Thinking Critically About AI

As AI-generated content becomes more common, it is increasingly important to recognize it. Signs of AI-generated text include: overly smooth and polished writing with no personality or strong opinions; repetitive sentence structures or overuse of certain phrases like "It's important to note that" or "In today's fast-paced world"; content that sounds confident but lacks specific sources or verifiable details; and responses that feel generic and could apply to almost any situation.

Look closely at hands and fingers — AI often draws too many or too few fingers, or creates fingers that bend the wrong way. Check for inconsistent details: text in the background that does not make sense, asymmetric jewelry or accessories, or backgrounds that blur strangely. Skin may look unnaturally smooth, almost plastic-like. Look at edges where a person meets the background — there may be blurring or warping.

Watch for unnatural blinking patterns or facial movements. The edges of the face may flicker or shift slightly. Lighting on the face may not match the lighting in the environment. Audio and lip movements may be slightly out of sync. If a video of a public figure is saying something shocking or out of character, be especially skeptical.

AI systems learn from data created by humans, and human data reflects human biases. This means AI can inherit and even amplify prejudices present in society. For example: an AI trained mostly on English-language data may give poorer results for other languages or cultures. AI image generators have been shown to over-represent certain demographics and underrepresent others. AI systems used in hiring, lending, or healthcare may produce results that disadvantage certain groups of people.

Bias in AI is not intentional — it is a reflection of the data the AI was trained on. But it is important to be aware of it, especially when AI is used to make decisions that affect people's lives.

Whenever you receive information from an AI, use this three-step process:

Step 1 — Ask: Does this claim seem reasonable? Use your own knowledge and common sense as a first filter.

Step 2 — Verify: Can I find this information from a trusted source? Check the claim against a reliable source like a government website, a well-known news organization, or a reference book.

Step 3 — Consider: Is this a topic where AI is likely to be reliable? AI tends to be more reliable on well-established facts and less reliable on recent events, niche topics, numbers and statistics, and personal advice.