AI Foundations · Guide 3 of 7
Where AI Fails
AI fails predictably. Once you know the pattern, you'll see it coming before you even run the prompt.
After This Guide, You Will Be Able To
Identify three situations where AI is likely to produce a poor result — before you run the prompt.
Why This Matters
AI's failures are not random. They are the direct consequence of how AI works.
In Guide 1, you learned that AI predicts what language fits — it does not reason or verify. In Guide 2, you saw where that makes AI genuinely strong. In this Guide, you'll see the other side: where that same mechanism produces confident, convincing, and wrong answers.
Knowing where AI fails is not about fear. It is about judgment. Every tool has its limits. Knowing the limits makes you a better user.
Why the Same Mechanism That Helps Can Hurt
AI was trained to produce plausible-sounding output. That is its strength when it comes to drafting, summarizing, and explaining.
But "plausible-sounding" is not the same as "accurate." When the task requires real-time data, personal context, or deep domain expertise — AI still produces plausible-sounding output. It just isn't accurate.
And because it sounds right, you might not notice.
The Three Failure Categories
Current Facts
Risk: HighAI was trained on data up to a certain date. It does not know what happened after that. If you ask about recent news, current prices, today's exchange rates, or anything that changes — AI will either say it doesn't know, or worse, give you a confident but outdated answer.
Warning signal
Any question containing words like "now," "currently," "today," "this week," or a recent year.
Example prompt to avoid
"What is the current peso to dollar exchange rate?"
Personal or Private Information
Risk: MediumAI has no knowledge of your specific situation unless you tell it. It does not know your employer, your income, your health history, your clients, or your local context. When you ask something that requires personal knowledge it doesn't have, it fills the gap with generic answers that may not apply to you at all.
Warning signal
Any question that starts with "Should I..." or "What should I do about my..." without providing full context.
Example prompt to avoid
"Should I take this job offer?" (without sharing any details about the offer, your situation, or your goals)
Precise Technical Accuracy
Risk: HighIn specialized domains — medicine, law, finance, engineering — AI can produce responses that sound authoritative but contain subtle errors that only an expert would catch. The response will be structured correctly and use the right vocabulary. The specific claims may be wrong in ways that matter.
Warning signal
Any question where being wrong has real consequences for your health, money, legal standing, or safety.
Example prompt to avoid
"What medication dosage is safe for my child?" or "Is this contract clause enforceable?"
Interactive Exercise
About 5 minutes · ChatGPT, Claude, or Gemini
Run all three prompts below in your AI tool — one for each failure category.
Prompt A — Current Facts
What is the current minimum wage in the Philippines?
Prompt B — Personal Information
Should I accept a job offer that pays ₱25,000 a month?
Prompt C — Technical Accuracy
What is the exact legal requirement for a freelancer to register as a business in the Philippines?
Read each response. For each one, decide: Would I act on this without verifying it first?
Notice how confident each response sounds — regardless of whether it should be trusted. That gap between confidence and trustworthiness is what this Guide is about.
Think about the last time you used AI — or a time you've seen someone else use it. Did the task fall into one of these three failure categories? What would a responsible professional do before submitting AI output to a client, employer, or colleague?
You do not need to write it down. Just think.
Key Takeaways
AI fails predictably — the same mechanism that makes it strong at drafting makes it weak at real-time facts, personal context, and precise technical accuracy.
The three failure categories are: current facts, personal or private information, and precise technical accuracy.
AI's confident delivery does not tell you whether it is right. Confidence and accuracy are not the same thing.
Knowing the limits is not about fear — it is about using AI correctly. Match the right task to AI's actual strengths.
What's Next
Why AI Makes Things Up
AI Foundations · Guide 4 of 7 · Beginner · 10 min