Spend enough time in Google Search Console across legal properties and a pattern emerges that has nothing to do with how people search. The queries are too long, too structured, and too consistent: a company name followed by a comma-separated liability checklist — litigation, lawsuits, settlements, court rulings, legal liabilities. An instruction phrased as a task: evaluate this legal company on a specific mass tort. Research strings stacked with a dozen site-exclusion operators, the kind of query syntax no one thumbs into a phone. They arrive in clusters, they repeat with small variations, and the pages they surface often sit at position one for queries no human ever composed.
These are machine queries. They are the visible retrieval step of AI assistants doing homework — scanning a company for litigation exposure before summarizing it, checking whether a legal information brand is legitimate before citing or recommending it, assembling a source set with exclusions already applied. The assistant's answer is downstream; what Search Console captures is the due diligence.
What the machines are asking
Three query families keep recurring. The first is the entity liability scan: a company name plus a standardized string of legal-exposure terms, run the same way across different companies — the signature of an agent template, not a person. The second is the evaluation prompt: an instruction to assess a named organization on a named topic, which reads like a fragment of a larger workflow that escaped into the search box. The third is the operator-heavy research query, prebuilt to exclude social platforms and forums, optimizing for primary and editorial sources.
Two things about these queries should get any managing partner's attention. They rank instantly — the pages answering them face almost no competition, because nobody writes for queries nobody types. And they are evaluative: the assistant is not looking for a phone number, it is deciding whether an entity is credible enough to put in front of a user. Whatever page wins that query becomes the machine's evidence file.
The reference check has no one picking up
Here is the gap. When a machine asks whether a legal brand is legitimate, the honest answer lives in fragments — a bar number on one page, an editorial policy on another, press citations on a third, and a corporate name that appears in four slightly different forms across four properties. An assistant assembling a legitimacy judgment from that gets a shrug. The fix is not clever; it is editorial hygiene treated as infrastructure. A page that states plainly who runs the operation, under what license, with what disclosures. Organization and Person schema that agree with each other everywhere the entity appears, with the same identifiers and the same connected profiles. Press citations marked up so a machine can verify that third parties have vouched for the entity. Naming consistency, so the brand the machine evaluates is the brand you actually operate.
None of this is speculative on the traffic side. We have followed assistant-referred visitors from the citation, through the landing page, into a completed intake form with full attribution intact. The due-diligence query is what happens before that referral gets made. It is the reference check — and right now, across most of the legal web, nobody is answering the phone.
What to do about it
Treat the evaluation queries as a content brief. If assistants ask whether your operation is legitimate, publish the page that answers in the first hundred words. If they scan companies for litigation exposure, your coverage of those companies should be precise, dated, and sourced, because it will be quoted without you in the room. And audit your entity graph the way you would audit a citation in a brief: every property, one name, one set of identifiers, no contradictions. The assistants are already reading. The only question is whether what they find was written on purpose. This publication's editorial standards page exists for exactly that reader — the one that isn't human.