AI search for BigLaw

Institutional buyers use AI systems the way they used to use a colleague with a long memory: to produce a shortlist of firms for a matter that has just landed. Joe Ashta studies those shortlists as a GEO / AEO consultant for law firms. This page covers what general counsel and procurement teams ask, how those queries differ from consumer ones, what the published leaderboard shows in general terms, which pages these systems cite, and what the leaderboard does not measure.

The buyers and their questions

The questions are credential questions and trigger-event questions. Which firms have former regulators in their FCPA practice. Who handles cross-border M&A in Toronto and New York. Which firms have defended a Section 337 investigation. Who should be on a panel for employment work across four jurisdictions. Which firm advised the other side on a deal like the one now on the desk. Behind each is an event: an investigation opened, a transaction signed, a regulator's letter received, a panel review due. The asker is a general counsel, a deputy, or a procurement lead building a list for a committee, and the purpose is to narrow a market of hundreds to a handful worth calling.

How institutional queries differ

Three differences change the work. There is no local pack and no proximity signal, because a buyer selecting counsel for a cross-border matter cares about the credential rather than the distance. The grammar of the query is practice area plus credential plus jurisdiction, in the profession's own vocabulary, so the system matches against specialist language rather than plain description. And the output is a shortlist meant to be compared, often with a reason per firm, rather than one recommended name. The practical effect is that a large firm competes on the specificity and the corroboration of its published record, practice by practice, against peers of the same size.

What the leaderboard shows

Joe Ashta publishes the BigLaw GEO Leaderboard, which ranks 118 US law firms by AI search citations reported by Ahrefs across ChatGPT, Gemini, Google AI Mode and AI Overviews, refreshed monthly. The firms are grouped into four leaderboard segments: Big Law, Global, Mid-Law and Boutique. Its general finding is that citation counts vary widely between firms of similar size and standing. Firms with comparable revenue, headcount and reputation are cited at very different rates, so size does not settle whether AI systems read a firm's pages. Domain Rating is shown beside each firm for context, and it does not explain the variation on its own.

The pages these systems cite

Large firms publish more than almost anyone, and three kinds of page carry most of these answers.

• Partner biographies. Credential questions are answered by people. A biography stating a government or regulatory role, the matters handled, the bar admissions and the industries served is directly responsive to "which firms have former regulators in this practice". A biography listing only education and a practice group answers nothing.
• Client alerts. Dated, specific, written in the buyer's vocabulary and published in volume when a regulation changes, these are the pages most often reached for on a trigger event. They carry the firm's name into an answer about the event.
• Practice and industry pages. These establish that the firm does the work at all, and connect the lawyers and the alerts into one claim of capability.

What those source links prove, and what they do not, is set out on AI citations. The same firm can be read very differently by a summary above the search results and by a conversational session, which is the distinction between AI Overviews and AI Mode.

What the leaderboard does not measure

A citation proves a platform read a page. It does not prove the platform recommended the firm, and the leaderboard makes no claim about which firms a buyer is told to call. It also says nothing about any firm's legal ability or the outcomes it achieves for clients. It reports vendor data from one source, which is the difference between vendor-reported and measured visibility. Reading it as a quality ranking of law firms would be a mistake. It ranks how often a firm's pages appear as sources in AI answers.

What a large firm can measure and fix

Work practice by practice rather than firm-wide. Take the credential and trigger-event questions a buyer would ask for one practice, in the jurisdictions it covers, and establish which firms the platforms name and cite for them today. Then fix what the answers expose: biographies that hide the credential the buyer asks about, alerts published without being retrievable, practice pages that describe the group without describing the work, and lawyer records that disagree between the firm site, the directories and the bar. Report it as a mention rate for the practice. Measuring that and fixing what it exposes for such a firm is what Joe Ashta does as a GEO / AEO consultant, and SEO and Google Ads engagements run through NearMe Marketing. The overview is Law firm AI search, the role is GEO consultant for law firms, and the canonical facts are on the entity record.

Related: AI search for criminal defence firms, AI search for personal injury firms (coming soon) and AI search for family law firms.