The adoption and efficacy of large language models in US consumer financial complaints

Minkyu Shin; Jin Kim; Jiwoong Shin

doi:10.1038/s41562-026-02409-4

The adoption and efficacy of large language models in US consumer financial complaints

Minkyu Shin, Jin Kim, Jiwoong Shin

Department of Marketing

Research output: Journal Publications and Reviews › RGC 21 - Publication in refereed journal › peer-review

Abstract

This research explores the impact of large language models (LLMs) on consumer complaints submitted to the US Consumer Financial Protection Bureau. Analysing 1,134,512 complaints from 2015 to 2024, we document a sharp increase in LLM usage following the release of ChatGPT. An instrumental variable analysis estimates that LLM usage increases the probability of obtaining favourable relief by 6.9 percentage points (95% confidence interval, (4.9, 8.9)). The analysis also reveals evidence of negative selection, where consumers otherwise prone to adverse outcomes are more likely to adopt LLMs. To further substantiate these findings and test the mechanism, we conducted three online controlled experiments (total N = 1,010 US participants); these demonstrate that LLMs can increase the likelihood of obtaining relief by enhancing the presentation of complaints without altering factual content. These findings suggest that LLMs can act as an equalizer, highlighting the need for policies that expand access to these technologies.

© The Author(s), under exclusive licence to Springer Nature Limited 2026.

Original language	English
Number of pages	12
Journal	Nature Human Behaviour
Online published	18 Feb 2026
DOIs	https://doi.org/10.1038/s41562-026-02409-4
Publication status	Online published - 18 Feb 2026

Funding

We thank the Department of Marketing at the City University of Hong Kong for financial support (grant ID 7200769 to M.S.). The funder had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. We also thank Y. Li, L. Guo, K. Uetake, S. Ghili, C. Wang, H. Fong, T. Chan, Y. Chen, O. Urminsky, T. L. Griffiths, G. Zauberman, A. Dukes, K. Sudhir, P. Arcidiacono, C. Fuchs, G. Packard, S. Puntoni, J. Joo, S. Ma, H. Park, L. Su and C. Song for their constructive feedback, as well as seminar participants at the Annual Business and Generative AI Workshop by Wharton, the Symposium on AI in Marketing, the Fisher AI in Business Conference, the China Marketing Science Conference, the AIM Conference, the AI ML and Business Analytics Conference, the Hong Kong Joint School Marketing Conference, Korea University, KAIST, CKGSB, the Informs Marketing Science Conference and the Marketing Exchange Forum.

Access Output

10.1038/s41562-026-02409-4

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@article{f200a728878e4ee9bb41122a6d1d729b,

title = "The adoption and efficacy of large language models in US consumer financial complaints",

abstract = "This research explores the impact of large language models (LLMs) on consumer complaints submitted to the US Consumer Financial Protection Bureau. Analysing 1,134,512 complaints from 2015 to 2024, we document a sharp increase in LLM usage following the release of ChatGPT. An instrumental variable analysis estimates that LLM usage increases the probability of obtaining favourable relief by 6.9 percentage points (95% confidence interval, (4.9, 8.9)). The analysis also reveals evidence of negative selection, where consumers otherwise prone to adverse outcomes are more likely to adopt LLMs. To further substantiate these findings and test the mechanism, we conducted three online controlled experiments (total N = 1,010 US participants); these demonstrate that LLMs can increase the likelihood of obtaining relief by enhancing the presentation of complaints without altering factual content. These findings suggest that LLMs can act as an equalizer, highlighting the need for policies that expand access to these technologies.{\textcopyright} The Author(s), under exclusive licence to Springer Nature Limited 2026.",

author = "Minkyu Shin and Jin Kim and Jiwoong Shin",

year = "2026",

month = feb,

day = "18",

doi = "10.1038/s41562-026-02409-4",

language = "English",

journal = "Nature Human Behaviour",

issn = "2397-3374",

publisher = "Nature Publishing Group",

}

TY - JOUR

T1 - The adoption and efficacy of large language models in US consumer financial complaints

AU - Shin, Minkyu

AU - Kim, Jin

AU - Shin, Jiwoong

PY - 2026/2/18

Y1 - 2026/2/18

N2 - This research explores the impact of large language models (LLMs) on consumer complaints submitted to the US Consumer Financial Protection Bureau. Analysing 1,134,512 complaints from 2015 to 2024, we document a sharp increase in LLM usage following the release of ChatGPT. An instrumental variable analysis estimates that LLM usage increases the probability of obtaining favourable relief by 6.9 percentage points (95% confidence interval, (4.9, 8.9)). The analysis also reveals evidence of negative selection, where consumers otherwise prone to adverse outcomes are more likely to adopt LLMs. To further substantiate these findings and test the mechanism, we conducted three online controlled experiments (total N = 1,010 US participants); these demonstrate that LLMs can increase the likelihood of obtaining relief by enhancing the presentation of complaints without altering factual content. These findings suggest that LLMs can act as an equalizer, highlighting the need for policies that expand access to these technologies.© The Author(s), under exclusive licence to Springer Nature Limited 2026.

AB - This research explores the impact of large language models (LLMs) on consumer complaints submitted to the US Consumer Financial Protection Bureau. Analysing 1,134,512 complaints from 2015 to 2024, we document a sharp increase in LLM usage following the release of ChatGPT. An instrumental variable analysis estimates that LLM usage increases the probability of obtaining favourable relief by 6.9 percentage points (95% confidence interval, (4.9, 8.9)). The analysis also reveals evidence of negative selection, where consumers otherwise prone to adverse outcomes are more likely to adopt LLMs. To further substantiate these findings and test the mechanism, we conducted three online controlled experiments (total N = 1,010 US participants); these demonstrate that LLMs can increase the likelihood of obtaining relief by enhancing the presentation of complaints without altering factual content. These findings suggest that LLMs can act as an equalizer, highlighting the need for policies that expand access to these technologies.© The Author(s), under exclusive licence to Springer Nature Limited 2026.

UR - https://www.webofscience.com/wos/woscc/full-record/WOS:001693661500001

UR - http://www.scopus.com/inward/record.url?scp=105030518209&partnerID=8YFLogxK

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-105030518209&origin=recordpage

U2 - 10.1038/s41562-026-02409-4

DO - 10.1038/s41562-026-02409-4

M3 - RGC 21 - Publication in refereed journal

SN - 2397-3374

JO - Nature Human Behaviour

JF - Nature Human Behaviour

ER -

The adoption and efficacy of large language models in US consumer financial complaints

Abstract

Funding

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this