SEO & AI search
We build for both the results page and the answer engine, measure them separately, and report which one produced the case. A page that ranks and is never quoted is failing at half its job. The method and the fee are on this page.
Build for extraction, not only for ranking
What an answer engine can actually quote
Ranking and being quoted are different problems. A results page rewards a document that covers a topic thoroughly. An answer engine rewards a passage it can lift without editing, attribute cleanly, and stand behind. Those pull in opposite directions more often than people expect.
A page that opens with three paragraphs of framing before it answers the question it titled itself with is unquotable, however well it ranks. A page that answers in the first sentence, attributes the claim, and carries the date is quotable even if it is short. The practical change is small and almost nobody makes it: put the answer first, then the reasoning, then the caveats.
Three things make a passage extractable. It states a fact rather than a position. It carries its own source, so the engine does not have to guess where the number came from. And it survives being read on its own, without the paragraph above it. We write to that standard and we report which pages get cited, separately from which pages rank, because a page that ranks and is never quoted is doing half its job.
page does not change its meaning. This site is built that way; its own llms.txt is public.| Requirement | Why it matters to an answer engine |
|---|---|
| A direct definitional answer in the first paragraph, before any framing | Extractable as a standalone response to the query |
| Facts written so a paragraph can be lifted without losing meaning | Synthesis breaks context; self-containment survives it |
Real <table> markup with headers, not a visual grid of divs | Tabular data is parsed as tabular data |
| Accurate Article, FAQPage, HowTo, Dataset schema, never stuffed | Inaccurate schema is worse than none; it teaches the engine to distrust the domain |
| A named author, a reviewer where relevant, and a real last-reviewed date | Accountability signals, and in health-and-legal topics they carry weight |
| An llms.txt summarising the site and its standing caveats | A machine-readable statement of what the site does and does not claim |
What we measure, from day one
AI-referral traffic is instrumented before the first page ships: ChatGPT, Perplexity, Claude, Gemini and Copilot segmented as named sources rather than dumped into direct or referral. Firms that add this instrumentation later cannot tell whether the channel grew or their measurement did.
| Channel | Leading indicator | Lagging indicator |
|---|---|---|
| Classic organic | Non-branded impressions and clicks on target clusters | Signed cases attributed to organic |
| Answer engines | Citations and referral sessions by engine | Signed cases with an AI-assistant touchpoint |
| Brand | Branded query volume | Direct inquiries, a lagging measure of everything else |
Content that earns its place
We publish to a standard, not a calendar. A page ships when it does something the current best result does not, when every statistic in it links to primary material, and when a named author and, where relevant, a named reviewer stand behind it. If that yields two pages in a month, we publish two. The editorial standard we hold ourselves to is the one we apply to client work, and the full playbook is published free.
What we report
| Metric | Role | Reported as |
|---|---|---|
| Signed cases attributed to organic | Primary | Headline. |
| Citations in AI answers, by engine | Primary | Tracked as its own channel from day one, not folded into "organic". |
| Referral sessions from AI assistants | Primary | Segmented by engine in analytics before the first article ships. |
| Non-branded organic inquiries | Diagnostic | Branded traffic is a lagging measure of everything else you do. |
| Rankings | Operational | Appendix. A ranking is not an outcome. |
| Phase | Length | What happens | What you get |
|---|---|---|---|
| 0 · Diagnostic | 3 weeks | Market sizing from our data, audit of current spend against cost per signed case, screening-capacity model | A written plan and a decision. Credited in full against the first three months of any retainer. |
| 1 · Instrumentation | Weeks 1–4 | Tracking rebuilt so every inquiry carries its source through to the case management system | A funnel you can audit. Nothing is optimised before it can be measured. |
| 2 · Build | Weeks 3–10 | Channel build, creative, landing pages, intake scripts and screening rubric | Assets you own outright, in your accounts, under your name. |
| 3 · Operate | Ongoing | Weekly optimisation against cost per signed case, monthly against cost per case surviving review | A monthly report that leads with signed cases, not impressions. |
When this is the wrong service for you
If you need cases this quarter, this is the wrong line to start with. Organic and answer-engine work compounds over quarters, not weeks. Start with paid search for flow and run this alongside it. Equally, if your firm will not put a named attorney’s byline and credentials on published pages, do not buy content from anyone, unattributed content in a health-and-legal category is the weakest asset you can own.
Related
Paid search & Local Services Ads
The most expensive inventory in legal, bought against a signed-case target rather than a click target.
Paid social
Demand creation for a population that does not know a claim may still be live. Handled without exploiting grief.
Websites & landing pages
The landing pages and forms, built for a parent reading on a phone at 2am and for a partner auditing the funnel.
What this costs → · Book a call → · Read the free library first →
Authorship & review
- Author
- Birth Injury Marketing. About the company.
- Last reviewed
- Legal review
- Not yet assigned. This page describes marketing practice, not law, and gives no legal advice. Rules cited are linked to the primary source so you can read them directly. See Editorial Standards.
- Corrections
- Found an error? Tell us and it goes in the dated corrections log, whether or not it flatters us.