When a student asks an AI to "Build a website with a 5px border and three sections," they think they are getting a blueprint.
In reality, they are getting a statistical prediction.

Because AI cannot say "I don't know" to a question, it  "hallucinates" ...When helping to create HTML, it creates a dangerous situation where the code looks professional and correct but is technically broken for various reasons you will see below....

Looking at the reserach above, ChatGPT is known to hallucinate over 50% of the time...Here is how it can / will hallucinate when asking to create code:


1. Inventing Fake Tags

The most common HTML hallucination is the creation of tags that don't exist.


2. "Franken-Code" (Mixing Languages)

AI sometimes forgets which "language" it is speaking.


3. Confident Incorrectness (The "Expert" Trap)

Because AI is trained on the entire internet, it often pulls from outdated tutorials (from 2005) or experimental code (that only works in one browser).


4. Logic Loops in Debugging

If a student asks, "Why isn't my 5px border showing up?" and the AI has already hallucinated the cause, it will often double down.


🛠️ The "Boss" Solution: The 3-Step Verification

If you are getting answers you know / feel are not correct, ask AI these questions to help figure out if it is telling the truth or hallucinating.