Phase 4: Research, Analysis & Decision-Making · 35 min · ChatGPT (with search), Claude, Perplexity, Gemini

Spotting Hallucinations & Using Web-Connected AI

AI doesn't know if something is true. It knows if something sounds true. Those are different things.

Hiring signal: What changes when you use this: you stop getting burned by AI mistakes because you've built the verification habit — and you know when to use web-connected AI for grounded answers.

What you will learn

Spot the tells of a hallucination in AI output
Know the danger zones where hallucinations are most likely
Build the verification habit for anything that matters
Use web-connected AI tools responsibly and understand what 'grounded' answers mean

The Problem

AI hallucinates. It makes up facts, statistics, citations, dates, and names with total confidence. This isn't a bug — it's how the prediction engine works. When asked for a statistic, it predicts what number would plausibly appear in a helpful answer. When asked for a citation, it predicts what a citation looks like. The result looks real but might not be.

If you're using AI for anything that matters — research, decisions, published content — you need to know how to spot hallucinations and when to use tools that ground answers in real data.

The Tells of a Hallucination

Specific numbers without a source — "According to a 2024 McKinsey study, 73% of companies..." Check: does that study exist? Does McKinsey even publish on this topic?
Perfect-sounding citations — The citation looks exactly right — right format, right journal, right year — but the paper doesn't exist. Always verify citations by searching for them.
Confident claims about niche topics — The more niche the topic, the more likely the AI is hallucinating. It has less training data to draw from, so it fills gaps with plausible-sounding fiction.
Dates that are slightly off — "Apple released the iPhone X in 2018" (it was 2017). Close enough to sound right, wrong enough to matter.
Quotes that sound too perfect — If a quote sounds exactly like what the person would say but you can't find it anywhere else, it's probably fabricated.
Consistent confidence regardless of accuracy — The AI sounds just as confident about a hallucinated fact as a real one. Confidence is not a signal of accuracy.

Unlock the full lesson

You've read the first 2 sections. The rest of this lesson covers The Danger Zones, Web-Connected AI Tools, The Verification Habit, Your Practice Rep — plus a hands-on lab, quiz, and project artifact.

Create a free account to unlock Phase 0 and Phase 1 of every course — no credit card.