Schema markup for law firms — what AI engines actually use vs. what SEO checklists suggest

Open any "technical SEO for law firms" checklist and you'll be told to add a long list of structured data: LocalBusiness, Attorney, LegalService, FAQPage, BreadcrumbList, Review, Person, Article, and a few more. Implement all of it, the checklist implies, and you've "done schema." That advice is a holdover from a search world changing underneath it.

In AI search — where a prospective client asks ChatGPT, Google's AI Overviews, Gemini, Perplexity, or Claude "who's a good immigration lawyer in San Jose" and gets a short, named answer — structured data does something narrower and more specific than the checklists assume. Most of the schema on that list does little for AI visibility. A small subset does real work. The job is knowing which is which, and not mistaking a long markup file for a strategy.

This piece is for the person at a firm who owns the website — a marketing lead, a managing partner, or the agency they've hired — and wants to understand what schema is actually for in 2026, what to prioritize, and where the compliance landmines are.

What schema does — and what it doesn't

Schema.org markup is a vocabulary for telling machines, in a structured way, what the things on your page are: this is an organization, this is its name, this is a person, this is their role, this is a question and its answer. Search engines have used it for years to power rich results — the star ratings, FAQ accordions, and breadcrumb trails you see in Google. Two things are worth being precise about, because the hype in this category runs ahead of the evidence.

First, schema is descriptive, not persuasive. It doesn't make claims true or rankings higher on its own. It makes facts legible. If your site doesn't clearly establish who you are, schema helps a machine extract it; it does not invent authority you haven't earned.

Second — and this is where honesty matters — how much large language models lean on schema is not fully settled. Google has said its AI features draw on the same systems that already use structured data, and AI engines that retrieve from the live web benefit from pages whose facts are easy to parse. But no major AI provider publishes "we rank by schema." So the right posture is: schema is a high-confidence aid for entity understanding and fact extraction, a lower-confidence factor for direct ranking, and cheap enough to be worth doing well. Anyone selling schema as a guaranteed AI-ranking lever is overselling it.

Why extraction is the real job

It helps to picture what happens when an engine builds an answer. It retrieves a set of candidate sources, then tries to pull specific facts from each — who this firm is, where it practices, which attorney handles what, whether the page actually answers the question asked. The cleaner those facts are to parse, the more confidently the model can use them, and the less likely it is to drop your firm for a competitor whose facts were easier to read. That's the mechanism underneath the retrieval architecture we covered in how ChatGPT picks which lawyers to recommend. Schema doesn't argue your case to the model; it removes friction from the model reading you correctly. Framed that way, the priority list writes itself: mark up the facts that decide whether the engine can resolve and quote you, and skip the ornaments.

The schema that actually earns its place

For a law firm whose goal is to be understood and cited by AI engines, a short list does the heavy lifting.

1. Organization (and sameAs) — the single highest-value markup. The most important thing schema can do for AI search is help engines resolve which "Smith Law" you are among the dozens that share the name. An Organization block with your legal name, URL, logo, and a sameAs array pointing to your authoritative profiles — your LinkedIn company page, your bar-association listing, your Justia/Avvo profile, your Google Business Profile — ties those records into one entity. This is the structured-data half of the entity work we covered in entity consolidation for law firms. If you do only one thing, do this one.

2. Person, for attorney profiles. Each attorney bio should carry Person markup: name, jobTitle, the firm as worksFor, alumniOf for their law school, and sameAs linking their own authoritative profiles (bar listing, LinkedIn). AI answers that name a lawyer are resolving a person entity; clean Person markup makes your attorneys legible as the named experts they are. (Watch the compliance line here — more below.)

3. FAQPage, used sparingly and honestly. A genuine FAQ — real questions buyers ask, answered plainly — is one of the most extractable formats for AI answers, because the question/answer structure mirrors how a model assembles a response. The caveat is genuine: don't stuff marketing copy into fake "questions," and remember the answers are public claims subject to advertising rules.

4. Article / BlogPosting, on your insights. If you publish (and you should), mark articles up with author, publish date, and headline. This helps engines attribute your writing to your firm and your attorneys — reinforcing the same entity graph.

5. BreadcrumbList. Modest but free: it helps engines understand your site's structure and topical organization. Low effort, small upside, no downside.

That's most of the value. Notice what's not at the top of the list.

What one clean block looks like

Concretely, the highest-value markup — the Organization block with sameAs — is short. The point isn't volume; it's accuracy and consistency with what appears everywhere else:

{
  "@context": "https://schema.org",
  "@type": "LegalService",
  "@id": "https://smithlaw.example/#firm",
  "name": "Smith & Ruiz LLP",
  "url": "https://smithlaw.example",
  "logo": "https://smithlaw.example/logo.png",
  "areaServed": "San Jose, CA",
  "sameAs": [
    "https://www.linkedin.com/company/smith-ruiz-llp",
    "https://www.avvo.com/attorneys/...",
    "https://www.justia.com/lawyers/...",
    "https://g.page/smith-ruiz-llp"
  ]
}

Three rules make or break it. The name must match your legal name exactly as it appears on the bar record and every directory — "Smith & Ruiz LLP," not "Smith and Ruiz" on one surface and "Smith Ruiz Law" on another. Every URL in sameAs must be a profile you actually control and that actually names you, so the engine can corroborate the link both ways. And nothing in the block can contradict a visible fact on the page. A single accurate block like this does more than ten overlapping types stacked on top of each other.

