---
language: en
title: "Evidence on schema and AI citations"
description: "Evidence on schema and citations requires distinguishing association, intervention, and the stage being evaluated. Valid structure does not demonstrate a causal increase in AI presence."
canonical: "https://tropk.ai/en/resources/glossary/evidence-on-schema-and-ai-citations"
last_modified: "2026-10-12T16:09:58.337Z"
---

# Evidence on schema and AI citations

Reviewed on 2026-10-07.

## What it means

The challenge with evidence on schema and [AI citations](/en/resources/glossary/ai-citation) is distinguishing the effect of markup from other factors associated with a page. Finding more [structured data](/en/resources/glossary/structured-data) among cited pages demonstrates association, not a causal benefit from adding it. Studies may also evaluate different things, such as information extraction for a direct query or changes in citation frequency over time.

## Why it matters for marketing

For marketing, this distinction prevents turning a technical implementation into an unsupported promise of growth. Ahrefs analyzed pages that were already cited and then added [JSON-LD](/en/resources/glossary/json-ld), comparing their progress with controls, and found no clear gain in [ChatGPT](/en/resources/glossary/chatgpt) or [AI Mode](/en/resources/glossary/ai-mode). The result applies to that scope and method; it does not prove that every type of markup is useless in every context.

## Example in practice

Imagine an agency adds schema to equipment pages and rewrites their descriptions in the same week. It then sees more citations and attributes the increase solely to the markup. Reviewing the change log, it recognizes that both interventions happened together. It decides to separate future changes and compare similar pages, treating the schema hypothesis as a question to investigate.

## How to apply or verify

Before interpreting a study, check the population, systems, versions, query stage, period, and measured outcome. A test of price extraction from an artificial page, such as the one published by searchVIU in 2025, does not directly measure citation growth. In your own evaluations, record concurrent interventions and use a comparable baseline; changes observed without controls call for caution about the attributed cause.

## Important distinctions

Ahrefs' observational analysis is not a randomized experiment and acknowledges limitations involving selection, time window, and simultaneous changes. The searchVIU test also does not describe all current architectures or capabilities. No detectable effect does not prove an absolute absence of effect. Markup validity, understanding, [retrieval](/en/resources/glossary/information-retrieval), and citation are distinct outcomes that require evidence suited to each question.

## Sources and references

- [We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved.](<https://ahrefs.com/blog/schema-ai-citations/>) — Ahrefs
- [Schema Markup and AI in 2025: What ChatGPT, Claude, Perplexity & Gemini Really See](<https://www.searchviu.com/en/schema-markup-and-ai-in-2025-what-chatgpt-claude-perplexity-gemini-really-see/>) — searchVIU

## Continue a leitura

- [AI citation](</en/resources/glossary/ai-citation>)
- [Machine-readable content](</en/resources/glossary/machine-readable-content>)
- [Schema.org](</en/resources/glossary/schema-org>)

## About the author

[Luiz Gardelin](</en/team/luiz-gardelin>)

Luiz Gardelin is the founder of Tropk.ai and works in GEO/AEO, product vision, strategy, and growth, helping brands understand and improve their presence in AI responses.

---

[All pages in Markdown](<https://tropk.ai/sitemap.md>)