Does structured data change AI search citations?
Research proposal — no data collected yet. The method is published first so any result, positive or negative, can be checked against it.
Short answer
Open question with a proposed method. Google states no special schema.org markup is required for its AI features. This entry is a research proposal and reports no results yet.
Question
Section titled “Question”Does adding valid schema.org JSON-LD (e.g. Article, FAQPage, HowTo) to existing pages change how often they are cited as sources in AI search answers?
What is already documented
Section titled “What is already documented”Hypotheses
Section titled “Hypotheses”- H0: Citation rate does not differ between pages with and without added markup.
- H1: Pages with added markup are cited more often.
Proposed method
Section titled “Proposed method”| Element | Design |
|---|---|
| Sample | ≥ 60 existing, indexed pages on 2+ sites; matched pairs by topic and traffic |
| Treatment | Add JSON-LD to one page per pair; no visible-content changes |
| Control | Paired page unchanged |
| Prompts | Fixed panel of ≥ 100 queries, written before treatment, stored in the repo |
| Engines | Google AI Overviews / AI Mode, ChatGPT search, Perplexity; logged separately |
| Window | 4 weeks baseline, 8 weeks after treatment |
| Metric | Share of prompts in which the page is cited, per engine |
| Analysis | Paired difference-in-differences; report confidence intervals, not single runs |
Known threats to validity
Section titled “Known threats to validity”- AI answers are volatile run to run, so single samples are noise.
- Re-crawl timing differs per page. Check when each page was last fetched.
- Engines change during the window. Log dates of announced updates.
Status
Section titled “Status”Awaiting a second reviewer of the method before data collection.