AI WORLD HQHow to Spot AI-Generated Scam Messages Before You Click or Reply
A practical five-minute checklist for checking suspicious texts, emails and DMs with your usual AI assistant
In short: Modern scams often use AI to write surprisingly convincing messages, but you can paste the message into any AI assistant (ChatGPT, Gemini, Claude, Copilot, Le Chat, DeepSeek — whichever you already use) and ask it to flag the warning signs. This guide walks you through a simple five-step check before you click a link, reply, or send money.
Scammers used to write in clumsy English. Today, AI helps them write polished messages that look exactly like a real text from your bank, Australia Post, or a family member. The good news: the same AI assistants we use every day are also very good at spotting these tricks when you ask the right questions. In the next five minutes, you'll learn a repeatable check you can run on any suspicious message.
💡 Tip: tap a step’s number when you finish it — a green tick appears and your browser remembers how far you got.
- An account with at least one AI assistant you can chat with — ChatGPT, Gemini, Claude, Microsoft Copilot, Le Chat, DeepSeek, or similar. The free tier is enough for this.
- A phone or computer where you can copy the suspicious message and paste it into the chat.
- The original message itself (screenshot or text), plus the sender's number, email or username if you have it.
- Roughly 5 minutes for your first check; faster once you've done it two or three times.
A quick honest note: an AI assistant can flag likely warning signs. It cannot guarantee a message is safe, and it cannot detect every new trick. When in doubt, treat the message as suspicious and contact the organisation directly using a phone number from their official website.
Don't click anything yet; copy the message safely
The very first action is the most important: do not tap any link, reply, or call any number inside the suspicious message. Scam links can take you to convincing fake sites, and replying confirms your number is real. Instead, take a breath, then open the message and copy the full text. On most phones, press and hold the message bubble and look for a Copy option in the small menu that pops up. On a computer, highlight the text with your mouse and press Ctrl+C (or Cmd+C on a Mac). If the message includes an image (a screenshot of an invoice, for example), take a screenshot of that too — look for a button like Share or Screenshot on your device.
You'll know it worked when the suspicious text is sitting in your phone or computer's clipboard, ready to paste, and you have not tapped any link inside it.
Open your AI assistant in a separate window
Open a fresh browser tab or your AI app — whichever you normally use — so you can see the suspicious message and the assistant side by side. The reason for a separate window: if the scam message came through email, replying inside your email app can be risky, and a fresh chat keeps things clean. Look for the chat box, which is usually a large text field at the bottom of the screen, sometimes with a placeholder like "Message ChatGPT" or "Ask Gemini". If you don't see a chat box, look for a button like New chat or a + icon at the top-left. If your assistant is on a paid plan, this check works the same; free plans handle short messages easily.
You'll know it worked when you can see an empty chat with a text field ready for typing, and the original message is still in your clipboard.
Paste the message and ask for a scam-spotting breakdown
Tap inside the chat field and paste the message with Ctrl+V or Cmd+V (or long-press and Paste on a phone). Then, on a new line under the pasted message, ask the assistant to look for scam warning signs. A good prompt is short and direct — you don't need any special wording. If the assistant offers extra features (like web search, voice mode, or "deep research"), you can leave them off for this task; the default chat is plenty.
You'll know it worked when the assistant replies with a numbered or bulleted list pointing out specific things in the message (for example: "1. Urgency — the word 'urgently' is a pressure tactic").
Ask the assistant to check the link and the sender
Now add the sender's details and any link, and ask the assistant to look at those too. The From field on an email, the phone number on a text, or the username on a DM can all reveal patterns — for example, a "PayPal" email sent from a Gmail address, or a "bank" link that doesn't start with the bank's real website. Paste any link in full, exactly as it appeared (don't tap it). Then ask the assistant three things: who the sender claims to be, whether the link looks like a real address for that organisation, and whether the sender's address matches the company they say they represent. If a tool can browse the web, you can also ask it to summarise what the public link actually points to.
You'll know it worked when the assistant flags the mismatch — a Gmail address for a big company, or a link that almost spells the brand but is slightly off.
Read the summary with your own judgement
The assistant's list is a starting point, not a verdict. Take a moment to read its answer and ask yourself: do these warning signs match my gut feeling about the message? If the assistant flagged several red flags (urgency, money request, weird sender, off-brand link), treat the message as a scam — don't reply, don't click, and don't send any money. If you want to verify, contact the real organisation using a phone number from their official website (type the address into your browser yourself; don't use any number from the suspicious message). Delete the message afterwards. If the assistant found no red flags but the message still feels wrong, trust your feeling — scams evolve faster than any checklist, and no AI can promise a 100% clean bill of health.
You'll know it worked when you can clearly say to yourself either "I'm confident this is a scam, I'm deleting it" or "I'm confident enough to verify it through the official channel — and I still haven't clicked the link in the message."
- Tapping a link "just to look". A scam site can install harmful software the moment it opens, or quietly capture data. Always check links in the assistant's chat, never in your browser.
- Asking the assistant to "visit the link and tell me what's there". Some assistants can browse, but pasting an unknown URL into a tool you don't fully understand adds risk. Ask for an analysis of the URL itself (the words and structure) instead.
- Trusting the message because the writing sounds perfect. Perfect grammar used to be a sign of a real company. With AI, polished writing is now a common sign of a scam. Polished ≠ safe.
- Ignoring pressure words. "Urgent", "today only", "act now", "or your account will be closed" are some of the strongest scam signals. If you see them, slow down — that is exactly what the scammer doesn't want you to do.
- Replying to ask "is this real?". Replying tells scammers your number is active. If you genuinely want to verify, contact the organisation through a channel you found yourself, not through the message.
Find one message in your phone or inbox that you've been unsure about (not from the last hour — something older is fine). Paste it into your usual AI assistant with the prompt from Step 3, and read the first three red flags it lists. That's your first five-minute check.
❓ Quick questions
How long does this take?
About 7 minutes — the guide has 5 steps, and you can tick each one off as you go.
Do I need to prepare anything?
- An account with at least one AI assistant you can chat with — ChatGPT, Gemini, Claude, Microsoft Copilot, Le Chat, DeepSeek, or similar. The free tier is enough for this.
- A phone or computer where you can copy the suspicious message and paste it into the chat.
- The original message itself (screenshot or text), plus the sender's number, email or username if you have it.
- Roughly 5 minutes for your first check; faster once you've done it two or three times.
A quick honest note: an AI assistant can flag likely warning signs. It cannot guarantee a message is safe, and it cannot detect every new trick. When in doubt, treat the message as suspicious and contact the organisation directly using a phone number from their official website.
What mistakes should I avoid?
- Tapping a link "just to look". A scam site can install harmful software the moment it opens, or quietly capture data. Always check links in the assistant's chat, never in your browser.
- Asking the assistant to "visit the link and tell me what's there". Some assistants can browse, but pasting an unknown URL into a tool you don't fully understand adds risk. Ask for an analysis of the URL itself (the words and structure) instead.
- Trusting the message because the writing sounds perfect. Perfect grammar used to be a sign of a real company. With AI, polished writing is now a common sign of a scam. Polished ≠ safe.
- Ignoring pressure words. "Urgent", "today only", "act now", "or your account will be closed" are some of the strongest scam signals. If you see them, slow down — that is exactly what the scammer doesn't want you to do.
- Replying to ask "is this real?". Replying tells scammers your number is active. If you genuinely want to verify, contact the organisation through a channel you found yourself, not through the message.
Keep reading
✦ Original step-by-step guide by AI World HQ's AI editorial team. Written in plain language, reviewed for accuracy.
← Back to all stories