BuyTheBy: A dataset of 18,710 text-based paper mill advertisements with 51,812 timestamped prices

Reese AK Richardson, Spencer S Hong, Anna Abalkina · 2026

Read BuyTheBy: A dataset of 18,710 text-based paper mill advertisements with 51,812 timestamped prices by Reese AK Richardson, Spencer S Hong, Anna Abalkina (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-24576.txt.

Introduction

Richardson, Hong and Abalkina assemble BuyTheBy, a dataset of 18,710 text ads for paper-mill services, drawn from seven businesses in seven countries. Of these, 15,839 carry prices. They cover 20,598 positions for sale across 5,567 products in 14 categories, with 51,812 timestamped price points. The authors present basic analyses showing how the collection can support quantitative study of markets for academic fraud.

Catalog record

CREATORReese AK Richardson, Spencer S Hong, Anna Abalkina
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25228
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.24576

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

BuyTheBy: A dataset of 18,710 text-based paper mill advertisements with 51,812 timestamped prices Reese AK Richardson, Spencer S Hong, Anna Abalkina Abstract The study of paper mills and similar businesses operating in the market for academic and education fraud services is frustrated by the lack of market price data on their various offerings. Here, we assemble BuyTheBy, a large, annotated dataset of timestamped, text-based paper mill advertisements from seven businesses operating out of seven different countries. The dataset consists of 18,710 individual advertisements, of which 15,839 have prices listed. Among these there are 20,598 positions listed as for sale on 5,567 unique products in 14 different product categories with 51,812 timestamped price data points. We perform elementary analysis of this dataset to demonstrate its utility for quantitative understanding of markets for academic fraud services and suggest future use cases. Introduction Recent years have seen the apparent emergence of “paper mills”, typically understood to be commercial organizations that sell authorship positions on pre-written scientific manuscripts and collude with editors to have these manuscripts published without undergoing a rigorous peer review process [ 21 , 9 , 15 , 5 , 11 ] . Paper mill products are suspected to represent an increasing proportion of the published scientific literature [ 15 , 29 , 1 , 24 ] . We and others have recently suggested that paper mills should be conceptualized as just one aspect of the broader global market for academic fraud services, wherein many actors concurrently offer services related to diploma milling, essay milling, contract cheating, admissions fraud and patent milling [ 12 , 13 , 26 , 20 ] . Despite ongoing scholarship and journalism enhancing the detection of paper mill products [ 10 , 25 , 23 , 19 , 22 ] and describing paper mill operations [ 15 , 4 , 18 , 14 , 27 , 17 , 16 ] , there are many unknowns about paper mill operations and the market for paper mill products [ 1 , 8 ] . One gap in knowledge concerns the market pricing of paper mill products. Thus far, most of the data made available on the market prices of paper mill products has been sparse, anecdotal in nature or indirect. For instance, a 2013 investigation found authorship positions in articles in journals indexed by Web of Science being sold by a Chinese firm for 1,600 United States dollars (USD) to 26,300 USD [ 14 ] . A Russian company was found to advertise authorship positions for between 180 euros (EUR) to 5,000 EUR between 2019 to 2021 [ 4 ] . A 2024 investigation found that one editor had been offered 3,000 USD for every paper they accepted in their journal originating from one paper mill [ 15 ] , implying prices for authorship slots roughly at this level. A 2025 investigation detailing the operations of several paper mills lists prices for authorship positions between 20 USD and 1,500 USD [ 27 ] . To aid in the quantitative economic study of paper mill operations, here we assemble BuyTheBy, dataset of structured price data from largely unstructured text-based paper mill advertisements for a variety of services related to publishing and credential fraud. Data and methods To assemble BuyTheBy v1.0, we collected advertisements made in “channels” on messaging app Telegram from three businesses, using Telegram’s “Export chat history” feature. We selected these channels because they were publicly accessible channels where many text-based advertisements written mostly in English featuring price data were posted. We prioritized such datasets to assemble as large a dataset as possible with minimal manual cleanup and redaction. Only text-based messages were downloaded; any advertisements contained within PDF or image files attached to messages were not processed. From an additional four businesses, we collected advertisements posted on the paper mill’s website. Business 1 (B1) apparently operates in India and primarily caters to academics in India. In its Telegram channel, all messages are written in English and prices are listed in Indian rupees (INR). 1,101 messages were sent in this channel from 16 March 2022 to 16 July 2024. The channel was exported on July 16, 2024, after which the channel owner disabled the “Export chat history feature”. We considered each message to be equivalent to one advertisement. An example advertisement is shown in Figure 4 . Business 2 (B2) apparently operates in Iraq and primarily caters to academics in Iraq. In its Telegram channel, all messages are written in English and prices are listed in USD. Messages spanned two channels, one in which 31 messages were sent by B2 from 20 December 2023 to 26 February 2024 and one in which 53 messages were sent by B2 from 26 February 2024 to 15 June 2024. These channels were exported on 14 August 2024 and 26 June 2024, respectively. We considered each message to be equivalent to one advertisement. An example advertisement is shown in Figure 5 . Business 3 (B3) apparently operates in Uzbekistan and primarily caters to academics around Central Asia and the Middle East. In its Telegram channel, most messages are written in English with some messages written in Uzbek. Prices are listed mostly in USD with some prices listed in Uzbekistani sum (UZS). 1,126 messages were sent from 26 March 2020 to 16 March 2026. This channel was exported on 12 March 2025 and 16 March 2026. We considered each message to be equivalent to one advertisement. An example advertisement is shown in Figure 6 . Business 4 (B4) is registered in Latvia and appears to cater to academics around Eastern Europe and Central Asia. It has previously been described elsewhere [ 27 ] . On its website, B4 lists 178 article topics in English in a searchable table. Each entry lists up to eight authorship slots with prices listed in USD and a button labeled “BUY”, which prompts the user to submit a form with their contact information. We archived the state of this website on 11 November 2023. We considered each entry in this table to be equivalent to one advertisement. An example advertisement is shown in Figure 7 . Business 5 (B5) is registered in Russia and appears to primarily cater to academics in Russia and Kazakhstan. It has previously been described elsewhere [ 4 , 17 ] . On its former website, B5 hosted “publication service agreement” pages that list an article topic, an authorship position on that article and a price for this single authorship slot in Russian rubles (RUB). A button on the bottom of each contract redirects the user to a payment page. We archived the state of these pages on 1 October 2021. However, prior to this date, B5 had become aware that its services were being monitored by us and added 17,763 fake contracts where article titles were derived from previously published articles. We were able to discriminate between fake and real contracts because fake contracts did not have a functional purchase button and uniformly listed the price of advertised authorship slot as zero. We exclude these contracts from our dataset. We considered each of the remaining 7,907 pages, with one authorship slot each, to be equivalent to one advertisement. An example advertisement is shown

Read the complete text