Study citations at both the domain and page level. Domain counts show breadth within the collected answers; the individual pages explain what information was cited. This is a research method, not a published ranking of domains.
- Status
- Study design
- Language
- Spanish
- Results
- No dataset published
A protocol you can reuse
Proposed study design. No collection has been run and no results are implied.
Do Perplexity’s cited domains change between broad recommendations and specific comparisons?
Build 15 matched pairs of Spanish prompts. In each pair, ask a broad category question and a more specific comparison using the same category and market. Collect both prompts on three days: 90 planned answers.
Preserve citation order, exact URLs, and the statement each source supports. Count each domain once per answer. Record unavailable or inaccessible source pages without treating their absence as proof the answer had no source.
Compare domain coverage within each matched pair and question type. Keep answer-level evidence behind the domain totals. Do not compare the totals with a ChatGPT study unless prompt wording, market, dates, and collection coverage are matched.
Collection record
What the study can distinguish
- Compare citations within matched question categories instead of reporting only one aggregate list.
- Record whether a cited page belongs to the subject brand, a competitor, or an independent source.
- Treat source appearance as an observation. It does not establish why the platform selected that page.
Collection and interpretation
- Define the Spanish questions, target market, collection period, and counting rule.
- Retain complete answers and original citation URLs. Deduplicate repeated domain appearances within each answer for domain-level counts.
- Publish counts, denominators, collection gaps, and example source passages beside any ranking.