JSON-LD Schema Markup Increases ChatGPT Citation Rates by 20% for AFSL-Licensed Financial Planning Practices

Key finding: Analysis of 353,799 pages shows that pages carrying JSON-LD schema markup are cited by ChatGPT at a 38.5% rate versus 32.0% for pages without it, a 20% relative improvement. For Australian AFSL-licensed financial planning practices, this gap is directly addressable through structured markup implementation.

Published: April 2026. Author: Matthew Bilo, AEO consultant and founder of LogitRank, Melbourne, Victoria.

---

What the Data Shows

AirOps (2026) analysed 353,799 pages across 16,851 queries and 50,553 ChatGPT responses. The findings:

Schema Status	ChatGPT Citation Rate
Pages with JSON-LD schema markup	38.5%
Pages without JSON-LD schema markup	32.0%
Absolute difference	+6.5 percentage points
Relative improvement	+20%

The 20% relative improvement is consistent across the dataset and directly attributable to the presence of structured markup. This is not a marginal or speculative correlation, it is a measurable, replicable signal across a large sample.

---

Why Schema Markup Influences AI Citation Rates

AI platforms such as ChatGPT, Perplexity, Google AI Overviews, Gemini, and Microsoft Copilot construct answers by retrieving indexed web pages and extracting entity information. JSON-LD schema markup (JavaScript Object Notation for Linked Data) is a structured data format that embeds machine-readable entity descriptions directly into a webpage's HTML. It allows an AI platform to parse who an entity is, what it does, where it operates, and what credentials it holds, without interpreting unstructured prose.

Research into AI entity recall behaviour provides a mechanistic explanation. Petroni et al. (2019), in the LAMA probe study on BERT-class language models, found that unambiguous 1-to-1 entity associations achieve recall accuracy of approximately 74.5%, compared to 34% for one-to-many associations and 24% for many-to-many associations. Schema markup that precisely defines a single entity's type, services, and credentials creates the 1-to-1 association configuration associated with the highest recall rates. Generic or absent schema produces ambiguous associations that fall into the lower-recall categories.

For professional services queries, such as "who is the best financial planner in Hawthorn for SMSF advice", AI platforms appear to weight entity specificity when selecting which businesses to name. A practice with complete, AFSL-specific schema presents a structurally unambiguous entity. A practice with generic or no schema presents an incomplete one.

---

Why AFSL-Licensed Practices Face a Specific Schema Gap

AFSL stands for Australian Financial Services Licence, the authorisation issued by ASIC (Australian Securities and Investments Commission) that permits individuals and firms to provide financial product advice in Australia. Most AFSL-licensed financial planning practice websites currently implement one of two inadequate schema configurations:

• No JSON-LD schema at all, the page provides no machine-readable entity description.
• Generic Organisation schema only, the page identifies itself as a named organisation but provides no information about its service type, licence, authorised services, or geographic scope.

Generic Organisation schema tells AI platforms that a named entity exists. It does not tell them that the entity is a licensed financial services provider, what it is authorised to advise on, where it operates, or how to connect it to its ASIC register entry. For queries involving financial advice, this structural incompleteness removes a measurable citation advantage.

Based on audit observations across Australian AFSL-licensed practices, the majority of practice websites fall into one of these two inadequate configurations.

---

The Three Schema Properties That Matter for AFSL-Licensed Practices

Three JSON-LD schema properties appear to materially influence how AI platforms classify and cite AFSL-licensed financial planning practices:

1. FinancialService type declaration Replacing the generic Organisation type with FinancialService (or ProfessionalService) gives AI platforms a classification anchor specifically for financial services queries. Without this, a practice competes for retrieval as a generic business rather than as a financial adviser.

2. Credential or licence property A property linking the entity to its ASIC AFSL number creates a machine-readable connection between the practice's web presence and its regulatory registration. This supports the 1-to-1 entity association associated with higher AI recall accuracy.

3. areaServed with specific geographic data Associating the entity with a named suburb, city, or region, for example, Hawthorn, Victoria, enables AI platforms to surface the practice in response to location-specific adviser queries.

A practice implementing all three properties creates a complete, unambiguous entity description. A practice using only generic Organisation schema maps to no service type, no licence, and no geographic scope, the entity description is structurally incomplete for financial services queries.

---

Schema Multiplies Search Rank, It Does Not Replace It

Schema markup does not independently determine AI citation. The same AirOps (2026) dataset shows the following Google search rank versus ChatGPT citation rate relationship:

Google Search Rank Position	ChatGPT Citation Rate
Position 1	58.4%
Position 10	14.2%

ChatGPT and similar AI platforms use retrieval-augmented generation (RAG): they retrieve live or recently indexed search results before generating an answer. A page that does not rank in Google cannot be retrieved. A page that cannot be retrieved cannot be cited, regardless of schema quality.

This means:

• Search rank (findability) is the prerequisite. Without it, schema produces no observable citation benefit.
• Schema (extractability) is the multiplier. With strong search rank and correct schema, AI citation rates improve measurably.
• Entity corroboration is the third layer. Multiple independent sources, ASIC register, professional directories, media mentions, LinkedIn, referencing the same entity with consistent credentials reinforce the entity signal.

