be 18 or older. A responsible course for younger children uses tools designed for their age or runs AI tools through adult-controlled accounts, and never asks a child to sign up on their own.

Do AI course certificates help with college admissions?

Certificates from kids' AI courses carry little weight on their own. What helps a teenager is a small portfolio of real projects, with the code and a short explanation of what each one does and where it fails, which they can talk about confidently in an application or interview.