Show HN: Dashboard tracking which sources Perplexity uses (geo-optimization.researchly.at)

🤖 AI Summary
A new dashboard maps which domains the Perplexity Q&A/assistant service actually cites, revealing that corporate websites dominate the citation pool — roughly 76% of sources — while editorial, comparison/portal, UGC and review sites are cited far less. The dataset shows a long tail: the top 10 most-cited domains account for only 12.5% of citations and no single site exceeds a ~20% share (YouTube is the most-cited individual domain). Frequently cited corporate domains include ionos.de, monday.com, zendesk.de, botpress.com, zapier.com and many enterprise SaaS vendors; editorial and review mentions include handelsblatt.com, omr.com and thedigitalprojectmanager.com. The dashboard also surfaces category breakdowns, citation frequency, URL mentions, visibility and sentiment metrics and can generate prompts to probe domain coverage. For the AI/ML community this matters because it provides transparency into retrieval behavior and source bias in a mainstream retrieval-augmented assistant. Heavy reliance on corporate content can skew answers toward vendor-centric framing, affect neutrality and increase risk of missing critical, independent verification — important for model auditing, benchmark design, and downstream trust. Practically, the findings suggest teams should monitor source composition, apply weighting or filtering in RAG pipelines, diversify retrieval indexes, and surface provenance to users. The dashboard offers a simple operational tool for researchers and site owners to quantify citation patterns and prioritize mitigation or outreach.
Loading comments...
loading comments...