Schema in a week: five JSON-LD moves that move AI citation

Every AI Visibility audit Aday Interactive, Inc. runs on a professional-services firm finds the same top issue: schema coverage is thin, and the schema that exists does not cross-reference cleanly. This post names the five schema moves that close that gap. They can genuinely ship in a week, in the sequence below. They are also the deliverable inside the [Schema in a Week](/pricing#ai-visibility) execution tier.

None of these moves are novel. What is uncommon is running all five together, in the right sequence, with the entity `@id` cross-references wired correctly. That is where citation performance moves.

1. Organization + LocalBusiness: the foundation graph

Every page on the site emits Organization schema. The homepage also emits LocalBusiness. The Organization `@id` is a stable URI (typically `/#organization`). The LocalBusiness `@id` is a separate stable URI (`/#localbusiness`). It references the Organization via `parentOrganization`.

The fields matter. Legal name (exactly matching the firm\'s registration: "Firm Name, P.A." not "The Firm"), address, geo coordinates, telephone, email, sameAs array (Google Business Profile, LinkedIn, bar directory profile, etc.), aggregateRating if applicable and honest. AI engines combine these fields into the entity anchor. Every later schema block on the site hangs off that anchor.

One common failure: the LocalBusiness `@id` doesn\'t reference the Organization `@id`. AI engines then see two entities. This is the entity split we flag every time we run a [GEO Audit](/geo-audit). The schema is technically valid. But the graph is disconnected, and citation confidence collapses.

Here is the wire in plain terms. The disconnected version emits an Organization with an `@id` ending in `#organization`, and a separate LocalBusiness with an `@id` ending in `#localbusiness`, and the two never point at each other. A model reads them as two businesses that happen to share an address. The connected version adds one reference: the LocalBusiness names the Organization in its `parentOrganization`, using that exact `#organization` id. Now the model reads one firm, with one address, one phone, and one set of reviews. Same facts, one extra line, and the entity resolves.

2. Person / Attorney / Physician / RealEstateAgent: the people graph

Every team page or bio page carries Person schema. Vertical subtypes matter. Law firms emit Attorney (or the extended type via schema.org/Attorney). Medical practices emit Physician. Wealth firms emit FinancialAdvisor via the Person base with jobTitle set correctly. Luxury real estate emits RealEstateAgent.

Each Person block references the Organization via `worksFor`. `sameAs` links to the vertical-specific directory profiles: Florida Bar directory for attorneys, IAPD or BrokerCheck for RIAs, Healthgrades or state license lookup for physicians, MLS profile or brokerage-specific ID for RealEstateAgent.

The upgrade opportunity: bar admissions, jurisdictions, credentials, board certifications, medical specialties. These fields exist in the schema spec. Most firms\' current schema does not emit them. AI engines cite these fields when a prompt is jurisdiction- or specialty-specific: who is a board-certified estate planning attorney admitted in Florida and Texas.

3. FAQPage with specificity-recipe answers: the extraction layer

FAQPage is the single highest-leverage schema type for AI citation. It lets AI engines extract answers directly from the page rather than inferring them from prose. But it only works when the answers meet the specificity recipe.

Every FAQ answer carries at least one of: a named regulation with rule number, a named vendor or tool, a dollar or time anchor, a worked pattern by firm size or vertical sub-segment, or an internal link. Answers that carry none of these read as generic to LLMs. The engines skip them for a competitor\'s more specific answer.

The recipe in one line. Generic answer: "We help wealth firms stay compliant with marketing rules." An engine reads that as boilerplate and moves on. Specific answer: "Advisor bios and client testimonials are structured to meet the SEC marketing rule, 206(4)-1, including the required disclosures for endorsements." Named rule, concrete work, extractable and attributable. The second answer can be cited. The first one cannot.

The scale to hit is 12 to 20 FAQ entries per practice area or specialty. Not five. Not 40. AI engines credit density. They also credit variety. The 12 to 20 range covers the common questions plus the sub-common ones that reveal domain depth.

4. Service, or the vertical subtype

Every service-line page emits Service schema, or better, the vertical subtype: LegalService for law-firm practice areas, MedicalBusiness or MedicalTherapy for practice specialties, FinancialService for wealth-management service lines, RealEstateAgent for brokerage service pages.

`provider` references the Organization; `serviceType` is the vertical-standard label (e.g. "Business Litigation", "Concierge Medicine", "Wealth Management for High-Net-Worth Families"); `areaServed` names the metro or state (Miami-Dade, Florida); `hasOfferCatalog` optionally lists sub-services within the practice area.

Most firms miss this pattern: emit one Service block per practice area, not one block for the whole firm. AI engines match query intent against `serviceType`. A prospect asks about "commercial real-estate litigation Coral Gables". They need to find a Service block with `serviceType: Commercial Real Estate Litigation`, not a generic "Legal Services" block.

5. Article + BlogPosting: the editorial layer

Every blog post emits Article schema. The blog index emits BlogPosting entries for each post. Fields: `headline`, `description`, `datePublished`, `dateModified`, `author` (referencing the Person `@id` of the author, not just a string), `publisher` (referencing the Organization `@id`), `mainEntityOfPage`.

The `datePublished` and `dateModified` fields matter for AI engines\' recency scoring. ChatGPT and Perplexity both credit recent, actively-maintained editorial content over stale pages. Any date more than 18 months old and never updated reads as static. Refreshing the content and updating `dateModified` is a citation-lift move by itself.

The `author` reference makes the byline citation-eligible. Wire the Article schema to his Person `@id` and AI engines cite Brandon Aday of Aday Interactive, Inc. by name. Without that reference, the citation collapses to just the firm name. The individual attribution is lost.

The sequence, and the week

The five moves ship in this exact order over the week:

• Day 1 to 2: Organization + LocalBusiness sitewide, with the entity graph wired.
• Day 3: Person / vertical-subtype on every team page.
• Day 4 to 5: Service or vertical subtype on every service-line / practice-area page.
• Day 5 to 7: FAQPage on the top three FAQ candidates, with 12 to 20 entries each, authored to the specificity recipe.
• Day 7: Article + BlogPosting on the blog corpus.

Ship all five together and citation performance moves. Ship three of the five and it does not. The entity graph stays incomplete. AI engines default to citing a competitor whose graph is whole. That is why this is a Schema in a Week engagement, not a Schema Whenever engagement.

What comes next

Schema is the foundation. Above it sits the editorial cadence: ongoing, specificity-rich content that keeps the entity fresh in AI-engine training and embedding cycles. That layer is scoped in the [90-day AI Visibility Plan](/pricing#ai-visibility). The Schema in a Week credits fully into it.

Before either engagement, run the [free AI Visibility Scorecard](/lp/visibility-scorecard). It is a 10-minute self-audit of the citation surfaces AI engines actually check. It shows which of the five moves your firm is missing today.