AI search is fragmenting. What law firms should do about it.

When I wrote about Google AI Overviews erasing law firm SEO last year, the prescription was straightforward: optimize content for AI engines, not just Google. The premise was that ChatGPT was eating the top-of-funnel queries and firms needed to show up there. That premise was correct. It's also already out of date — because the AI search market that existed when I wrote that post is not the market that exists today.

In the past twelve months, the AI search ecosystem went from a near-monopoly to a four-way contest. ChatGPT's referral share dropped from 89.1% in October 2025 to 62.6% in March–April 2026. Gemini climbed from 2.31% to 8.65%. Claude jumped from 0.30% to nearly 3%. Perplexity is up to 7.07%. The market has fragmented faster than any analyst predicted. For law firms that built their AI visibility strategy around "show up in ChatGPT," that strategy is now optimizing for a shrinking — though still dominant — channel.

This post is about what changed, why it matters specifically for law firm marketing, and what the practical fix is.

The numbers, briefly.

I'll keep this short because there's not much value in restating reports you can read elsewhere. The headline numbers, drawn from the recent Statcounter, Goodie, and BrightEdge analyses:

• ChatGPT: referral share fell from 89.1% (Oct 2025) to ~62.6% (Mar–Apr 2026). Still dominant. Still the largest single source. But the dominance is eroding monthly.
• Gemini: climbed to 8.65% by March 2026, claiming the #2 spot from Perplexity. The biggest jump came in December 2025 to January 2026, after Gemini 3 Flash became the default model across Google's AI-powered surfaces.
• Perplexity: ~7.07%. Lost the #2 position to Gemini in early 2026 but holds the highest reputation among publishers for citation quality.
• Claude: jumped from 0.30% in April 2025 to roughly 2.9% in March 2026 — a 10× increase year-over-year, with a notable surge in week 12 of 2026 driven by a news-cycle wave of users switching from ChatGPT.
• Copilot: holding around 4%, mostly enterprise-adjacent.

The total AI referral footprint is still small in absolute terms — Chartbeat estimates AI chatbots account for less than 1% of publisher page view referrals overall — but it's growing extremely fast (Adobe data shows a 357% increase year-over-year), and the conversion data tells a different story than the volume data. AI-referred users convert roughly twice as well as standard organic search traffic. Claude specifically generates a 16.8% conversion rate; ChatGPT 14–16%; Perplexity 10.5%; Gemini 3%; against a Google organic baseline of 1.76–2.8%. For law firms, where one signed retainer pays for a quarter of marketing spend, the conversion quality of AI traffic matters far more than the headline volume.

Why this matters specifically for law firms.

Two reasons. First, the people consulting AI engines about legal questions are disproportionately high-value prospects. Someone who asks ChatGPT "do I have a personal injury case" is much earlier in the buying journey than someone who Googles "personal injury lawyer near me," but they're also pre-qualifying themselves in a way the cold paid-search click doesn't. By the time they click a citation in an AI engine, they've already absorbed contextual information about their situation and are warmer than the equivalent organic-search visitor.

Second, the engines disagree about which sources to cite. Research analyzing 83,670 AI citations across ChatGPT, Claude, and Perplexity found that the engines overlap on fewer sources than you'd expect — each has its own preference for source authority, citation patterns, and content depth. A firm that's been optimized for ChatGPT specifically is almost certainly underperforming on Gemini and Claude, and likely invisible on Perplexity. That used to be tolerable because ChatGPT had 89% of the share. At 62%, the gaps are getting expensive.

A firm that's been optimized for ChatGPT specifically is almost certainly underperforming on Gemini and Claude, and likely invisible on Perplexity. That used to be tolerable. It isn't anymore.

What each engine actually rewards.

Generalizing across the recent citation-analysis research, the engines have meaningfully different content preferences:

ChatGPT.

Still favors authoritative, structured content with clear extractable lead paragraphs (40–60 words). One-third of ChatGPT citations originate from deep pages (subdirectory content, not just home pages), which means specific topic-cluster pages and long-form articles outperform top-level marketing pages.

Gemini.

