Can ChatGPT tell you if a website is a scam?
%201%20(1).png)
Key Takeaways
TL;DR: ChatGPT can help you spot the warning signs in a suspicious message, but on its own it doesn't check a link against threat data. A brand-new scam site may have no history for it to find, and it can sound sure while being wrong. To check a link without leaving the chat, add the Guardio app in ChatGPT, which checks links against Guardio's known threats and shows the result in your conversation.
You get a text about a package on hold, or you find a deal that looks almost too good. So you paste the link into ChatGPT and ask: is this a scam? It's a sensible instinct, and ChatGPT is a good second opinion on a suspicious message. It isn't a link checker, though, and that difference matters when your card details are one click away.
This guide covers what ChatGPT can and can't tell you about a website, and how to check a link with the Guardio app without leaving the chat.
What ChatGPT is good at when you suspect a scam
ChatGPT is useful for reading a message the way a scam-aware friend would. Paste the text or describe what you got, and ask what looks off. It can point out:
- Pressure and urgency: "Your account will be closed today" or "Your package will be returned unless you pay now."
- Odd ways to pay: requests for gift cards, wire transfers, or cryptocurrency.
- Sender details that don't match: a message that claims to be from one company but comes from an unrelated email address or phone number.
- Deals that are too good: a big discount on something that's usually expensive and rarely on sale.
- Generic greetings: "Dear customer" instead of your name or order details.
- Links that don't match the button: a "Track your order" button that points to an unfamiliar address.
It can also explain what kind of scam you might be looking at, what the scammer is after, and sensible next steps, such as contacting the company through its official app or website instead of the link.
Tip: Paste the message, not your secrets. Never share passwords, card numbers, or one-time codes in any AI chat.
Where ChatGPT on its own falls short
ChatGPT is a language tool, not a security scanner. Even when it searches the web or looks a page up, it isn't checking the link against a list of known threats. That leaves a few gaps.
- Brand-new sites have no history. Scam sites appear on new addresses all the time. If a site went up this week, there may be no reviews, reports, or search results for ChatGPT to find.
- Copies can look like the real brand. Fake stores borrow real logos, product photos, and layouts. An address that's one letter off, or ends in an unusual domain, is easy to miss, and a search may show the real brand instead of the copy.
- Sites change and move. A scam site that gets taken down can come back under a new address, so an older answer may not fit the link in front of you.
- AI can sound sure and be wrong. ChatGPT aims to give a fluent, helpful answer. That can read as reassurance even when there's no evidence behind it.
- The padlock isn't proof. A padlock means the connection is encrypted, not that the site is honest. Scam sites use them too. Our guide to the HTTPS padlock myth explains why.
So use ChatGPT for the message, and use a real check for the link.
How to check a link with the Guardio app in ChatGPT
The Guardio app connects ChatGPT to Guardio's link checks. You ask about a link in your chat, Guardio checks it against its known threats, and the answer appears in the conversation. In ChatGPT's app directory, it is listed as Guardio.
Before you start
- You need a Guardio account. You sign in once and approve access.
- If you are asked to connect again later, repeat the same steps.
Step 1: Add the Guardio app to ChatGPT
Go to guard.io/mcp, choose ChatGPT under Platforms, and follow the steps shown. Or open ChatGPT, search its app directory for Guardio, and add it.
Step 2: Connect your Guardio account
Click Connect, sign in to Guardio, and approve access.
Step 3: Ask Guardio to check the link or message
Mention Guardio in your message, then ask in your own words. For example:
- "Guardio, check this link before I open it: [link]"
- "Guardio, is this link safe to open? [link]"
- "Guardio, does this text look like a scam? [message]"
Got a suspicious delivery text or deal email? You can paste the whole message. Guardio checks the links in it without the message body being sent to Guardio.
Step 4: Read the result and act on it
- If Guardio flags the link as dangerous, don't open it. Delete the message, and if it claims to be from a company you use, contact them through their official app or website.
- If no known threats are found, that's a good sign, not a guarantee. A link result reports known threats at the time of the check, and it doesn't vouch for the whole message. If the message still pushes you to pay fast or share details, trust that feeling and check with the company directly.
For more ways to check a link before you click, see our guide on how to check if a link is safe.
Want to set it up now? Head to Guardio for AI assistants and connect Guardio to ChatGPT.
