Most law firms still think of AI search as an emerging concern — something to worry about in 2027 or 2028. That assumption is incorrect. AI Overviews already appear on roughly 70% of legal queries. ChatGPT processes over 800 million conversations a week, many of them about local professional services. The question is no longer whether prospects are using AI to find firms. The question is which firms AI is recommending.
To find out, we ran the same query against ChatGPT, Perplexity, and Google's Gemini AI Overview: 'I was in a car accident in Grand Rapids and need to find a personal injury lawyer. Who should I consider?' The three responses revealed more about the state of AI legal search than any whitepaper could.
Before sharing what happened, a methodology note. The query was issued from a fresh browser session with no logged-in account, no location personalization beyond the implicit 'Grand Rapids' in the prompt, and identical phrasing across all three systems. The results were captured on the same afternoon to control for indexing recency. The goal was not to evaluate any specific firm's marketing — it was to see what mechanisms AI systems use when they recommend.
What ChatGPT did
ChatGPT (running on the GPT-5 model layer) returned a structured response with four firm recommendations, each with a one-paragraph summary describing the firm's practice areas, notable case results where available, and a citation to the source. The citations clustered around three sources: each firm's own website (especially the about and practice-area pages), legal directories (Avvo, Justia, Martindale-Hubbell), and regional news coverage where firms had been quoted or covered.
What was telling was what ChatGPT did not surface. Firms whose websites lacked clear practice-area definition were absent, even when their Google rankings were strong. Firms whose attorney bios were thin or generic were ranked lower in the recommendation list. The firms ChatGPT confidently recommended all had three things in common: detailed practice-area pages, deep attorney profiles with credentials and case examples, and strong third-party validation through directory presence and news coverage.
What Perplexity did
Perplexity returned a more aggressive synthesis. Its response opened with a direct recommendation of one firm by name — with reasoning — and then listed three alternatives. Perplexity's citation panel was visible alongside the response, which let us trace exactly which sources informed each ranking. Approximately 60% of citations were to firm websites, 25% to legal directories, and 15% to news coverage and law-related industry sites.
Notably, Perplexity was the only system that explicitly weighted recent content. A firm with fresh 2025 and 2026 articles on personal injury topics in Michigan was placed first. Firms with stale blog content from 2021–2023 were placed lower, even when their core directory presence was stronger. Perplexity's emphasis on recency is consistent with its broader product positioning around real-time research, and it has direct implications for legal content strategy.
What Gemini AI Overview did
Google's Gemini AI Overview returned the most conservative response. Rather than naming firms directly, it surfaced a general framework for choosing a personal injury attorney in Michigan — what to look for, what questions to ask, what credentials matter — with sources cited to legal education and consumer-protection content rather than to specific firms.
The named-firm recommendations appeared below the AI Overview, in the traditional results section, where the local pack and organic listings still operate on classic local SEO signals. The lesson here is that Google's AI layer is treating legal services with extra caution, surfacing frameworks rather than direct firm endorsements. Firms hoping to appear in the AI Overview itself need to publish authoritative, educational content that the framework can cite — not just service-oriented marketing content.
The four patterns across all three systems
Despite their different mechanisms, four consistent patterns emerged. The first pattern: firms with deep, current practice-area content were recommended across all three systems. Thin practice-area pages — common in the industry — translate to invisibility in AI search. The second pattern: attorney profile depth mattered everywhere. Firms whose attorney bios listed credentials, case examples, professional affiliations, and specialty focus appeared more confidently than firms with one-paragraph bios.
The third pattern: legal directory presence is still doing real work. Even in the AI era, Avvo, Justia, and Martindale-Hubbell are foundational citation sources for all three AI systems. Firms treating directory profiles as set-and-forget projects are giving up significant ground. The fourth pattern: third-party validation in news coverage, regional business publications, and respected legal commentary sites consistently lifted firms above peers with similar website quality. Earning those mentions is unglamorous, sustained PR work — but the AI search return is measurable.
What this means for Grand Rapids law firms
The firms that appeared in all three AI systems were not the firms with the biggest marketing budgets. They were the firms whose digital footprint was complete: detailed practice pages, deep attorney bios, current content, strong directory presence, and third-party mentions. None of these elements is exotic. None of them requires an enterprise platform. What they require is the discipline to treat every one of them as primary infrastructure rather than an afterthought.
Most firms underinvest in two or three of the five and assume the others will compensate. AI systems are not making that compensation. They are looking for completeness, and the firms that are complete are the firms being recommended. The rest are watching a market they cannot see decide their pipeline.
We re-ran the same test two weeks later to check stability. Two of the three systems returned substantially the same firm recommendations. Perplexity rotated in a new firm at the second position — one that had published a fresh blog post on Michigan no-fault auto reform during the intervening period. This single data point suggests that publishing cadence directly influences Perplexity's ranking, which makes regular content production a higher-leverage tactic than many firms assume.
One additional observation: none of the three AI systems surfaced firms based on paid search advertising. AI recommendation is downstream of organic and entity signals. Firms spending heavily on Google Ads were no more likely to be AI-recommended than firms with no paid presence at all. This is consistent with AI systems' general policy of separating advertising from recommendation, but the implication for firm marketing budgets is worth noting.
A separate observation worth flagging: the three systems disagreed more on the second and third positions than on the first. The 'best' firm tended to be obvious to all three because the firm's signal completeness was unusually high. The 'next-best' firms varied more, because each system weights its underlying source mix differently. This means the firms that appear consistently across all three systems are scoring well on multiple distinct signal types — directory, content, and third-party validation — not just maximizing one channel. Diversified entity strength is the only reliable way to be cited across AI platforms with different retrieval mechanics.
Testing ChatGPT, Perplexity, and Gemini on a personal injury lawyer query for Grand Rapids revealed four consistent patterns: deep practice-area content, comprehensive attorney profiles, active legal directory presence, and third-party validation through news coverage all drove recommendations across systems. Marketing budget and paid search spend did not. The firms appearing across all three AI systems shared a discipline of digital completeness, not a budget advantage.







.webp)