I’m trying to figure out whether a piece of writing I received was created by AI or written by a real person. The wording feels a little too polished and repetitive, and now I need help knowing what signs to look for before I respond. I’m looking for practical ways to check for AI writing without accusing someone unfairly.
There’s no perfect test. Most “AI detectors” throw false positives all the time, esp with clean business writing or non-native writers.
What to check instead:
-
Repetition.
Same sentence shape over and over. Same filler ideas repeated with new wording. -
Safe vagueness.
Lots of polished text, low detail. Few names, dates, numbers, or firsthand specifics. -
Weird consistency.
No rough spots. No personal quirks. No strong opinion. Every paragraph sounds like the same temperature. -
Generic examples.
Examples feel invented and broad, not tied to real context. -
Fact check it.
AI often slips in wrong citations, fake sources, or mushy claims.
Best method, ask for the draft history. Google Docs version history helps a lot. Ask follow-up questions too. A human usually explains choices fast. Someone pasting AI text often gets fuzzy or defensive. That isnt proof, but it’s a better signal than detector sites.
I’d add one thing to what @viajantedoceu said: don’t treat “too polished” as the smoking gun. Some people just write clean copy, especially if they work in marketing, legal, HR, academia, etc. Detector tools love flagging that stuff and they’re wrong a lot.
What I’d look at instead:
-
Ask for a rewrite in a different tone.
If the writer can quickly make it more blunt, more casual, shorter, or more opinionated, that tells you a lot. AI users often submit the first polished version and can’t really steer it well after. -
Check whether the text actually answers the prompt.
AI often circles the topic without fully committing. It sounds relevant but somehow avoids the hardest part. -
Look for “balanced for no reason.”
Sometimes the writing gives both sides of every tiny point, like it’s scared to sound human. Real people are usually a bit messier and pick a lane. -
See if the vocabulary fits the person.
If someone normally writes plain emails and suddenly hands in something full of “moreover,” “furthermore,” and “in today’s landscape,” yeah, that’s worth side-eyeing. -
Ask about one specific sentence.
Why did you phrase it this way? Why this example? A real writer usually has some logic, even if it’s imperfect. AI-assisted stuff gets fuzzy fast.
Also, human writing can be repetitive too. People are weird. So I wouldn’t try to “prove” AI from style alone. You’re mostly looking for mismatch, context, and whether the person can own the text when pressed a littel.