Schema markup used to be a rich-snippet competition — bolt on the right structured data, get stars, FAQ accordions, sitelinks, the cosmetic stuff that made your listing bigger on the page. That game is mostly over. The new game is whether an AI Overview, an SGE answer panel, or a ChatGPT response will cite your firm — and schema’s role has shifted from snippet decoration to citation eligibility. Most schema advice law firms are still being sold is calibrated to the old game. Here’s what actually matters now.
Schema is not magic. It does not, by itself, rank a page. It’s a structured way to tell a machine what’s on the page so the machine doesn’t have to guess. When the machines were just Google’s classic indexer, that meant slightly better snippets and the occasional rich result. When the machines now include Gemini, Claude, ChatGPT’s web tool, Perplexity, and AI Overviews, “tell the machine what’s on the page” is much higher-leverage — but only for the schema types that map to how those systems extract and cite.
The brief version of my position: a small handful of schema types are structurally more important than they used to be, and a few that agencies still implement wholesale are essentially decorative. If you’re paying for “comprehensive schema markup” right now, there’s a real chance most of the work is in the second bucket.
The old schema story (and why it’s mostly over)
From 2014 through 2023, schema markup for law firms was about competing for visual space on the SERP. Review schema produced stars. FAQ schema produced accordions. HowTo schema produced step-by-step results. LegalService schema fed the knowledge panel.
Between late 2023 and 2025, Google quietly took most of it away. Review stars for self-hosted reviews got deprecated. FAQ accordions stopped showing for the vast majority of queries. HowTo rich results were retired. The visual real estate got swallowed by AI Overviews, Local Service Ads, and the crowded ad load above the organic results. If you’re still hearing schema pitched as “we’ll add FAQ schema and you’ll get the dropdowns under your listing,” the agency is selling you 2019.
The new schema story — AI citation eligibility
When an AI Overview renders for a legal query — “what is comparative negligence in Arizona,” “how long do I have to file a personal injury claim in Phoenix” — it pulls facts from a small number of source pages and cites them. The pages that get cited share a pattern: clean information architecture, clear authorship signals, structured fact extraction, and an indexable answer that maps cleanly to the question.
Schema’s new job is to make your page easier to extract for that citation. Not flashier on a SERP. Easier to parse for a machine deciding whether to quote you. Same dynamic for ChatGPT’s web tool, Perplexity, Gemini, and whatever Apple ships. They all reward structured, source-rich pages — and penalize pages where facts are buried in JavaScript or impossible to attribute. Schema doesn’t get you cited. The content has to be worth citing. But if the content is good and the schema is missing, you make it harder for the machine to choose you.
The schema types that still pay for law firms
Four types, in priority order. Implement these well and you’ve done 80% of the useful schema work for a law firm site in 2026.
LegalService (or Attorney). The foundational layer. Done right, it tells every machine: this organization is a law firm, here’s the practice area, here’s the service area, here are the credentials. Most firms have a shell of this and most agencies fill it with the obvious fields. Where it falls apart is the practice-area enumeration. A firm doing PI, family, and estate should have those expressed as discrete service offerings, not lumped into a single description string. That detail is what gives an AI summarizer enough information to confidently cite you for the right practice area. Detail on schema for law firms here.
Person (for the attorneys). Every attorney bio page should have Person schema with credentials, bar admissions, years of practice, and affiliations. The reason isn’t a knowledge panel — most firms won’t get one. The reason is that AI systems are trained to weight cited information higher when the source is attributable to an identifiable expert. Person schema is the structured form of “this article was written by a real lawyer with real credentials.” Skip it and you’re indistinguishable from generic legal content in the AI’s source-quality calculation.
FAQPage — used differently than before. FAQ rich results in the SERP are gone for legal queries. FAQPage schema as a structured-fact-extraction signal is more valuable than ever. It pays on Q&A pages built around high-intent queries — “what is the statute of limitations on a car accident in Arizona.” Mark up as FAQPage, write the answer concisely as the first sentence, then expand. AI Overviews disproportionately cite pages structured this way because the machine can extract the precise answer fragment without interpreting a paragraph. More on what schema law firms actually need here.
BreadcrumbList. The least sexy of the four and one of the most under-implemented. Breadcrumb schema helps every crawler understand your information hierarchy — practice area → subtopic → answer page. For AI citation, that hierarchy is a confidence signal. A page in a clear topical neighborhood is more likely to be cited than an orphaned page that could be about anything. Most law firm sites don’t have it at all. The fix is twenty minutes.
Schema isn’t about looking better in the search results anymore. It’s about making it easier for an AI to decide your page is worth quoting. The firms that get that distinction right are the ones who’ll still be visible when the SERP is mostly AI.
What’s now overrated (and what we tell clients to skip)
Self-published Review schema. Google’s 2023 change — pulling review snippets for self-hosted reviews — was not reversed. Adding Review or AggregateRating schema with testimonials you collected yourself does almost nothing for visibility now and runs you into ABA Rule 7.1 territory in several states. If you want review signals to count, get them on Google itself. More on what still works in legal SEO with AI search here.
Wildly complex Organization markup. Nesting sameAs links to every social profile and directory listing is a relic of “fill the knowledge panel.” Modern AI systems don’t need 30 sameAs entries to figure out who you are. A clean LegalService block with firm name, address, phone, URL, logo, and three or four genuinely authoritative sameAs entries does the same work — and simpler markup is less likely to break.
HowTo and Speakable schema. HowTo was retired by Google — stop implementing it. Voice search for legal queries never became the volume channel agencies promised in 2018. If either is on your invoice, that’s a budget line that should be doing something else.
The 90-minute schema audit you can do this week
Minutes 0–20. Pull your homepage and three top practice pages into Google’s Rich Results Test and Schema.org Validator. The most common findings: LegalService present but missing service-area precision; Person schema missing entirely from bio pages; FAQPage applied to pages that aren’t actually Q&A; BreadcrumbList missing site-wide.
Minutes 20–50. On the LegalService block, expand service offerings into a discrete list. If you do PI, family, and estate, those should be three Service entries, each with a description that mirrors how a prospect would describe the problem. The highest-impact schema move for most firms — and almost nobody does it.
Minutes 50–70. Add Person schema to every attorney bio with bar admissions, years of practice, alumni-of (law school), and member-of (relevant bar associations). Five minutes per attorney once the template is set.
Minutes 70–90. Implement BreadcrumbList site-wide. On the top three Q&A pages, ensure FAQPage schema mirrors the actual on-page Q&A and that the answer’s first sentence is a complete, extractable response. Anything else can wait. More on how a ranking practice page is structured here.
That’s it. If your agency bills you for more than this, ask what’s in the rest of the bucket. The honest answer is mostly busywork or deprecated rich-result hunting. The wins come from the four types above, implemented carefully, on pages worth citing. The schema is the wrapper. The content is still the product.
— The owner, PHX Search Co.