What the checklists overrate

Review and AggregateRating markup is the headline example. SEO checklists love it because star ratings are eye-catching in traditional results. But self-serving review markup is both heavily discounted by Google and a genuine bar-advertising risk for law firms — testimonials and endorsements are among the most regulated categories of attorney advertising. Marking up reviews you've selected and rendered on your own site invites exactly the kind of scrutiny you don't want. For most firms this is a place to be conservative, not aggressive.

LocalBusiness / Attorney / LegalService types are worth including for a firm with a real office and a service area, but they're not the lever the checklists imply. They help with local/Maps context more than with AI citation, and piling on multiple overlapping legal types doesn't compound the benefit. Use one clear type, fill it accurately, move on.

Volume for its own sake. Ten schema types implemented carelessly — with mismatched names, missing fields, or claims that contradict your visible page content — are worse than three implemented cleanly. Inconsistency between your schema and your on-page facts is a signal of low quality, and it undercuts the entity-resolution job schema is supposed to do.

The compliance overlay nobody puts on the checklist

Here is the part generic SEO advice misses entirely: structured data is still lawyer advertising. A claim does not become exempt from ABA Model Rule 7.1 or your state bar's advertising rules because it's wrapped in JSON-LD instead of a headline. A few concrete tripwires:

• "Specialist" in Person or jobTitle markup. California, Texas, Florida, and roughly fifteen other states reserve "specialist" for attorneys certified by an approved program. An attorney bio that hard-codes "jobTitle": "Immigration Specialist" in schema can be a violation in those jurisdictions just as surely as it would be in visible text. Use "practice focused on" or a named certification instead. (We went deep on this rule surface in AI search and ABA Rule 7.1.)
• Superlatives in descriptions. "Best," "top-rated," "leading," "premier" inside an Organization description carry the same exposure as on the page. Schema is not a backdoor for claims you couldn't otherwise make.
• Review/testimonial markup as noted above — the highest-risk category; treat with caution and your jurisdiction's disclaimers.
• Past-results data. If you mark up case results, the same disclaimer requirements that apply on the page apply to the structured version.

The practical rule: anything you put in schema should pass the same compliance review you'd run on visible marketing copy. At Legal Torch that review is the C in the VERDICT™ methodology — every output, structured data included, screened against ABA Model Rules and the relevant state-bar rules before it ships. It's the part most agencies skip, and structured data is exactly where it quietly gets skipped. (To be precise about roles: Legal Torch is not a law firm, and that screen is a marketing-content review, not legal advice — the firm's attorneys decide and approve.)

A sensible implementation order

If you're starting from a typical firm site, prioritize like this:

1. One clean Organization block with sameAs linking your real, authoritative profiles. Get your legal name and URL exactly consistent with how they appear elsewhere.
2. Person markup on every attorney bio, compliance-checked, with sameAs to each attorney's bar and professional profiles.
3. A genuine FAQPage on the pages where you actually answer buyer questions.
4. Article markup on your insights, and BreadcrumbList site-wide.
5. One accurate local/legal type if you have a physical presence — then stop.

Skip, or handle with caution: self-rendered review/rating markup, redundant overlapping legal types, and any claim you wouldn't put in a headline.

Then verify it — don't assume it

Schema you never test is schema you don't know works. Two free checks catch most problems. Google's Rich Results Test shows you what Google can parse from a live URL and flags missing required fields; the Schema.org validator catches syntax and vocabulary errors. But the check that matters most for AI search isn't a tool — it's consistency: read your rendered markup next to your visible page and confirm every fact agrees, then confirm the same name, URL, and profile links appear identically across your site, your directories, and your bar listing. An engine that finds your firm described three slightly different ways learns to trust none of them. Re-run the checks whenever you rebrand, move offices, add an attorney, or change a profile — the moments when schema and reality quietly drift apart.

The point

Schema isn't magic, and it isn't a checklist to be maxed out. For AI search, its real job is narrow and valuable: make your firm and your attorneys resolvable as coherent, consistent entities, and make your genuine answers easy to extract — all while staying inside the advertising rules that apply to everything a law firm publishes. Do that small set well and you've done more for your AI visibility than a page full of markup no engine trusts.

If you want to see where your firm currently stands — both your AI-search citation position and your existing schema/compliance posture — the free AI Visibility Audit covers both in about 60 seconds.

New to the topic? Start with the complete guide: GEO for Law Firms — what it is, what it costs, and what the citation data actually shows.