AI Visibility for Law Firms: What We Found Auditing the Towson Legal Market

We picked law firms for a specific reason.

Legal services is one of the highest-intent, highest-trust categories in local search. When someone asks an AI to recommend an attorney, they're not browsing — they're looking for someone to handle something important. The stakes of that recommendation are real. And yet when we scored Towson-area law firms for AI visibility, the results were stark enough that we built an agency around what we found.

What We Measured

The MAKIF Audit Framework evaluates 46 signals across five layers: Trust, Entity, Retrieval, Citation, and Measurement. Each layer addresses a distinct stage in how AI systems decide whether to find, select, and cite a source. We applied this framework to a sample of law firms across the Baltimore–Towson corridor.

We won't name individual firms. But we'll share what the data showed, because the patterns were consistent enough to be meaningful.

What We Found: The Entity Layer Was the Most Common Failure

Across the firms we scored, the Entity layer was the most consistently weak. Most had unclaimed or minimally completed Google Business Profiles. Many had schema markup either absent or incorrectly implemented — LocalBusiness markup without the required attorney-specific type declarations, or Organization markup that conflated the firm entity with individual attorney entities.

This matters because Google's Knowledge Graph builds entity confidence from cross-web consistency. A law firm whose name appears four different ways across its website, Google Maps listing, Yelp page, and legal directory listings is, from the AI's perspective, potentially four different businesses. That ambiguity is enough to exclude it from AI recommendations entirely.

The fix is specific: claim and verify the Business Profile, implement correct LocalBusiness and Attorney schema markup, audit all third-party directory listings for name/address/phone consistency, and establish cross-web entity signals through platforms like Avvo, Justia, and FindLaw — the legal directories AI systems actively draw from when answering "recommend an attorney" queries.

The Citation Layer: Where Almost Everyone Failed

No firm we scored had meaningful citation triggers in their content. Virtually all law firm websites follow the same content pattern: practice area pages with broad descriptions of services, attorney bios with credentials, and a contact form.

None of that content structure is what AI systems treat as authoritative. According to the KDD '24 GEO research from Columbia and Princeton Universities, the content patterns that drive AI citation rates are: embedded statistics with sources, direct citations from credible references, and content structured to answer specific questions. For Law & Government queries specifically, Statistics Addition was the highest-performing GEO method tested.

A law firm content page that says "we have over 20 years of experience and fight hard for our clients" provides no citation triggers. A page that says "Maryland injury victims who retain legal counsel receive settlements averaging 3.5x higher than those who negotiate alone, according to the Insurance Research Council" — with that statistic sourced, linked, and contextualized — is a completely different signal to an AI system.

Most law firm content is written to sound credible to a human reading it. It needs to be written to signal credibility to a system evaluating it.

The Retrieval Layer: A Technical Gap Most Firms Don't Know They Have

Several firms we reviewed had JavaScript-heavy websites — marketing agency templates that load content dynamically. AI crawlers, unlike Google's full rendering engine, often cannot execute JavaScript. Content that only appears after JavaScript executes is invisible to these crawlers.

This is a silent failure mode. The firm's website looks fine in a browser. It may even rank on Google. But for the AI crawlers feeding systems like ChatGPT and Perplexity — the content isn't there.

The fix requires either server-side rendering, static HTML fallbacks, or at minimum an llms.txt file that directs AI crawlers to the content they can access. None of the firms we scored had implemented llms.txt.

The Pattern: Structured for Yesterday's Search

The consistent finding across every firm we evaluated was the same: websites optimized for how search worked ten years ago, in a world where AI search didn't exist.

Generative engines don't reward keyword density. They reward cited authority, entity clarity, and direct-answer content structure. The KDD '24 research demonstrated this on 10,000 queries across multiple AI systems — keyword stuffing performed worse than doing nothing, while citation and statistics signals produced 30–40% visibility improvements.

What a 0–100 Score Looks Like in Practice

For a typical Towson-area law firm in our pilot, the scores looked roughly like:

*Composite score: 7–15 out of 100.*

Not zero — but functionally invisible. An AI system encountering that signal profile has no strong reason to choose this firm over a nationally-known legal resource that happens to mention the same location.

The Opportunity

The Towson legal market has a specific structural advantage right now: most firms haven't started. The first law firms to establish strong AI visibility signals in this market will capture recommendation real estate that's currently empty — not competitive, just empty.

That window closes as awareness grows and other firms begin optimizing. The MAKIF Audit is the starting point: a scored baseline, a competitor comparison, and a prioritized action list. Everything that comes after that is execution.

Sources: KDD '24 GEO Research (Aggarwal et al., Columbia & Princeton Universities) · Google Search Central — Local Business Structured Data · Google Knowledge Graph API Documentation · RAG Survey, Gao et al. (arXiv 2312.10997) · MAKIF Audit Data, Towson MD Pilot, 2025