Retrieval Collapses When AI Pollutes the Web Conference: Proceedings of the ACM Web Conference 2026; April 13–17, 2026; Dubai, United Arab Emirates Proceedings of the ACM Web Conference 2026 (WWW ’26), April 13–17, 2026, Dubai, United Arab Emirates DOI: 10.1145/3774904.3792955 ISBN: 979-8-4007-2307-
Hongyeon Yu, Dongchan Kim, Young-Bum Kim · 2026
Read Retrieval Collapses When AI Pollutes the Web Conference: Proceedings of the ACM Web Conference 2026; April 13–17, 2026; Dubai, United Arab Emirates Proceedings of the ACM Web Conference 2026 (WWW ’26), April 13–17, 2026, Dubai, United Arab Emirates DOI: 10.1145/3774904.3792955 ISBN: 979-8-4007-2307- by Hongyeon Yu, Dongchan Kim, Young-Bum Kim (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-2602-16136.txt.
Description
Hongyeon Yu leads a team of three researchers warning that AI-made web content can break retrieval itself. Search engines and retrieval-augmented systems now draw heavily on text made by large language models. A two-stage failure, called Retrieval Collapse, can then set in. First, AI content dominates results and erodes source diversity. Second, low-quality or adversarial content slips into the pipeline.
Controlled tests cover search-optimized content and adversarially crafted content. In the search-optimized case, a pool about two thirds synthetic leads to far higher exposure in results. Answer accuracy stays stable, yet the state is homogenized and deceptively healthy. Under adversarial contamination, classic BM25 exposes a notable share of harmful content, while LLM-based rankers suppress it better. The authors call for retrieval-aware strategies to stop a self-reinforcing slide in quality.
Catalog record
| CREATOR | Hongyeon Yu, Dongchan Kim, Young-Bum Kim |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 17 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-16136