Fake ChatGPT, Claude, Gemini Login Pages: Best How to Spot
Spot fake ChatGPT, Claude, and Gemini login pages before they steal your password and MFA code. A 20-minute, step-by-step defence guide.
# Beware Fake AI Login Pages: How to Spot [ChatGPT](https://chat.openai.com/), Claude, and Gemini Phishing
A reader in Manchester emails us a screenshot: a "[ChatGPT Plus](/blog/chatgpt-plus-for-less-official-vs-shared-plans) annual deal, 70% off" ad, a login box, and a prompt for her six-digit authenticator code. She typed it in. Ninety seconds later, her account was posting crypto spam to 400 contacts. The whole attack took her under two minutes to fall for, and it will take you about 20 minutes to inoculate yourself against it using the steps below. That is the best how to do trade we know of: 20 minutes now versus a credential dump later.
The pattern is consistent across fake ChatGPT login pages, Claude phishing kits, and Gemini ad portals. Attackers buy search and social ads, clone the real sign-in screen pixel for pixel, and harvest both your password and your live MFA code. The MFA relay is the nasty part: the fake page passes your code to the real service in real time, so your authenticator app never flags anything wrong. This guide walks you through spotting the fakes, hardening your accounts, and knowing what to do in the first ten minutes after a mistake.
## What you'll need
- A [password manager](/blog/ai-security-crisis-how-hackers-breached-openai-and-what-you-can-do) with autofill (Bitwarden free tier is fine, 1Password or Dashlane if you want polish). Budget: $0 to $60 per year.
- An authenticator app or hardware key. Aegis or Google Authenticator are free; a YubiKey Security Key NFC runs about $29.
- Access to your ChatGPT, Claude, and Gemini account settings (15 minutes).
- A browser with a visible address bar and bookmark bar enabled (Chrome, Firefox, Safari, Edge all work).
- About 20 minutes total, split across the steps. No paid tools required.
## Step-by-step
### Step 1. Bookmark the real sign-in pages today
Goal: eliminate the need to ever search for an AI login page again.
Open each service directly by typing the address, then bookmark the login screen. The canonical sign-in URLs are chatgpt.com/auth/login, claude.ai/login, and gemini.google.com. Drag each into your bookmark bar and rename them "ChatGPT login," "Claude login," "Gemini login."
From now on, reach the login page only through the bookmark. This single habit defeats the majority of ad-based phishing, because the fake portal depends on you clicking a sponsored result.
**Tip:** Put the bookmarks in a dedicated folder called "Logins" and hide the bar if you like a clean screen. The muscle memory matters more than the visibility.
### Step 2. Learn the three-second URL check
Goal: train your eye to catch the domain swap before you type anything.
Real domains end in exactly these strings: chatgpt.com, openai.com, claude.ai, anthropic.com, gemini.google.com, and google.com. Fake portals love lookalikes such as chatgpt-plus[.]com, claude-ai[.]app, and gemini-ai[.]login. They also use subdomain tricks: chatgpt.com.secure-login[.]xyz is not OpenAI, no matter what comes first.
Read the domain from right to left, starting at the dot before the final slash. That is the only part that matters.
**Watch out:** Padlock icons mean nothing. Phishing sites have valid TLS certificates too. The lock proves the connection is encrypted, not that the site is honest.
### Step 3. Kill the ad-click reflex
Goal: stop feeding the ad portals that harvest credentials.
Search results and social feeds are now the main delivery channel for AI phishing. The sponsored result often sits above the genuine one, and the copy is tailored: "[ChatGPT Plus](/blog/how-to-cancel-your-chatgpt-subscription-and-when-it-s-worth-it) at $6.99/month," "[Claude Pro](https://claude.ai/) student discount," "Gemini Advanced free for 3 months." If a deal sounds better than the official pricing page, assume it is bait until proven otherwise.
Check official pricing before you click anything. As of late 2026, ChatGPT Plus is $20 per month, Claude Pro is $20 per month, and Google's Gemini Advanced tier sits inside the Google One AI Premium plan at $19.99 per month. Anything materially cheaper is a red flag, not a bargain.
**Tip:** Install an ad blocker such as uBlock Origin. It removes most sponsored results, which removes most of the attack surface.
### Step 4. Turn on phishing-resistant MFA
Goal: make a stolen password useless even if you slip.
Go to your [account security](/blog/best-two-factor-authentication-apps-and-methods-compared) settings and prefer a passkey or hardware security key over SMS or app codes. In ChatGPT, open Settings, then Security, and add a passkey. In Claude, open Settings, then Security, and enable two-factor authentication with an authenticator app, then add a passkey if offered. For Google, visit myaccount.google.com/security and enroll a passkey or a security key under "How you sign in to Google."
Passkeys are bound to the real domain, so a fake login page cannot use them. That is the difference between a bad afternoon and a breached account.
