Would you notice anything wrong if an email said your ChatGPT Plus payment didn't go through? On a busy morning, most of us wouldn't. You pay for a handful of subscriptions and you barely think about them; that's the whole point of a subscription. So when the email says update your card, the reflex is to fix it quickly, before anything gets cut off. That reflex is exactly what the scam is built around.

Why is a billing notice a better lure than a scary warning?

Most phishing tries to frighten you: your account was hacked, your parcel is held, your bank has locked you out. A failed-payment email doesn't need any of that. It's mundane. Cards expire. Banks decline things for no reason. A billing hiccup is boring and completely believable, which is precisely why it gets clicked.

It also asks for the one thing the attacker actually wants, and asks for it in a way that feels perfectly normal. Nobody is surprised that a payment-update page wants a card number. That's what a payment-update page is for.

Security firm Check Point flagged this exact campaign in its Q2 2026 brand phishing report, published on July 23, 2026. Their description is short and worth quoting in full: "One example from June involved a fake ChatGPT Plus billing email, built to look exactly like an OpenAI payment failure notice, that led to a page designed to harvest full credit card details."

Note the phrase "built to look exactly like." There is no bad logo to spot here.

And it isn't a handful of emails. In a September 10, 2026 post, Microsoft's security team described a ChatGPT-themed campaign that "sent up to 100,000 emails in a single day," tricking people into updating their ChatGPT Plus payment information and stealing personal and credit card data.

Why AI subscriptions now? Because millions of people started paying for one in the space of a couple of years, and most of them have never seen what a real billing email from that company looks like. Check Point's report has ChatGPT/OpenAI entering the ten most impersonated brands for the first time this quarter, corroborated by Infosecurity Magazine and Help Net Security. It's still far from the top: Microsoft remains the favorite at 23% of attempts, per Check Point, and the top five brands together account for more than half of everything. The news isn't that ChatGPT is the biggest target. It's that it's now on the menu at all.

What did one of these emails actually look like?

Check Point described the campaign but didn't publish the addresses involved. An Australian email security company, MailGuard, intercepted and documented a version of it on July 10, 2026, and their write-up is the concrete one. Here's what they observed:

And one nasty detail worth remembering: MailGuard notes the page threw fake error messages and retry prompts, so a victim would type their card details in more than once. If a payment page ever tells you "something went wrong, please try again," treat that as a reason to stop, not a reason to retype. 😅

Senderimi2001.co.jpdisplay name said "Chat GPT"; the domain says otherwise
Payment pageargentina.alwaysdata.neta free hosting subdomain wearing Stripe's design

This isn't even a clever lookalike. Nobody swapped a letter or registered openai-billing.com. It's an unrelated domain that presumably got compromised or was simply bought cheap, which tells you how little effort the design of the email required to work. The people who clicked weren't fooled by the domain. They never looked at it.

The lure isn't new, either. Someone posted a near-identical complaint to OpenAI's own community forum back in May 2025: a "your payment method stopped working" email, sent from a random address with no connection to OpenAI. Different year, same script.

Which check actually works? Read the domain after the @

Every email has two sender fields, and they are nothing alike:

  1. The display name. "Chat GPT," "OpenAI Billing," "Support." Free text. The sender types whatever they want. It carries no information at all.
  2. The actual address. The part after the @. This is a real domain that someone had to register and control. It's the one part of the email the design team can't fake.

On most mail apps the address is hidden behind the display name until you tap or hover on the sender. Do that. Then read the domain right to left: the real owner is the last two chunks before the first slash. billing.openai.com is owned by openai.com. openai.com-billing-update.info is owned by com-billing-update.info, and the brand name at the front is bait. (We went through that reading trick in detail in read the domain right to left.)

Then there's the check that costs nothing and beats all of this: don't use the link. If you genuinely might have a billing problem, open a new tab, type the service's address yourself, and look at your billing page there. A real failed payment will be waiting for you inside your account. A fake one won't exist.

"Not flagged" is not the same as "safe." A domain nobody has reported yet is just a domain nobody has reported yet, and a phishing page put up this morning is exactly the kind nothing has caught up with. Every check is a signal for your judgment, not a verdict.

Where does IP Tracker fit?

IP Tracker is a free Chrome extension. Paste in a full email address, the whole thing, something@imi2001.co.jp, and it pulls out the domain and checks that. The free tier is 50 checks a day, no account, no tracking. Only the value you paste is looked up.

For a sender domain, it does three things at once. It compares the domain against the official domains of about 125 widely impersonated brands, normalizing look-alike characters first (the number 1 and capital I become a lowercase l, zero becomes the letter O, common Cyrillic look-alikes map back to Latin), so a letter-swap fake gets named even when your eye slid past it. It checks whether Google Safe Browsing (Google's list of reported dangerous sites) has flagged it, and how many security vendors flag it on VirusTotal. And it shows the domain's registration date, because a domain created eleven days ago that's emailing you about your subscription is a signature worth seeing.

What you get back is plain English: flagged as a lookalike, on a blocklist, or nothing found, plus who owns the domain and when it was registered, so you can decide.

What can't it do?

Honesty matters more than comfort here, so four real limits:

The design of the email is the part the attacker controls completely, so it's the part that will always look perfect. The domain it came from is the part they have to actually own, which is why that's the part worth two seconds of your attention.

To summarize:

Nothing about a subscription is ever that urgent. Let the panic wait five minutes. 😉

Sources