Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ShareChat: A Dataset of Chatbot Conversations in the Wild

Yueru Yan, Tuc Nguyen, Bo Su, Melissa Lieffers, Thai Le

TL;DR

ShareChat addresses the gap in public LLM evaluation by preserving interface context across multiple platforms. The authors assemble 142,808 conversations with over 660k turns from five platforms (ChatGPT, Claude, Gemini, Perplexity, Grok), capturing platform-specific features, longer context, and multilingual content. They demonstrate three analyses—conversation completeness, source-domain usage, and temporal dynamics—to reveal how design choices shape user interaction and grounding behavior. The dataset's public-sharing provenance yields lower toxicity and reveals platform-specific interaction patterns, offering a valuable resource for studying real-world human-AI collaboration and for improving evaluation and alignment of multi-turn chat systems.

Abstract

While Large Language Models (LLMs) have evolved into distinct platforms with unique interface designs and capabilities, existing public datasets treat models as generic text generators, stripping away the interface context that actively shapes user interaction. To address this limitation, we present ShareChat, a large-scale, cross-platform corpus comprising 142,808 conversations and over 660,000 turns collected from publicly shared URLs across five major platforms: ChatGPT, Claude, Gemini, Perplexity, and Grok. ShareChat distinguishes itself by preserving native platform affordances often lost in standard logs, including reasoning traces, source links, and code artifacts, while spanning 101 languages over the period from April 2023 to October 2025. Furthermore, ShareChat offers substantially longer context windows and greater interaction depth than prior datasets. We demonstrate the dataset's multifaceted utility through three representative analyses: (1) analyzing conversation completeness to measure user intent satisfaction; (2) evaluating source citation behaviors in content generation; and (3) conducting temporal analysis to track evolving usage patterns. This work provides the community with a vital and timely resource for understanding authentic user-LLM chatbot interactions in the wild.

ShareChat: A Dataset of Chatbot Conversations in the Wild

TL;DR

ShareChat addresses the gap in public LLM evaluation by preserving interface context across multiple platforms. The authors assemble 142,808 conversations with over 660k turns from five platforms (ChatGPT, Claude, Gemini, Perplexity, Grok), capturing platform-specific features, longer context, and multilingual content. They demonstrate three analyses—conversation completeness, source-domain usage, and temporal dynamics—to reveal how design choices shape user interaction and grounding behavior. The dataset's public-sharing provenance yields lower toxicity and reveals platform-specific interaction patterns, offering a valuable resource for studying real-world human-AI collaboration and for improving evaluation and alignment of multi-turn chat systems.

Abstract

While Large Language Models (LLMs) have evolved into distinct platforms with unique interface designs and capabilities, existing public datasets treat models as generic text generators, stripping away the interface context that actively shapes user interaction. To address this limitation, we present ShareChat, a large-scale, cross-platform corpus comprising 142,808 conversations and over 660,000 turns collected from publicly shared URLs across five major platforms: ChatGPT, Claude, Gemini, Perplexity, and Grok. ShareChat distinguishes itself by preserving native platform affordances often lost in standard logs, including reasoning traces, source links, and code artifacts, while spanning 101 languages over the period from April 2023 to October 2025. Furthermore, ShareChat offers substantially longer context windows and greater interaction depth than prior datasets. We demonstrate the dataset's multifaceted utility through three representative analyses: (1) analyzing conversation completeness to measure user intent satisfaction; (2) evaluating source citation behaviors in content generation; and (3) conducting temporal analysis to track evolving usage patterns. This work provides the community with a vital and timely resource for understanding authentic user-LLM chatbot interactions in the wild.

Paper Structure

This paper contains 23 sections, 10 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Data Collection
  3. Filtering of Identifiable Information.
  4. Preliminary Analysis
  5. Basic Statistics
  6. Platform-Specific Interaction and Metadata Attributes
  7. Toxicity Analysis
  8. Topic Analysis
  9. Use Cases
  10. Conversation Completeness Analysis
  11. Response Source Analysis
  12. Timestamp Analysis
  13. Discussion
  14. Conclusion
  15. Appendix
  16. ...and 8 more sections

Figures (10)

  • Figure 1: Turn Length Distribution Across datasets
  • Figure 2: Distribution of the top 10 languages in ShareChat.
  • Figure 3: Turn Level Toxicity Comparison by Platform. These bar charts compare toxicity detection rates at the individual message level across five AI platforms using two different detection methods: Detoxify (language filtered) and OpenAI Moderation API. The left panel shows LLLM message toxicity rates, while the right panel shows user response toxicity rates.
  • Figure 4: Average topic distribution of user requests across 5 platforms.
  • Figure 5: Conversation completeness analysis across five platforms. Panel (a) shows the proportion of user intentions receiving complete, partial, incomplete, or unknown verdicts at the intention level. Panel (b) displays the distribution of conversation-level completeness scores as violin plots, with the bold horizontal line in each violin indicating the median score. Panel (c) presents the distribution of the number of extracted intentions per conversation on a logarithmic scale, with median values labeled above each box.
  • ...and 5 more figures