What else the Guardio app can do in ChatGPT
- Check for data leaks: ask whether your email address or phone number appears in known data breaches. Results describe known exposure and don't reveal leaked passwords. If your email shows up, change the passwords for accounts that use it.
- Review your protection: see your Guardio protection setup, scan activity, and security recommendations, and ask what to fix next.
- Set up browser protection: ask Guardio to guide you through setting up its protection in a supported browser.
A few things it doesn't do: the app doesn't scan files on your device, and connecting it doesn't give you continuous browser protection. Account features depend on your access and plan.
What Guardio can and cannot see
It's fair to ask before you connect anything to your AI assistant. Here's how Guardio describes it:
- It sees only your own Guardio account. Family members' data stays with them.
- It can look things up, not change them: no settings, no purchases.
- Guardio never sends your assistant the pages you visit or your messages. It only checks the links and texts you ask about.
- To disconnect, remove Guardio from your assistant's app or connector settings.
Using Guardio with Claude and other AI assistants
The same checks work beyond ChatGPT. Guardio for AI assistants connects through MCP, short for Model Context Protocol, an open standard that lets AI assistants use outside tools. Guardio's setup page lists Claude, Claude Code, Cursor, and Grok Bot, with more on the way. The server address is https://mcp.guard.io/mcp, and the setup steps for each assistant are on guard.io/mcp.
This matters more as assistants start shopping for people. If you ask an assistant to find the best price on something, you can ask it to check the store's link with Guardio before you buy. Before you let any assistant pay for you, read our guide on whether it's safe to let an AI shopping agent buy for you.
On-demand checks vs. always-on protection
The Guardio app in ChatGPT is for the moments you stop and ask. That's useful, but scam links don't wait for you to feel suspicious. Guardio's browser extension for Chrome and Edge, along with its iOS and Android apps, helps flag phishing sites, phishing emails, and scam texts on iOS in the background across your devices. Use the ChatGPT app for a quick check, and let Guardio's protection cover the moments you don't think to ask.
Troubleshooting the Guardio app
- Can't find Guardio in ChatGPT: go to guard.io/mcp and choose ChatGPT under Platforms.
- The wrong Guardio account is connected: remove Guardio from ChatGPT, sign out at guard.io, then connect again and sign in with the right account.
- ChatGPT says it has no Guardio tools: remove Guardio and connect again, then start a new chat so ChatGPT reloads its tools.
- It asks you to connect again: connect again and you're set.
The bottom line
ChatGPT is a helpful second opinion on a suspicious message, but it isn't a link checker. When a link matters, let ChatGPT call Guardio and check it before you open it. If you're sorting through delivery texts and deal emails this season, our guide to package delivery text scams shows what to look for, and popular AI scams explained covers the newer tricks.
How well do you spot scam links and texts?
Take Guardio's free scam quiz to see how well you spot the scam texts and links that catch people out.
FAQs
Can ChatGPT tell if a link is safe?
Not reliably on its own. ChatGPT can flag warning signs in the message around a link, and it can search the web, but it doesn't check the link against threat data. A new scam site may have no history to find. The Guardio app in ChatGPT adds that check and returns the result in the same chat.
Is a "no known threats" result a guarantee?
No. It means Guardio found no known threats at the time of the check. It isn't a guarantee that a site or a whole message is safe, so if the message still pressures you to pay or share details, check with the company directly.
Does Guardio see my messages or ChatGPT chats?
No. Guardio says it never sends your assistant the pages you visit or your messages, and that it only checks the links and texts you ask about. When you paste a message, Guardio checks the links in it without the message body being sent to Guardio.
Do I need a Guardio account to use the app in ChatGPT?
Yes. You sign in to your Guardio account once and approve access.
Can I use Guardio with Claude or other AI assistants?
Yes. Guardio's setup page lists Claude, Claude Code, Cursor, and Grok Bot, with more on the way. The steps for each one are on guard.io/mcp.
What should I do if I already clicked a scam link?
Don't enter anything on the page, and close it. If you typed a password, change it, along with any other account that uses the same one, and turn on two-factor sign-in. If you shared card details, call your bank or card issuer. You can report the scam to the FTC at ReportFraud.ftc.gov.
Safe BrowsingWhat Are Clickbait Ads? Guide to Deceptive Marketing Tactics





