AI Source Scraper

About

Capture cited web sources and metadata from Claude, Gemini and ChatGPT. Export CSV/JSON for digital methods research.

AI Source Scraper captures every cited web source — with metadata — from the source/activity panels of Claude, Gemini, and ChatGPT, for longitudinal source-recurrence analysis.

Built to operationalise the Longitudinal Prompt Method for Understanding AI Search (https://www.researchgate.net/publication/408580170_Longitudinal_Prompt_Method_for_Understanding_AI_Search), this tool supports digital methods research projects that seek to advance source studies of single or cross-AI models by turning the citation layer of AI chatbots into structured data.

HOW IT WORKS
1. Open a Claude / Gemini / ChatGPT response that searched the web.
2. Open the Sources / Activity panel.
3. Click "Scan sources" in the floating panel — the tool auto-scrolls the panel to load every card, including virtualised lists that only render a few items at a time.
4. Export CSV or JSON.

WHAT IT CAPTURES
Per source: URL, clean URL (tracking parameters stripped), hostname, registrable domain, title, description, favicon, citation position, and whether the source appeared inline in the answer or only in the activity panel — a direct read on citation selectivity.

FEATURES
- Session label field to tag each capture with your prompt
- Buffer that accumulates captures across a session before exporting
- Structure-based detection (no fragile CSS selectors), designed to survive interface redesigns
- Console API (window.aiSourceScraper) for chaining into other tooling

PRIVACY
The extension runs entirely in your browser. It collects no data, sends nothing anywhere, and has no analytics. Exports are downloaded directly to your machine.

Open source: https://github.com/jannajoceli/ai-source-scraper
Citation: Omena, J. J. (2026). AI Source Scraper. Zenodo. https://doi.org/10.5281/zenodo.20945556