Copy the suspicious message, paste it into an AI chatbot, and ask it to list the red flags and what the sender wants you to do. AI models are good at spotting the patterns scammers reuse: urgency, an unexpected payment problem, a link to a lookalike website, a request for a code. Treat the answer as a second opinion, then confirm by contacting the company through its official app or a phone number you look up yourself.
If the AI runs on your phone, the message and any personal details the scammer included in it never go to another company’s server.
How common are scam texts? #
Very common. Consumers told the Federal Trade Commission they lost more than $12.5 billion to fraud in 2024, up 25% from the year before. Email was the most reported way scammers made first contact, followed by phone calls and then text messages. The share of people who reported fraud and actually lost money rose from 27% in 2023 to 38% in 2024.
What are the signs a text message is a scam? #
The FTC’s guide to spam text messages lists the classics. Scam texts often:
- Promise free prizes, gift cards or coupons.
- Offer a low or no-interest credit card.
- Claim they can help with student loan debt.
- Say there’s suspicious activity on your account.
- Report a problem with a payment, or send a fake invoice.
- Announce a delivery problem with a package you weren’t expecting.
Look at the form as well as the content. A bank “alert” from a 10-digit personal number or an email address, a link to a domain that’s almost the company’s name, a deadline measured in hours, and a request to move the conversation to WhatsApp or Telegram are all warning signs.
| Scam type | What the text says | The tell | What to do |
|---|---|---|---|
| Package delivery | “Your parcel is on hold, update your address” | You weren’t expecting anything; the link isn’t the carrier’s site | Track the package in the carrier’s own app |
| Bank alert | “Unusual sign-in, verify now” | A link or callback number inside the text | Open your bank’s app or call the number on your card |
| Unpaid toll | “Pay your outstanding toll to avoid fees” | Small amount, urgent deadline, odd web address | Check your account on the toll agency’s official site |
| Job offer | “Earn $300 a day, reply to learn more” | You never applied; the pay is too good | Ignore and block |
| Wrong number | “Hi, is this Emma? Nice to meet you anyway” | A friendly chat that drifts toward investing | Don’t engage |
| Verification code | “Send me the code I just texted you” | Anyone asking for a code | Never share codes |
Does your phone already check for scam texts? #
Partly. In March 2025 Google added Scam Detection to Google Messages. It flags conversational scams in SMS, MMS and RCS messages, with all processing on the phone. It launched in English in the U.S., U.K. and Canada and is on by default for messages from non-contacts. The Phone by Google app got a similar feature for calls on Pixel 9 and later phones, using the on-device Gemini Nano model; it launched for English speakers in the U.S. and is off by default.
iPhones can filter texts from unknown senders into a separate list and let you report junk from the conversation. Carriers also block a lot of spam before it reaches you.
These filters catch many scams, but not all. The ones that get through tend to be the better-written ones, which is exactly when a second opinion helps.
How to check a suspicious text with AI, step by step #
- Don’t tap links or reply. Replying, even with STOP, tells the sender your number works.
- Copy the text or take a screenshot. A screenshot keeps the sender’s number and any link preview.
- Paste it into an AI chatbot. If it includes your name, address, account digits or a tracking number, use a private model (more on that below).
- Ask a specific question. A prompt that works well:
Here is a text message I received. I did not expect it. List every sign that it might be a scam, explain what the sender wants me to do, and tell me how I could check whether it’s real without using any link or phone number in the message.
- Read the reasoning, not just the verdict. If the AI points out that a link goes to a lookalike address (say, a made-up “usps-redelivery-help” site instead of usps.com), check that detail yourself.
- Verify through a channel you choose. Open the company’s official app, type its web address yourself, or call the number printed on your card or statement.
- Report and delete. Forward the text to 7726 (SPAM) so your carrier can act on it, report it at ReportFraud.ftc.gov, then block the sender.
Why use an offline AI to check scam messages? #
Scam texts are often personalized. They include your first name, part of a card number, a real tracking number or your street, and a screenshot can show contact names and earlier messages too. Pasting all that into a cloud chatbot hands it to one more company.
Personal LLM runs the model on your phone, so the message stays there. Every model in its catalog can read images once you add the image-support file (195 MB to 1 GB), so you can attach the screenshot instead of retyping it; our guide to asking AI about photos offline walks through that. Turn on thinking mode and the model shows its step-by-step reasoning in a collapsible panel above the answer, which makes it easier to spot where it might have gone wrong. It also works in airplane mode, so you can check a message without being online at all.
What AI can and can’t tell you about a scam #
| AI can | AI can’t |
|---|---|
| Spot urgency, threats and pressure tactics | Tell whether a link is live or malicious right now |
| Notice lookalike domains and odd sender formats | Look up who registered a domain or owns a number |
| Explain what a scam script is trying to get you to do | Know about scam campaigns that began after its training data ends |
| Suggest safe ways to verify | Guarantee that a message is legitimate |
An offline model has a training cutoff and no internet connection, so it doesn’t know about this week’s scam wave or whether your bank really sent an alert this morning. We explain that limit in does offline AI know current events. Cloud chatbots can sometimes search the web, but they’re still guessing about your specific message.
AI makes both kinds of mistakes. It can call a genuine fraud alert from your bank suspicious, and it can miss a carefully written scam. A “looks legitimate” verdict is never a reason to click a link.
What if you already clicked or replied? #
Act quickly:
- If you entered a password, change it on the real site and anywhere else you used it.
- If you shared card or bank details, call your bank using the number on your card and ask it to block the card or watch the account.
- If you gave someone a verification code, assume they tried to get into that account. Sign in, then check recent activity and connected devices.
- If the message got you to install an app or profile, remove it.
- If your identity may have been stolen, go to IdentityTheft.gov for a recovery plan.
Frequently asked questions #
Can AI detect scam text messages? #
It can spot most common patterns and explain its reasoning, which makes it a useful second opinion. It can’t check links or phone numbers in real time, and it can be wrong in both directions. Always confirm through a channel you find yourself.
Is it safe to paste a scam text into ChatGPT? #
The message itself is harmless to paste; the risk is the personal information in it. If it contains your name, address or account details, remove them first or use an on-device model. For more on what cloud chatbots keep, see is it safe to paste emails into ChatGPT.
Should I reply STOP to a spam text? #
Only to messages from companies you actually signed up with. For unknown senders, replying tells them your number is active. Block the sender, forward the text to 7726 and delete it.
Does Google Messages read my texts to find scams? #
Google says its Scam Detection feature analyzes messages on the phone itself and doesn’t send them to Google for that check. When it spots a likely scam, it warns you in the conversation.
Can I use AI to check a suspicious email too? #
Yes, the same prompt works for email. Include the sender’s full email address, which scam messages often hide behind a familiar display name, and never open a link just to see where it goes.