How AI assistants are changing birth injury attorney discovery, and what to do about it without guessing.

What has actually changed

Three things are observable today without any speculation. First, the major assistants and Google's AI modes answer medical and legal-process questions directly and cite a small number of sources. Second, those sources skew toward pages that are specific, well structured, consistent with primary medical and legal authorities, and clear about uncertainty. Third, the conversation is multi-hop: the answer to the medical question leads to the preventability question, which leads to the process and attorney questions, often in the same session. A firm's content can be cited at any hop, but the preventability and process hops are where a family is deciding whether to talk to anyone.

What has not changed

Nobody has published reliable volumes for these conversations, and anyone quoting "AI search volume" for birth injury queries is estimating. Citation is not a ranking: the same question returns different sources on different days. And a citation is not a conversion: the family still has to visit, trust the page, and decide to call. The old rules still govern the result: accurate content, fast pages, clear disclosures, and intake that answers.

How to be cited accurately

• Answer one question per page, in the first screen. Assistants extract; a page that buries its answer under a thousand words of preamble is skipped for one that states it plainly and then supports it.
• Agree with the primary sources. Condition content that contradicts NIH, ACOG or AAP, or that presents a diagnosis as proof of negligence, is less likely to be cited and more likely to be cited as an example of what not to trust.
• Be explicit about uncertainty and jurisdiction. "In most states, with exceptions" is extractable and honest; a flat national rule is neither.
• Make the page verifiable. Named authorship and review where true, dated updates, sources linked, structured data that reflects the visible content, no hidden text aimed at crawlers.
• Keep crawlers allowed. Blocking AI crawlers in robots.txt removes you from the answer; it does not remove the question.

What not to do

Do not publish "AI-optimized" pages stuffed with question variants; they duplicate intent and are a thin-content problem under a new name. Do not inject prompt text or invisible instructions for assistants; it is detectable and it reads, when found, exactly as it is. Do not pay for "AI visibility" reports that cannot show you the question, platform, date and run count behind every number.

How to measure it without inventing numbers

A fixed set of questions a family actually asks, run on each platform three times a month, with the sources cited, whether your firm appears, and whether the answer was accurate, recorded with the date. Report the share of questions where you appear per platform and the sources that beat you. That is the whole method; it fits in a spreadsheet, and it is the only honest form of this measurement we know. We run it for clients as part of SEO and content, and we publish our own question sets so the method can be checked.