8 min read
Leaked customer logins at 40 big websites: 1,380 a month
We searched stealer logs and combolists for 40 retail, travel and food delivery sites. 89,262 of their customers' logins were posted in the last 30 days.
By BreachDetect Research
On this page
In the last 30 days, 89,262 customer logins for 40 well-known shopping, travel and food delivery websites were posted to the Telegram channels and bots we collect from. The median site had 1,380. The busiest had 9,158, about 300 a day.
On 15 November 2026 we searched our index for 40 consumer websites: 14 online retailers, 13 airlines, hotel groups and booking sites, and 13 food delivery and restaurant brands. For our study of 99 companies we searched by email domain, which finds employees. This time we searched by website. That finds every leaked login someone saved for the site, whoever it belongs to, and for a consumer brand that mostly means customers.
We're not naming the sites, apart from two food delivery brands further down whose results need explaining. They're household names with large security teams, and nothing here means any of them was hacked. These logins came off customers' own computers and out of other companies' breaches. That's the point. This is what the problem looks like for a big consumer brand even when its own systems are fine.
The numbers
- 386,270 leaked login records for customers of the 40 sites, belonging to 267,919 distinct logins. 39 of the 40 hit our limit of 10,000 results, so these are the newest records, not all of them.
- 89,262 were posted in the last 30 days and 46,782 in the last seven. Those two counts are complete for every site, for reasons explained at the end.
- 27% came straight from stealer logs, copied out of an infected browser. Every other record was a combolist line that also named the website, the format stealer logs get trimmed down to before they're resold.
- 10% were saved on a sign-up page and 4% at checkout.
- 63% of the logins that were email addresses were Gmail. A search for leaked @yourcompany.com addresses can't see any of them.
- 34 of the 40 sites also had their own staff's logins mixed in with the customers', 5,120 in all.
10,000 records only reached back to the summer
Our search returns the newest records first and stops at 10,000. For a small business that's usually years of history. For these sites it was a few months.
For 36 of the 40, the newest 10,000 customer records were all posted after mid-July 2026. For one retailer and one travel site, the newest 10,000 covered less than six weeks.
The postings come in waves rather than a steady trickle. August was the busiest month in the sample, even though for most sites we only saw the end of it. The first two weeks of November brought more records than the whole of October. A quiet month tells you very little about the next one.
By sector
| Retail | Travel | Food and restaurants | |
|---|---|---|---|
| Sites | 14 | 13 | 13 |
| Median logins posted in the last 30 days | 2,320 | 1,792 | 521 |
| Lowest site | 430 | 635 | 2 |
| Highest site | 9,158 | 8,862 | 2,822 |
| Records from stealer logs | 35% | 26% | 19% |
| Logins that weren't email addresses | 12% | 42% | 17% |
| Staff logins found | 2,263 | 1,315 | 1,542 |
Retail sites had the most, and the largest share from stealer logs. Four retailers had more than 6,500 in 30 days.
The food figures come with two caveats. Uber Eats customers sign in through Uber's own sign-in page on uber.com, so most of their leaked logins are filed under Uber, not under ubereats.com. And one brand had almost nothing, which is the next section.
The site without passwords had almost nothing to leak
Wolt, the food delivery company, had two leaked customer logins posted in the last 30 days. Leaving out Uber Eats, the next lowest site had 219, and the median food site had 521.
Wolt doesn't use passwords for customers. You type your email address or phone number and it sends you a link or a code. There's no password for a browser to save, so an infostealer has very little to copy and a combolist has very little to carry. We did find some Wolt records, 869 in the past year, but at a rate closer to a handful a week than the hundreds a day the big retailers get.
This is the clearest result in the study, and it isn't really about Wolt. Every other site in the sample still lets customers sign in with a password, and every one of them had hundreds or thousands of new leaked logins this month. Passkeys and one-time codes don't stop malware infecting a customer's laptop. They do mean that when it happens, there's no reusable password to steal. The malware can still take the session cookie, but that stops working when the session ends. A password keeps working until someone changes it.
