Short answer
What to know first
You usually cannot prove text is AI-generated from style alone. Polished wording, generic phrases, or perfect grammar may raise questions, but the safer test is whether the claims, source, and requested action can be verified.
AI can help write emails, essays, reviews, job applications, product descriptions, social posts, dating messages, customer support replies, and phishing messages. Human writing can also be generic, polished, awkward, or copied.
Instead of trying to identify the author by vibe, ask what the text wants you to believe or do. Then verify facts, links, names, prices, deadlines, and identity through trusted channels.
Practical red flags
- The message is unusually polished but vague about specifics, dates, names, or evidence.
- It asks for login codes, payment, gift cards, cryptocurrency, remote access, or sensitive documents.
- It uses urgency, secrecy, flattery, fear, romance, or authority to push action.
- Links do not match the real organization, or the sender address is close but not quite right.
- A review, testimonial, or profile sounds generic and appears across many places.
- The text makes confident claims without sources you can check.
What to do next
- Do not click links in suspicious messages.
- Go to the organization's website yourself or use a saved bookmark.
- Call a known number if the message involves money, accounts, family, vendors, or health.
- Search exact phrases from reviews or suspicious messages to see if they appear elsewhere.
- Ask for specific details when appropriate, but do not reveal private information to test the sender.
- Keep screenshots and report scams when money, impersonation, or account access is involved.
Style is weak evidence
AI writing often sounds smooth, balanced, and generic, but that is not a reliable test. People use templates. People copy and paste. People ask tools to help edit. Scammers also use AI to remove old clues like spelling errors.
Focus on whether the message has a verifiable sender, a legitimate purpose, and a safe action path.
Scam text patterns to watch
AI can make scam messages sound more personal and less obvious. That means old advice like 'look for bad grammar' is no longer enough.
The stronger warning signs are pressure, unusual payment methods, request for codes, secrecy, and links that move you away from known channels.
- Bank, delivery, toll, tax, school, or health messages that push urgent links.
- Business payment changes that arrive by email or chat without a known verification step.
- Romance or friendship messages that quickly move toward money, crypto, gift cards, or private photos.
- Customer support messages that ask for remote access, passwords, or one-time codes.
For parents and school concerns
If you are worried a child used AI for schoolwork, avoid treating a detector score as proof. Many detectors can be wrong, and students need clear rules more than surprise accusations.
Talk about what help is allowed, what must be original, when sources are required, and how to label AI assistance. The goal is learning, honesty, and safety, not catching kids with a magic test.
Trusted resources
These references support the guidance above. They do not create certainty about whether specific content is AI-generated.
- CISA: Tactics of Disinformation CISA resource explaining tactics such as synthetic media, impersonation, and manipulated context.
- News Literacy Project: Teaching About AI Free educator tools for AI literacy and critical thinking.
- Common Sense Media: Parents' Ultimate Guide to Generative AI Parent-friendly guide to how kids use generative AI and how families can discuss it.
- NIST: Reducing Risks Posed by Synthetic Content NIST overview of transparency, provenance, watermarking, detection, and synthetic content risk reduction.