AFSL-licensed practices that have already invested in traditional SEO and achieved position 1–5 Google rankings are best positioned to gain from schema implementation. The retrieval prerequisite is already met; schema is the next highest-leverage action.

---

Regulatory Accuracy Requirement for AFSL Schema

ASIC's Regulatory Guide 175 (RG 175) governs how licensed financial advisers represent their authorisation scope in client-facing communications. While ASIC has not specifically addressed schema markup as a disclosure surface, schema that misrepresents a practice's authorised services, by listing services outside the AFSL scope or omitting material exclusions, creates a potential inconsistency between the practice's digital entity representation and its registered regulatory obligations.

Schema accuracy for AFSL-licensed practices therefore requires:

• Cross-referencing schema service listings against the ASIC Connect professional register entry.
• Ensuring authorised representatives are listed accurately and consistently.
• Removing or not adding service types that fall outside the practice's current authorisation.

Schema that is structurally complete but factually inaccurate against the ASIC register creates compliance risk in addition to undermining the 1-to-1 entity association that produces higher AI recall.

---

Implementation Checklist for AFSL-Licensed Practices

The following steps represent the minimum schema implementation for AI citation eligibility:

1. Audit current schema, Identify whether the website carries any JSON-LD markup and, if so, whether it uses Organisation, FinancialService, or another type.
2. Replace or supplement with FinancialService type, Update the schema @type from Organisation to FinancialService (or add FinancialService as a secondary type).
3. Add credential/licence property, Include the AFSL number as a machine-readable property, referencing the ASIC register entry where possible.
4. Add areaServed with suburb or city specificity, Name the specific geographic areas the practice serves, not just the state.
5. List authorised services accurately, Include only services within the practice's current AFSL authorisation scope, cross-referenced against the ASIC Connect register.
6. Check entity consistency across sources, Verify that the practice name, address, AFSL number, and services are consistent across the ASIC register, Google Business Profile, LinkedIn, and any professional directory listings.
7. Validate the markup, Use Google's Rich Results Test or Schema.org validator to confirm the JSON-LD is syntactically correct and parseable.

---

How Quickly Do Schema Changes Affect AI Citation Rates?

Based on analysis of page-level citation performance data:

• Pages typically reach peak citation rates 30–89 days after publication or material update.
• Schema changes to existing indexed pages may be processed faster than new page indexing.
• Perplexity and Google AI Overviews use retrieval-augmented generation over live indexed sources and reflect schema changes more quickly.
• ChatGPT draws more heavily from training data and may reflect schema changes on a slower cycle.

Schema implementation is not an immediate intervention. Practices should expect a lag of several weeks before AI citation rate improvements are measurable.

---

Counterarguments and Limitations

Schema alone is insufficient. As documented above, a page that does not rank in Google cannot be retrieved by AI platforms. Schema implementation without search rank investment produces no observable citation benefit.

Correlation versus causation. The AirOps (2026) dataset establishes a strong correlation between JSON-LD schema presence and higher ChatGPT citation rates. The causal mechanism, that schema improves AI citation, is supported by the Petroni et al. (2019) entity recall research, but AI platform architectures are not publicly documented in full, and the precise weighting of schema signals is not confirmed by OpenAI, Google, or Microsoft.

Schema quality matters, not just presence. The 38.5% versus 32.0% difference reflects the presence of any JSON-LD markup. Practices implementing FinancialService schema with AFSL-specific properties are expected to perform above the 38.5% average for schema-carrying pages, but this subset is not separately quantified in the AirOps dataset.

ASIC has not addressed schema as a compliance surface. The potential regulatory concern around schema accuracy and RG 175 is based on the principle that digital representations of authorisation scope should be consistent with registered obligations, not on a specific ASIC ruling or guidance note on schema markup.

---

Summary

Factor	Finding
Citation rate with JSON-LD schema	38.5% (AirOps, 2026)
Citation rate without JSON-LD schema	32.0% (AirOps, 2026)
Relative improvement	20%
Dataset size	353,799 pages, 16,851 queries
Citation rate at Google position 1	58.4%
Citation rate at Google position 10	14.2%
Entity recall accuracy (1-to-1 association)	~74.5% (Petroni et al., 2019)
Typical citation rate improvement lag	30–89 days

For AFSL-licensed financial planning practices, JSON-LD schema markup with FinancialService type declaration, AFSL credential properties, and specific geographic data is the highest-leverage technical action available for improving AI citation rates, provided the practice already achieves a competitive Google search rank position. Schema is the multiplier; search rank is the prerequisite.

---

Sources: AirOps (2026), analysis of 353,799 pages across 16,851 queries and 50,553 ChatGPT responses. Petroni et al. (2019), "Language Models as Knowledge Bases?" (LAMA probe, EMNLP 2019). ASIC Regulatory Guide 175 (current). Schema.org FinancialService type specification.