Robot pen pals: a multidisciplinary analysis of recent trends in scientific correspondence
BACKGROUND
This summer, we received a suspicious letter to the editor commenting on our recently published results. We noticed the letters author had gone from zero letters ever published in 2024 to more than 80 in 2025. Due to this unusual uptick in frequency, we wondered whether generative AI could have played a role and decided to explore the topic further.
METHODS
We downloaded the records of all letters to the editor available in PubMed from 2005 through September 2025 which totalled more than 730,000. We used descriptive statistics to explore the general trends, identified authors showing an unusual growth in their correspondence corpus and assessed their proportional contribution to all letters published every year. Finally, we calculated the Gini coefficient and Herfindahl-Hirschman Index to assess the inequality and concentration of the corresponding market.
RESULTS
The mean number of letters to the editor published every year has increased in three distinct peaks, one in 2013 due to indexing changes in MEDLINE, one in 2020 coinciding with the pandemic, and one in 2023 coinciding with the popularization of generative AI. The surge is particularly strong among a small group of authors whose collective output went from the bottom 5% in the period 2005–2022 to the top 5% in 2023–2025. This was largely at the expense of all others, which signals Darwinian competition. The economic indexes also reflect a sharp increase in inequality in the correspondence market starting in 2023.
CONCLUSIONS
The diverse metrics we used identify 2023–2025 as a period of abnormal growth in the number of published letters to the editor, which seems attributable to the popularization of generative AI.
Web | DOI | PDF | Preprint: Research Square | Open Access
Carlos Chaccour; Gonzalo Arrondo; Tommaso Cancellario; Fhabián S Carrión-Nessi; Itzel De Haro; Javier García-Manglano; Matthew Rudd; Mirko Abbritti
BACKGROUND
This summer, we received a suspicious letter to the editor commenting on our recently published results. We noticed the letters author had gone from zero letters ever published in 2024 to more than 80 in 2025. Due to this unusual uptick in frequency, we wondered whether generative AI could have played a role and decided to explore the topic further.
METHODS
We downloaded the records of all letters to the editor available in PubMed from 2005 through September 2025 which totalled more than 730,000. We used descriptive statistics to explore the general trends, identified authors showing an unusual growth in their correspondence corpus and assessed their proportional contribution to all letters published every year. Finally, we calculated the Gini coefficient and Herfindahl-Hirschman Index to assess the inequality and concentration of the corresponding market.
RESULTS
The mean number of letters to the editor published every year has increased in three distinct peaks, one in 2013 due to indexing changes in MEDLINE, one in 2020 coinciding with the pandemic, and one in 2023 coinciding with the popularization of generative AI. The surge is particularly strong among a small group of authors whose collective output went from the bottom 5% in the period 2005–2022 to the top 5% in 2023–2025. This was largely at the expense of all others, which signals Darwinian competition. The economic indexes also reflect a sharp increase in inequality in the correspondence market starting in 2023.
CONCLUSIONS
The diverse metrics we used identify 2023–2025 as a period of abnormal growth in the number of published letters to the editor, which seems attributable to the popularization of generative AI.
Web | DOI | PDF | Preprint: Research Square | Open Access