The singer Raye says ticket touts used AI bots to create "hundreds of thousands of fake registrations" for an upcoming show, overwhelming the sign-up process.
There were 1.8 million sign-ups for a one-off 5,000-seat performance at the Royal Albert Hall on 23 October. That is 360 registrations for every seat.
What she did about it
Raye had already tried to design the touts out. Rather than run a pre-sale, which she felt scalpers could exploit, her team reviewed the sign-up list by hand to produce a shortlist of people they believed were real fans. Those selected get a pre-sale link and code by email; remaining tickets go on general sale.
"I've never witnessed bots on this scale," she said, adding that it was "so important" to her that fans could buy at face value. "I'm not sure what the future holds for ticketing, I do know this experience has really opened my eyes."
Why the defence failed, and what it says
The interesting part is that her approach was the sophisticated one. She correctly identified the pre-sale as the exploitable step and removed it. The touts simply attacked the stage before it.
That is the general shape of this problem. Any free, unverified registration form can be filled a million times by software for almost no cost, and the only reliable filters are ones that impose a real-world cost on each entry — a verified payment method, a phone number that can receive one message, an identity document. Each of those excludes legitimate fans too, which is why nobody uses them.
The fallback was 1.8 million rows reviewed by people. That worked, and it does not scale to a stadium tour.
What it means in Bangladesh
Bangladesh has the same mechanism without the glamour, and it shows up wherever something scarce is given away through a web form.
Cricket tickets are the obvious case. But the pattern is more consequential in queues that are not entertainment: online registration for exams and admissions, government service appointments, and limited-stock flash sales on local e-commerce, where a product priced below market disappears in seconds and reappears for resale.
The defence is the same one, with the same cost. A registration that requires a verified mobile number via OTP is substantially harder to automate at scale than one that requires only an email address — which is why the operators who are serious about it already use OTP, and the ones who are not say they are investigating.
And there is a diagnostic worth knowing, because it applies to any such system here: if a sign-up list is many times larger than the thing being given away, the organiser does not have a popularity problem. They have a verification problem, and the two look identical in a press release.
Why this keeps getting worse
Bot defence used to rest on the assumption that filling a form convincingly was hard. That assumption has quietly stopped being true.
A registration form asks for a name, an email address and sometimes a reason for wanting to attend. Each of those was once a small barrier — a bot filling ten thousand forms produced ten thousand obviously fake entries. Generated text removes that tell completely, and at a cost per entry close to zero. The volume is not the new part; the plausibility is.
Which is why the manual review Raye's team fell back on worked at all. A person can still spot patterns across a list that no individual row reveals — a thousand addresses at the same domain, a burst of registrations in the same second, names that follow a template. That is a different skill from inspecting one entry, and it is the one that still beats the machines.
Source: BBC Newsbeat.
