BRENNEN

Industries / Media, arts and recreation / Media and entertainment / Ticketing

Fraud and bots

Fraud and bots covers the ways tickets are taken or faked: software that buys faster than people can (bots), fake accounts that get around purchase limits, counterfeit tickets, scams and stolen data.

6

Milestones achieved, from 2016 to 2024.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 2016

    1,012 tickets in a minute

    In January 2016 New York's attorney general published "Obstructed View", a report on ticket sales. It found that when U2 tickets went on sale on 8 December 2014, a single broker bought 1,012 tickets to one show at Madison Square Garden in one minute, and by the end of the day it and one other broker held more than 15,000 tickets to U2's North American shows. It found that more than half of the tickets to popular concerts were held back for insiders or presales before the public could buy, that limits on tickets per buyer were not regularly enforced, that bots bought at least tens of thousands of tickets a year, and that brokers marked tickets up by an estimated 49% on average.

    LessonA limit posted on a website holds only if something enforces it.

    ag.ny.gov

    Credit

    New York State Office of the Attorney General

  2. 2016

    A federal law against bots

    The Better Online Ticket Sales (BOTS) Act, signed on 14 December 2016 as Public Law 114-274, made it unlawful in the US to get around a security measure or access control that a ticket seller uses to enforce posted purchase limits, and to sell tickets obtained that way, with enforcement by the FTC.

    govinfo.gov

    Credit

    US Congress

  3. 2017

    Registering fans before sales

    In March 2017 Ticketmaster expanded Verified Fan, which uses customer data such as purchase history to pick out likely fans before a sale and invite them to buy, using it for presales of Ed Sheeran's tour after months of tests on smaller tours, Hypebot reported. By 2022 it was the gate to the largest sales: 3.5 million people registered through it for the Eras Tour.

    hypebot.combusiness.ticketmaster.com

    Credit

    Ticketmaster

  4. 2018

    Britain bans ticket bots

    From 5 July 2018 it has been an offence in the UK to use automated software to buy more tickets than an event allows, punishable by an unlimited fine.

    gov.uk

    Credit

    UK Department for Digital, Culture, Media and Sport and Department for Business, Energy and Industrial Strategy (Margot James, minister)

  5. 2021

    First cases under the law

    On 22 January 2021 the FTC announced what it called "the first case brought under the BOTS Act", against three New York brokers accused of using automated ticket-buying software, software to hide their internet addresses, and hundreds of fictitious Ticketmaster accounts and credit cards. The judgments totalled more than $31 million in civil penalties, mostly suspended because the brokers could not pay; they paid more than $3.7 million.

    LessonMore than four years passed between the law and the first case.

    ftc.gov

    Credit

    Federal Trade Commission; US Department of Justice

  6. 2024

    Stolen data, leaked barcodes

    On 20 May 2024 Live Nation found unauthorized activity in a third-party cloud database holding company data, mainly from Ticketmaster, and on 27 May a criminal group offered what it said was the company's user data for sale on the dark web, the company told the SEC. In July 2024 a group calling itself Sp1d3rHunters posted what it said were barcodes for 170,000 Eras Tour tickets for shows in Miami, New Orleans and Indianapolis and demanded $2 million from Ticketmaster, Security Affairs reported. Whether any leaked code would have worked at a gate was not confirmed by an outside party.

    LessonA ticket seller's database holds both buyers' personal details and the codes that open the gates.

    sec.govsecurityaffairs.com

    Credit

    Live Nation (disclosure); Security Affairs (reporting)

Watch for

  1. A broker's 379,776 tickets

    The FTC alleges that the broker Key Investment Group, its affiliates Epic Seats, TotalTickets.com and Totally Tix, and three executives used thousands of Ticketmaster accounts to buy at least 379,776 tickets in just over a year for nearly $57 million and resold them for about $64 million; for one Eras Tour concert they allegedly used 49 accounts to buy 273 tickets, past a limit of six.

    Achieved whenthe court rules against the defendants or they settle with binding terms.

    ftc.gov

    Claimed by

    FTC, August 2025, in a complaint filed in federal court in Maryland

Open

  1. 01

    A detector for scam posts

    An open, labelled set of social media posts offering tickets, each marked as a likely scam or not, and a small classifier that fan groups and moderators can run to flag likely scams.

    The brief
    Credit · unclaimed
  2. 02

    A test bench for on-sales

    An open simulation of a high-demand on-sale, with a sample ticket shop, a waiting room and 10,000 simulated buyers, some of them scripted bots and AI browser agents, to measure how many tickets bots win under first-come queues, randomized queues, lotteries and pre-registration.

    The brief
    Credit · unclaimed

Open 01 · the brief

A detector for scam posts

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open, labelled set of social media posts offering tickets, each marked as a likely scam or not, and a small classifier that fan groups and moderators can run to flag likely scams.

Why it is still open

Lloyds Bank said that 90% of the concert ticket scams reported by its customers during the UK Eras Tour sales started on Facebook, and estimated UK fans had lost more than £1 million. Searches found general spam classifiers for Facebook and Twitter, but no public dataset or model for ticket scams.

brightonandhovenews.org

Built with what exists

Public posts collected by hand from open groups, with names and contact details removed; labels based on warning signs published by banks and police; a small open-weight model such as Liquid AI's LFM3, fine-tuned on one GPU (see the Note "One GPU to fine-tune"), or a scoring model such as Gemini Decision 1 that returns probabilities for yes-or-no questions (see the Note "A thousand questions, one pass"); scikit-learn baselines.

  1. First step

    Collect and label 300 public posts about one sold-out tour, and measure how often two labellers agree.

  2. Done looks like

    Three thousand labelled posts released under an open license; on posts from a later sale held back for testing, at least 90% precision at 80% recall, scored against labels from two people working separately.

Open 02 · the brief

A test bench for on-sales

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open simulation of a high-demand on-sale, with a sample ticket shop, a waiting room and 10,000 simulated buyers, some of them scripted bots and AI browser agents, to measure how many tickets bots win under first-come queues, randomized queues, lotteries and pre-registration.

Why it is still open

Figures on bots in on-sales come from sellers and the vendors that defend them: Ticketmaster blamed "bot attacks" for part of the Eras Tour traffic, and the FTC alleges brokers with thousands of accounts got past Ticketmaster's limits. Searches found vendor material but no public benchmark comparing ways of allocating tickets against the same bots.

business.ticketmaster.comftc.gov

Built with what exists

pretix, an open-source ticket shop; AWS's Virtual Waiting Room, archived in November 2025 but still published under the Apache 2.0 license; Locust for load testing; Browser Use for agent-driven buyers (see the Note "Deleting 80% to start over"). Everything runs on servers the builder controls, never against a live seller.

github.comcrowdhandler.com

  1. First step

    Set up pretix and the waiting room in a test account and run 1,000 simulated buyers through a first-come queue.

  2. Done looks like

    A public repository and report covering at least four allocation methods against at least three bot strategies, each run 20 times with 10,000 simulated buyers, giving the share of tickets bots took, the median wait for simulated people and the share of people wrongly blocked.