The New Reality
AI-generated text is now indistinguishable from human writing in many cases. AI images win photography contests. AI voices clone people from seconds of audio. AI videos show politicians saying things they never said.
This isn't a reason to distrust everything - it's a reason to upgrade your verification habits.
Spotting AI Text: The Useful Tell
The old tells (phrases like "in today's fast-paced world", em-dash overuse) are fading as models improve. Real detection now works differently:
- Ask where it came from: text posted with no author, no source, no history is more suspicious
- Check the logic: AI writes fluently but can still contradict itself across paragraphs
- Look for missing specifics: real people cite real numbers, names, and dates
- Search a distinctive phrase: quote part of it in a search - AI slop often appears across multiple sites
The honest truth: you can't reliably spot AI text by reading it. You spot it by checking provenance - where it came from and whether the source checks out.
Spotting AI Images: The Practical Checklist
- Hands: AI still struggles with fingers (less than before, but still)
- Text in the image: misspelled signs, gibberish logos, blurry text
- Background details: melted edges, extra limbs, physics that doesn't work
- Over-perfection: that "too good to be true" studio-quality photo of a person who has no other photos anywhere
- Reverse image search: if it's real, it exists elsewhere
Remember: image generators keep improving. Today's checklist catches today's fakes. Verify anything important.
Deepfakes: The Ones That Actually Matter
Voice and video clones are the dangerous ones because they bypass our natural skepticism - we trust faces and voices.
- The family code word (from the safety lesson) is your first line of defence
- Callback rule: any voice call asking for money or urgent action gets a callback on a known number
- Look for the unnatural: odd blinking, audio-video sync issues, weird lighting - then verify, don't just squint
- Ask a question only the real person knows: a real family member knows your childhood pet's name. A clone doesn't.
The Politician / Celebrity Rule
Before sharing a shocking image or video of a public figure, assume it might be fake and check three sources. If it's real, it's everywhere. If it's only on one weird account, it's generated. This single habit will stop 90 percent of the misinformation you'd otherwise spread.
Try This Today
Find any AI image generator's free tier (or ask a chatbot to describe one) and generate a photo of "a person holding a coffee cup, detailed hands". Look closely at the fingers. Congratulations - you've trained your eye on a live example.
The One-Line Summary
You can't always spot AI content by looking, so check provenance, verify the important stuff, and keep the family code word handy.