In about two years, asking an AI assistant “who’s the best lawyer for…” went from novelty to habit. The firms those assistants name by name are winning consultations their competitors never hear about, and most law firms cannot see whether they are among the named or the invisible. This piece is for solo and small-firm owners and the people who run their marketing. It explains what generative engine optimization is, how the engines decide whom to cite, where a firm can realistically win, and why acting early builds a lead that is genuinely hard to reverse — while being honest that a lead is not a lock.
What changed: AI answering went mainstream
The shift is not a forecast; it already happened. ChatGPT passed one billion weekly active users in 2026 (OpenAI, reported July 2026). Google’s AI Overviews — the AI answer that sits at the top of a search — reach about 2.5 billion people a month, and Google’s AI Mode passed one billion monthly users within a year of launch (Google, I/O 2026). The Gemini app crossed one billion monthly users in August 2026 (Google, Aug 2026). Adoption at this speed has no precedent in consumer technology.
Legal is not exempt. In iLawyerMarketing’s 2026 survey of 1,110 U.S. consumers, 41.9% said they would use ChatGPT to research which lawyer to hire — up from 28.1% a year earlier — and roughly one in ten would research a firm through AI alone, with no Google and no directories (iLawyerMarketing, 2026). The consumers adopting fastest are not the youngest; in that data, the 45-to-60 age group leads, and higher-income households adopt at the highest rates — the clients many firms most want are the ones most likely to be asking an assistant first.
This is a shift, not a collapse. Google still handles far more volume than every AI assistant combined, and most people who see an AI answer never click through to anyone: when a Google AI Overview appears, users click a traditional result on about 8% of searches, versus 15% without one (Pew Research Center, 2025). The click your website has always depended on is being absorbed into the answer. The firm named inside that answer gets the exposure whether or not anyone clicks.
What GEO actually is
Generative Engine Optimization (GEO) is the work of becoming the firm an AI assistant names when a prospective client asks it a question — across a panel of surfaces, not one. At Legal Torch the panel is six: ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Gemini, and Claude. Each answers differently, so visibility is measured per engine, several times per question, rather than from a single screenshot that proves nothing.
GEO overlaps with classic search-engine optimization but is not the same job. SEO works to rank a page a searcher then clicks. GEO works to make a firm the cited answer an engine assembles for the user — which means the target is not a ranking position but whether the model retrieves and repeats your firm as the response. A strong Google ranking still helps, and it modestly improves the odds of being cited, but ranking no longer decides citation the way it once did.
How AI engines decide which firm to name
The engines do not invent names. They assemble answers from sources they can retrieve and corroborate, and a consistent set of signals decides who makes the shortlist.
The first is the directories the engines lean on — Avvo, Justia, Super Lawyers, Expertise, Yelp, and the Google Business Profile. In our audit work, firms that appear across these directories are markedly more likely to be named than firms that do not. The second is reviews: their count, pace, and recency. Review volume is one of the strongest signals separating the firms an engine names from the ones it skips, and it is a signal that accumulates over time rather than one that can be bought. The third is structured data and dedicated pages — clean schema markup and real practice-area and city pages that answer specific client questions, which give the engines something concrete to cite. The fourth is entity consistency: the same firm name, address, and credentials repeated across the web, so the engine can be confident two mentions are the same firm. The last is substantiated real-world authority — verifiable credentials, years in practice, languages served, matters handled — the facts the engines repeat back when they explain why they named a firm.
Two honest caveats belong here. Google AI Overviews and AI Mode often cannot be captured by automated tools and have to be checked by a person in a normal browser, which is why the reliable method is a human cross-check, not a dashboard. And engines vary by phrasing and by day — the same question asked two ways can name two different firms — which is exactly why one reading is never enough.
Where a firm can actually win: pick a niche you can measure
The instinct is to chase the biggest term — “bankruptcy attorney,” “personal injury lawyer.” That is usually the wrong target. The head terms are the most crowded, the most review-gated, and often the least winnable for a smaller firm; the winnable ground is a specific practice area, in a specific community or language, in a specific metro.
Sizing that ground takes care, because keyword tools can mislead. Google suppresses metro-level search volume for sensitive categories — anything containing the word “bankruptcy” returns null at the metro level, which reads as “no demand” but is actually withheld demand (Google Ads financial-services policy). A firm that sized this lane on a keyword tool would conclude, wrongly, that no one is searching. The honest approach is to measure such lanes through the AI panel and rank tracking, and to size demand on the specific-intent pool — the “near me,” probate, and trust-administration queries that actually convert — rather than the informational head term that looks large but rarely hires.
