Promises and Pitfalls of Using LLMs for Scraping Web UIs

Rebecca Krosnick and Steve Oney

CHI2023-CompUI 2023

PDF

Abstract

ChatGPT and other publicly available large language models (LLMs) put AI into the hands of everyday computer users, offering the possibility of automating computer tasks. One candidate task is web scraping. We informally experimented with ChatGPT to explore the potential promises and pitfalls of using it for scraping data from web user interfaces. We share our observations and considerations for future human-LLM web scraping systems.

Cite this paper (BibTeX)

@inproceedings{krosnick2023promises,
  author    = {Krosnick, Rebecca and Oney, Steve},
  title     = {Promises and Pitfalls of Using {LLMs} for Scraping Web {UIs}},
  booktitle = {CHI 2023 Workshop: Future of Computational Approaches for Understanding \& Adapting User Interfaces},
  year      = {2023}
}