Leans heavily on Google's own index, which means content that already ranks well in traditional Google search will also tend to surface in Gemini citations. The upshot: traditional SEO investment compounds into Gemini visibility — but only if the SEO is current and not from the 2019 page-one playbook.

Perplexity.

Citation-first by design. Rewards recent, well-sourced content that cites its own sources clearly. Favors quotable, citable sentences embedded in clear structure. Perplexity has the widest sourcing net of any engine — domain diversity is high — which means specialized publishers, niche research pages, and local-market expertise have more entry points to win citations.

Claude.

Sends minimal direct referral traffic (it's not built primarily as a search engine), but its influence on professional-context decision making is rising. If your firm shows up well in Claude responses, the audience is small but the buying signal is strong. Claude's crawl footprint — by some measures 500,000:1 against its referral footprint — means it's reading content at scale regardless of whether it sends clicks. Blocking ClaudeBot is, in 2026, a defensive own-goal.

The four-channel optimization framework.

With that backdrop, here's the practical approach I'd take for a law firm's AI-visibility strategy in 2026. This is a re-architecting of the recommendations in the original AI Overviews post, updated for the fragmented market we now live in.

1. Publish for the four-engine baseline, not for ChatGPT specifically.

Every substantive piece of content the firm publishes should pass four tests: (a) clear lead paragraph with the topic's claim stated in 40–60 words (ChatGPT preference), (b) cited sources with linked attribution (Perplexity preference), (c) structured headers and Q&A format (Gemini preference, also helps Google), (d) genuine authority signals — named author, credentials, dated publication, real evidence (Claude preference). Doing all four for every piece is more work than optimizing for one — but the alternative is having visibility on one shrinking channel while three competitor channels grow without you.

2. Diversify the AI traffic mix toward 35% non-ChatGPT.

Goodie's recent analysis suggests publishers should target ≥35% of AI referrals from non-ChatGPT sources by Q3 2026. For law firms, that means deliberately building content that surfaces in Gemini, Perplexity, and Claude — not just hoping. A monthly review of your firm's AI citation footprint across all four engines should now be part of the marketing cadence.

3. Confirm AI crawlers are allowed.

I see law firms unintentionally blocking the AI crawlers in their robots.txt — sometimes because an SEO vendor added "block all bots" without thinking, sometimes because old defensive postures from 2024 about "AI scraping our content" never got reversed. If GPTBot, PerplexityBot, ClaudeBot, and Google-Extended aren't explicitly welcomed in your robots.txt, your firm is invisible to AI by choice. First fix to make before any other AI-visibility work.

4. Build a quarterly AI-visibility audit.

Pick 8–12 representative prospect queries in your practice area and your geography. Run them across ChatGPT, Gemini, Perplexity, and Claude. Document where your firm shows up, where competitors show up, and which sources are getting cited. This is the closest thing to a Search Console for AI engines that currently exists — you have to do it manually, but doing it quarterly tells you whether your investments are working and where the gaps are.

5. Treat attribution differently.

AI traffic doesn't show up cleanly in GA4. Most AI referrals get bucketed as "direct" because the engines strip referrers. Goodie's recent analysis suggests that even a conservative 5% reattribution from direct to AI more than doubles measurable AI share. Don't make budget decisions on the GA4 numbers alone — they understate AI by a wide margin. Build a separate AI-visibility scorecard alongside the traditional attribution stack.

The trajectory.

If Gemini's 47% monthly growth continues — and there's no obvious reason it won't, with Gemini 3 Flash now the default across Google's surfaces — the competitive landscape will look meaningfully different by late 2026. ChatGPT will probably still be the largest channel; Gemini will probably be close to half its share; Claude and Perplexity will keep growing in conversion quality even if not in raw volume. The firms that build for the multi-engine future starting now will look smart in eighteen months. The firms still optimizing for "AI" as if it's one channel will be playing catch-up on three.

The single-platform AI strategy was a 2024 strategy. We're in a different market now.

---

The diagnostic includes a multi-engine AI visibility audit against your top prospect queries — and a 90-day plan to close the gaps. Start there →