What this guide helps you decide
Measure whether AI assistants rely on the same sources for commercial questions.
The questions behind the decision
- Do ChatGPT, Gemini, Claude and Perplexity cite the same sources?
- Should overlap be measured by domain or URL?
- What can citation overlap not prove?
What this page adds
This page turns the question "Do ChatGPT, Gemini, Claude and Perplexity cite the same sources?" into a reproducible method for measure whether ai assistants rely on the same sources for commercial questions, with required evidence and explicit failure conditions.
Use two overlap levels
Measure exact URL intersection and normalized domain intersection. A provider may cite different pages from the same authoritative publisher.
Fix the question context
Use the same commercial question and market window. Preserve provider retrieval mode and model where observable.
Classify the source ecosystem
Separate owned product pages, independent publications, reviews, directories, community threads, documentation and research. This classification is often more actionable than one overlap percentage.
Publish limitations
Answers are stochastic and source interfaces differ. The study describes observed retrieval for its sample; it cannot expose proprietary ranking logic or guarantee future citation.
Where this fits
The methods and published data behind Eli's measurements are on the research page. For how assistants find, cite and recommend a company, see AI search for B2B.
Questions buyers ask next
Do ChatGPT, Gemini, Claude and Perplexity cite the same sources?
Provider citation overlap should compare exact canonical URLs and source domains for the same buyer questions, market and time window. Domain overlap shows shared authority patterns, while URL overlap shows whether providers retrieve the same specific evidence. Missing citations and provider failures must remain explicit. The study should not infer hidden ranking weights from a small sample.
Should overlap be measured by domain or URL?
Use the same commercial question and market window. Preserve provider retrieval mode and model where observable.
What can citation overlap not prove?
Answers are stochastic and source interfaces differ. The study describes observed retrieval for its sample; it cannot expose proprietary ranking logic or guarantee future citation.
Primary sources
Related guides
Check your own AI-search gap
Use the decision behind “Do ChatGPT, Gemini, Claude and Perplexity cite the same sources?” as your starting point. Run the free AI Citation Gap Checker to inspect the current public evidence. To keep monitoring the question and prepare a supported website improvement, Eli Free covers one site and one diagnostic, with the findings kept, no card and no expiry. External rankings and AI recommendations are never guaranteed.