Airline and hotel logins are membership numbers
42% of the travel logins weren't email addresses. At three airlines and one hotel group, more than 80% were a username or a membership number. At one airline it was 96%.
That makes sense, since loyalty programmes let you sign in with your member number. It also means these logins are invisible to every breach-checking service that works by email address. And a frequent flyer account is close to a wallet. Points can be spent on flights, upgrades and gift cards, transferred to other people or sold. Of all the accounts in this study, these are the ones an attacker can turn into money fastest.
Saved at sign-up and at checkout
A browser records the page a password was saved on. 39,225 of the customer records were saved on a sign-up page, which means the customer's browser stored the password at the moment they created the account. 16,587 were saved at checkout, often a guest turning into a member halfway through an order. 35 of the 40 sites had logins saved on their checkout pages.
Neither tells you anything bad about the site. What they show is that the customer's computer is infected, or was when the log was taken. Whatever password that customer picks next, on that same computer, will leak the same way. Resetting the password helps. Telling the customer their device needs cleaning helps more. We go into the details in how to check if your customers' passwords have leaked.
Three bots carried most of it
The records came from Telegram channels and bots, and the median site turned up in 20 different sources. But three Telegram bots carried 69% of all the records between them. One is named after ALIEN TXTBASE, the set of combolists Troy Hunt loaded into Have I Been Pwned in February 2025, which ran to 23 billion rows. A second also uses the Alien name.
This matches what we saw for employee logins. You don't have to watch hundreds of channels to catch most of what gets posted for a website. A handful carry most of it.
Some of these were staff
34 of the 40 sites had staff logins in the results. Some were the company's own email addresses used on its own site. Others were saved on staff pages, like the brand's single sign-on, webmail or supplier portal. One retailer had 1,208.
Those don't belong in the same queue as customer accounts. A leaked customer login risks one account. A leaked staff login for single sign-on risks the company. We keep them apart for that reason, and wrote up what to do in the first 24 hours when one turns up.
What we'd do with this
- If you run a site with a customer sign-in, assume new leaked logins for it arrive every day. A check you ran once in the spring says almost nothing now.
- Search by your website, not only by your email domain. That's the only way to find customers, and it often finds contractors and staff using personal addresses too.
- Check each leaked password against the account it belongs to. Your sign-in code already compares passwords with stored hashes, so you can run the same check without logging in to anything. Reset the ones that still match.
- Move customers off passwords where you can. The Wolt result is what that looks like in the data.
- Expect to be judged on it. Regulators have fined companies for not doing these things, and we've collected what credential stuffing has cost companies so far.
You can see what's out there for your own site with our free account takeover check. It shows how many logins have leaked, which pages they were saved on and when they were posted, with every customer's identity hidden.
How we did this
- We searched 40 websites on 15 November 2026 using the BreachDetect index of stealer logs and combolists. Each search covered the site's domain and its subdomains.
- We picked well-known names in each sector across the US, the UK, Europe and the Middle East. It's not a random sample, so treat the figures as examples rather than an average.
- A record is one leaked login, a username or email address and a password, for one website. The same login can appear many times when channels repost each other, which is why there are fewer distinct logins than records.
- We set aside logins that belonged to each site's staff: logins using the company's own email domain, and logins saved on staff pages like single sign-on or webmail. Everything else we counted as a customer.
- Searches stop at 10,000 records and return the newest first. For every site that hit the limit, the 10,000th record was posted more than 30 days ago, so the counts for the last 7 and 30 days are complete. The totals and percentages describe the newest 10,000 for each site, not its whole history.
- Posted means the date a record was posted to the source we collected it from, not the date the password was stolen.
- We only counted. We didn't request any passwords for this study, we didn't keep any email addresses or usernames, and we didn't try any of the logins. We don't know how many still work, and only each site can find out, by checking them against its own accounts.