Choosing the lane is also a strategy decision, not just a data one. A practice area with volatile, economy-driven demand is a weaker foundation than one with steady, non-cyclical demand; a lane already crowded with review-heavy incumbents is a harder win than an adjacent one no one has claimed. The strongest target is the intersection: real, measured demand, lower competition, and a claim the firm can substantiate.
The moat: a durable lead, not a permanent lock
Here is the part worth sitting with, because it is the reason timing matters. Getting visible faster and deeper than your competitors builds a compounding advantage — one that gets harder to unseat the longer you hold it. It is a durable lead, not a permanent lock, and being honest about both halves is how you set the right expectations and invest where it actually pays.
Four advantages compound and cannot be shortcut with money alone:
- Reviews. Count, pace, and recency build up over time, cannot be bought, and there are ethical limits on how fast anyone can gather them. A year’s head start compounds into a lead a latecomer cannot simply sprint past.
- Depth of authority. A deep body of practice-area content, clean structured data, and standing in the directories the engines trust. A competitor starting later has to rebuild the entire corpus, not just one page.
- Being the named answer. Once assistants consistently return you for a question, the citations reinforce themselves — being named produces more corroborating signals, which keeps you named.
- Time-only assets. Domain history and community reputation are earned over years, not purchased — an advantage simply unavailable to a new arrival.
And four honest limits, because a lead is not a guarantee:
- The copyable layer is catchable. Page content and markup can, with sustained effort, be matched by a determined, well-funded competitor. The durable part is the reviews, credentials, and history they cannot copy — not the pages they can.
- AI changes on its own. Engines update how they retrieve and cite. A position can move because the technology shifted, not because a rival outworked you.
- No one owns the rankings. There is no legal exclusivity over search results. A visibility lead is earned and defended, never owned outright.
- It needs upkeep. Reviews go stale and content ages. A lead that is maintained stays a lead; a lead that is neglected erodes.
The strategic reading is straightforward: front-load the advantages a competitor cannot shortcut — reviews, depth, and becoming the named answer — and keep them current. The most durable moat is not a promise about rankings; it is the incumbency built into your own name, which holds even as the technology and the field keep moving. And the case for moving now is sharpest precisely where a field is still soft: first-mover depth pays the most in a low-competition niche, before other firms arrive.
What this looks like in practice
Consider a North Orange County bankruptcy firm — anonymized here — that assumed it was invisible in AI search. A six-surface panel showed the opposite. On its core local queries, the assistants already named the firm, usually in the top three to five, and on a language-specific niche it had long served, it was usually named first across most engines. The foundation was sound.
The real gaps were narrower and more useful than “you’re invisible.” Outside its niche the firm was usually the third-to-fifth name rather than the first, and in the broader county-wide market it was absent across every engine — that lane belonged to firms with hundreds of reviews and a certified-specialist credential, a review moat that is a twelve-to-eighteen-month build, not a quick win. So the plan was not to fight the hardest, lowest-value fight. It was to measure honestly, defend the cells the firm already owned, and grow into an adjacent lane with real demand, lower competition, and steadier, non-cyclical economics — the lane a keyword tool alone would have hidden. Measure, defend, then grow where the ground is soft.
How to start, and what honest success looks like
Start by measuring. You cannot manage a surface you cannot see, and because AI answers move by engine, day, and phrasing, the only honest baseline is a panel run several times per question, depersonalized, with the evidence saved. That tells you where you actually stand — named or not, first or fifth, present in your niche but absent next door.
Then invest where the advantage compounds: the reviews, the authority corpus, and the named-answer entrenchment a later entrant cannot simply buy. Expect a durable, compounding lead, not a guaranteed ranking — no honest agency can promise the latter, and the engines and rules both forbid it. What you can expect is to know where you stand, to move from also-ran to a name the assistants return, and to hold that position as long as you keep investing in it. The firms that start now, while their niches are still open, are the ones that will be the answer when their competitors finally ask the question.
See where your firm stands — the free AI Visibility Audit →
Sources
Statistics verified against primary and authoritative sources the week of August 25–29, 2026. AI-adoption figures move quickly; re-verify before republication.
- OpenAI — ChatGPT tops 1B weekly active users (reported July 2026): link
- Google — AI Overviews (~2.5B/mo, Pichai at I/O 2026) and AI Mode (1B/mo): link
- Google — Gemini app 1B monthly users (official), August 11 2026: link
- Pew Research Center — click behavior with AI Overviews (8% vs 15%), 2025: link
- iLawyerMarketing — 2026 consumer survey (41.9% would use ChatGPT; n=1,110): link
- Google Ads Help — restricted financial-services categories (metro volume suppression context): link