**Watch out:** If a site asks for your MFA code and then asks you to "confirm" by entering a second code that appears on your phone, stop. That is a real-time relay attack. No legitimate login does this.
### Step 5. Audit your AI account sessions and connected apps
Goal: evict anything you do not recognize before an attacker uses it.
In ChatGPT, click your profile icon, then Settings, then Data controls, and review connected apps and active sessions. In Claude, open Settings, then Connected apps, and revoke anything unfamiliar. In Google, go to myaccount.google.com/permissions and remove third-party apps you do not use.
Sign out of all sessions, then sign back in on your own devices. It takes five minutes and clears any session an attacker may already hold.
### Step 6. Set up breach alerts
Goal: get told when your credentials surface, not when the damage lands.
Enroll your email in Have I Been Pwned's notification service (free) and turn on Google's dark web report if you use a Google account. In the US, UK, and EU, your bank and major email providers also offer free breach monitoring. Enable it.
**Tip:** Use a unique email alias for AI services so a breach there is easy to trace back to the source.
### Step 7. Practise the ten-minute recovery drill
Goal: know exactly what to do if you ever do type credentials into a fake page.
The moment you suspect a fake login, change the password on the real site from a device you trust, revoke all sessions, and rotate your MFA secret. Then check sent mail, payment methods, and API keys. If you entered a card number, call your bank and ask for a replacement card. In the EU and UK, you have strong chargeback rights; in the US, report fraud promptly to preserve them.
Report the fake page to the real provider (OpenAI, Anthropic, and Google all run phishing report channels) and to the ad platform. It takes two minutes and protects the next person.
### Step 8. Teach the household or team the bookmark rule
Goal: extend the defence to the people most likely to click.
Phishing works because it exploits urgency and habit, not stupidity. Show your family or team the bookmark trick and the domain check. Two minutes of explanation covers the people who will otherwise forward you a "great AI deal" next week.
## Common problems and fixes
- **The bookmark opens a warning page or a different domain.** Your bookmark was edited or your browser was hijacked by an extension. Remove recent extensions, then re-create the bookmark by typing the address manually.
- **Your password manager will not autofill on a login page.** Good. That is the manager protecting you. Autofill is domain-locked, so a missing prompt on a familiar-looking page usually means the domain is wrong. Stop and check the URL.
- **You already entered an MFA code into a fake page.** Assume the code is burned. Change the password, revoke sessions, and re-enroll MFA immediately. If you use SMS codes, move to an authenticator app or passkey today.
- **The fake page appears in search results above the real one.** Report it to the search engine using the "report inappropriate prediction" or ad-report link, then switch to the bookmark method permanently.
- **You cannot find a passkey option in a service.** Not every tier supports passkeys yet. Use an authenticator app with number matching, and never share a code with anyone who contacts you first, even if they claim to be support.
## Our take
Rank the defences by return on effort. First, bookmarks: free, instant, and they stop the most common attack. Second, a password manager with domain-locked autofill: Bitwarden at $0 is enough, 1Password at about $36 per year is the value pick if you want the smoothest experience. Third, a hardware key: a $29 YubiKey is the one to buy if you hold anything valuable in these accounts. Skip paid "[AI security](/blog/chatgpt-privacy-how-to-opt-out-of-ai-training-and-protect-your-data)" browser add-ons that promise phishing detection; the free combination above beats them, and the paid ones add an attack surface of their own. If you only do one thing, do Step 1.
## What to do next
Once your bookmarks, MFA, and breach alerts are in place, run the same audit on your other high-value accounts: email, banking, and cloud storage. Then set a calendar reminder for a quarterly session review, because connected apps accumulate quietly. Finally, forward this guide to the person in your life who forwards you deals.
## FAQ
**Can a fake AI login page steal my MFA code if I use an authenticator app?**
Yes, if the page relays the code to the real service in real time. Passkeys and hardware keys close that gap because they are bound to the real domain.
**Is a padlock icon enough to trust a login page?**
No. Certificates are free and phishing sites use them. Check the domain itself, reading right to left from the final dot.
**What should I do first if I typed my password into a fake ChatGPT page?**
Change the password on the real site, revoke all sessions, and re-enroll MFA. Do it from a device you trust, then report the fake page to OpenAI and the ad platform.
Frequently asked questions
Can a fake AI login page steal my MFA code if I use an authenticator app?
Yes, if the page relays the code to the real service in real time. Passkeys and hardware keys close that gap because they are bound to the real domain.
Is a padlock icon enough to trust a login page?
No. Certificates are free and phishing sites use them. Check the domain itself, reading right to left from the final dot.
What should I do first if I typed my password into a fake ChatGPT page?
Change the password on the real site, revoke all sessions, and re-enroll MFA. Do it from a device you trust, then report the fake page to OpenAI and the ad platform.
Tech-N-AI Talks
