A growing share of buyer research never touches a results page. Someone asks ChatGPT which accounting software suits a contracting business, or asks Perplexity what a panel upgrade costs, and gets a synthesised answer built from three to six sources. No ten blue links, no scrolling, no comparison shopping.
Being one of those cited sources is a different problem from ranking first, and most businesses are not structured for it at all. This guide covers what actually determines citation, how it differs from classic SEO, and what to change.
A search engine orders pages. An answer engine does something different: it retrieves passages, judges whether each one answers the question, and synthesises a response from the best of them. The unit of competition is the passage, not the page.
This has a counterintuitive consequence. A page ranking eighth can be cited while the page ranking first is skipped entirely — because citation depends on whether a specific block of text cleanly answers the specific question, not on the page's overall authority.
Cited passages share a consistent shape. They answer in the first sentence, they are self-contained, and they are specific enough to be worth quoting.
Classic SEO optimises a page to outrank other pages; AI search optimisation structures passages so a model can lift one cleanly. The two overlap heavily — both reward genuine depth, clear structure and topical authority — but they diverge in three ways worth knowing.
First, position matters less. Being in the retrievable set matters more than being first. Second, freshness weighs differently: answer engines often prefer recently updated sources for anything time-sensitive. Third, brand mentions matter more than links — models build associations between entities and topics from how often and how clearly they co-occur, not only from hyperlinks.
It helps, but less than clear writing does. Structured data makes your entities and their relationships explicit, which reduces the work a model has to do to understand what your business is and what a page covers. FAQPage, Organization and Service markup are all worth implementing.
But schema is a clarifier, not a substitute. A page with perfect markup and buried answers loses to a page with no markup that answers the question in its first sentence.
An llms.txt file is a plain-language summary of your business placed at the root of your domain, written for AI crawlers rather than for people. It states what you do, who you serve, what you charge, and what your key pages are.
It is not an official standard and no engine guarantees it will be read. But it costs almost nothing, and it gives a model an unambiguous, self-authored description of your business rather than one inferred from scattered marketing copy.
This takes about an hour and tells you exactly where you stand.
Classic search is a mature, brutally competitive channel where new domains wait quarters for traction. AI answer engines are neither mature nor saturated — most businesses have not restructured anything, and many have never checked whether they are cited at all.
That gap will close. Right now it is one of the few genuinely under-contested opportunities in organic visibility.
Answer the question in the first sentence beneath a heading phrased the way a person would ask it, keep each section self-contained, and be specific — numbers, named conditions, ordered steps. Answer engines retrieve passages, so each passage must stand alone and answer completely.
They overlap but diverge in three ways: position matters less than being in the retrievable set, freshness weighs more heavily for time-sensitive topics, and brand mentions matter alongside links because models build entity-topic associations from co-occurrence.
Yes, but less than clear writing does. Schema makes entities and relationships explicit and reduces ambiguity. It does not compensate for answers buried three paragraphs into a section.
Tell us your industry and city. We will send back a short snapshot of the questions buyers in your market are typing, and who is currently answering them.
We deploy for one business per niche, per metro. Thirty minutes tells you whether yours is open.
Book a 30-minute call