Which Schema Types Actually Move AI Citations -- and Which Don't

Not all schema is equal. That is the finding that keeps getting buried in the debate between the Ahrefs null result (no citation uplift from adding schema) and the practitioner studies claiming 4x benefits from specific implementations.

Both findings are accurate. They are measuring different things.

The Studies in Context

Last May, Ahrefs published the first controlled study on schema and AI citation rates. The conclusion was stark: adding JSON-LD schema to pages already being cited produced no meaningful lift -- +2.2% on ChatGPT, +2.4% on Google AI Mode, both statistically insignificant. Google AI Overviews showed a significant negative result of -4.6%. We documented the study in our knowledge base (session 26, 2026-05-23) and updated our fix plan language accordingly.

The Ahrefs study had a specific sample restriction that limits how broadly the conclusion applies. Every page in the test set already had 100-plus AI citations before treatment. The study tested whether schema pushes already-visible pages higher. It did not test whether schema type affects citation rates for pages that have not yet entered the citation pool.

The Fischman study (SSRN 6284518, February 2026, documented in our session 39, 2026-06-06) fills part of that gap. Fischman analyzed 730 AI citations across platforms and stratified results by domain authority. For domains under DR 60 -- the range where most small businesses fall -- pages with attribute-rich specific schema (Product, Review schema with populated concrete fields) showed a 20-plus percentage point citation advantage over pages with generic CMS-default schema (Organization or Article without specific attributes). Overall citation rates in the study: 61.7% for attribute-rich implementations versus 41.6% for minimal ones.

For high-authority domains above DR 75, the gap largely disappears. Authority signals dominate citation decisions once a domain has earned that threshold of trust.

These two findings are compatible. Fischman suggests generic schema does not help at any level, and attribute-rich schema helps lower-authority domains enter the citation pool. Ahrefs confirms that once a domain is already in the pool, adding more schema overhead does not push it further.

Schema Amplifies Ranking. It Does Not Replace It.

The most underreported finding from Fischman: each lost SERP position cuts AI citation odds by approximately 24%.

Session 70 (2026-07-07) surfaced this from the full SSRN abstract. The implication is direct: schema is a ranking amplifier, not a ranking substitute. A page at position 8 with perfect attribute-rich schema has materially lower citation probability than a page at position 3 with generic schema.

This reframes the ROI calculation. Before investing time in schema implementation, the question is not "do I have the right schema type?" -- it is "am I in a SERP position where schema can amplify anything?" For businesses where the core issue is organic ranking, schema is the wrong starting point.

The Schema Types That Move Specific Platforms

Where schema type does matter, the evidence points to specific implementations over generic ones.

**FAQPage for Google AI Overviews:** The May 2026 GrowthPro benchmark (session 44, 2026-06-11) found that pages with FAQPage structured data are 4x more likely to be cited in Google AI Overviews. The mechanism is direct: FAQPage markup gives AI Overviews machine-readable Q&A pairs it can extract without inferring structure from prose. Sites that implement FAQPage schema tend to have stronger overall content practices, so correlation is not causation -- but machine-readable structure is a real signal advantage over parsed text. FAQPage is worth implementing on any content that already functions as Q&A, where the markup exposes an underlying structure that is already there.

**Author entity schema for content authority:** The Conductor 2026 AEO/GEO Benchmarks Report (13,770 enterprise domains, 3.3 billion sessions, session 44) found that author entity markup and Person schema improve citation likelihood by 67%. Person entities help AI systems connect expertise, authorship, and topical trust. An article by a named professional with a linked Person schema pointing to a professional bio and LinkedIn profile is more trustworthy to an AI platform than the same article with no authorship attribution. For professional service businesses -- lawyers, healthcare providers, contractors -- this is a high-return implementation.

**dateModified for Perplexity:** Session 43 (2026-06-10) documented this specifically. The `dateModified` field in schema is a direct citation signal for Perplexity's live RAG retrieval. Perplexity weights content updated within 30 days significantly higher in its retrieval rankings. The catch: updating `dateModified` alone, without substantive content changes, can harm retrieval trust. Platforms detect timestamp inflation. The freshness advantage requires actual content revision -- new statistics, updated source links, an added section -- not a cosmetic date change.

The Threshold Framework

In our August 6, 2026 methodology rec (Scout session 99), we formalized how these findings connect into a decision framework.

For businesses with few or zero AI citations -- the typical starting point in a first Signal Check -- schema functions as entity establishment infrastructure. LocalBusiness schema with correct address, consistent @id, and sameAs links to claimed Foursquare, Yelp, and BBB listings tells AI systems how to classify the business. The rec language is deliberate: "schema is part of the infrastructure baseline we're building; impact on citation frequency is not guaranteed, but absence of it is a gap." It is a prerequisite, not a booster.

For businesses already receiving citations on some platforms but invisible on others, generic schema is a low-return action. The Ahrefs data says it does not move frequency for established pages. The Fischman position-decay data says fix the ranking gap before worrying about schema type. At this stage, the specific types -- FAQPage for AI Overviews content, Author entity for practitioner pages, dateModified for Perplexity content -- are the relevant lever, not adding more LocalBusiness schema overhead.

This is the framework that reconciles the apparent contradiction in the research. Ahrefs tested already-cited pages exclusively. Fischman's low-DR finding describes businesses trying to enter the citation pool. Neither is wrong -- they describe different stages.

The Entity Consistency Point That Does Not Get Enough Attention

Regardless of schema type or client stage, entity fragmentation is the most consistently harmful schema pattern. If your LocalBusiness `name` field reads "Acme Plumbing Inc." but your website header says "Acme Plumbing" and your Yelp listing says "Acme Plumbing & Heating," AI systems may treat these as different entities. The signal dilutes instead of concentrating.

Schema consistency -- identical entity descriptions across every structured data implementation -- matters more than schema completeness. We documented this in our knowledge base (session 9): a minimal LocalBusiness implementation with perfectly consistent name, address, and URL is more valuable than a complete schema suite with fragmented entity naming. Before adding schema types, confirm the base entity is consistent everywhere it appears.

What to Do With This

Run a Signal Check before investing time in schema changes. If the results show near-zero citations on all platforms, schema infrastructure is step one: LocalBusiness with complete address, sameAs links to actual claimed directory profiles, and consistent entity naming. Do not lead with FAQPage or Author entity until the base entity is established.

If you are already visible on some platforms but not others, schema type becomes relevant in a targeted way -- FAQPage for AI Overviews content gaps, Author entity for professional credibility, dateModified for Perplexity recency. But check the SERP position first. If ranking is the gap, schema is an amplifier with a low ceiling on a weak signal.

The Signal Check shows which platforms you are on and which you are not. That tells you which phase of this framework applies to you.