Thinking Critically About AI

AI is persuasive. It speaks with confidence, uses professional language, and rarely hedges. These qualities make it useful — and make it easy to over-trust. Critical thinking is the essential counterweight.

The confidence trap

AI systems generate responses that sound authoritative regardless of whether the underlying information is accurate. This is by design — a response that sounds uncertain is less useful even when uncertainty is appropriate. The mismatch between how confident AI sounds and how often it is wrong is the central challenge of using it responsibly.

Questions to ask of any AI output

• Can I verify this? Find the primary source. If the AI cites a study, find the actual study.
• Is this recent? Remember that AI knowledge has a cutoff date. Anything that could have changed — laws, statistics, company policies — should be checked.
• Is this relevant to New Zealand? Much AI training data is US-centric. Legal advice, medical guidelines, and cultural references may not apply here.
• Am I the right person to evaluate this? For medical, legal, and financial decisions, an AI output is a starting point for a conversation with a professional — not a substitute for one.

Healthy scepticism versus paralysis

Critical thinking about AI does not mean never using it. It means matching the level of verification to the stakes. For low-stakes tasks — brainstorming, drafting, explaining — AI is fine used at face value. For consequential decisions — health, legal, financial — verify before acting.

Trusting AI less is not the goal. Trusting it appropriately — knowing when it is reliable and when it is not — is.