What citation engine optimisation is
Citation engine optimisation is the work of getting your pages cited by AI engines when those engines answer a buyer's question. ChatGPT, Gemini, Perplexity, Claude. When someone asks one of them which firm to call, which tool to use, which auditor to hire, the engine picks two or three names. The picked names get a visitor who has already been recommended. Everyone else gets nothing.
It is not SEO. SEO is the practice of appearing on a list of ten links a human chooses from. This is different. There is no list. There is a paragraph. The paragraph names sources. You are in it or you are not.
Why this is becoming the real question
A buyer with a problem used to open Google and run a query. They scrolled, they clicked, they compared. They were yours to win on the search results page or lose there. That is changing faster than most accounts are reporting it.
The same buyer now opens ChatGPT and asks the question in plain language. The engine names three vendors. The buyer opens those three in new tabs and never sees a Google results page because no query was ever run. The dashboards that measure your traffic do not show this. The visitor arrives looking like direct traffic, or sometimes like a referral from a tab that was open. The session shows up. The cause does not.
Recent research from Adobe Digital Insights puts AI-referred traffic growth above 1,200% year on year through 2025. Exposure Ninja's 2025 numbers find that 88% of AI engine citations come from URLs that are not in Google's top 10. The two pools are diverging. A business that wins SEO and loses citation share is on a slow timer most owners cannot yet see.
This shift is well past the early stage in B2B technology, professional services, medical specialties, and considered consumer purchases. It is slower in transactional local trades where people still call the first number they see. Every category is moving in the same direction at a different pace.
The five things that decide whether you get cited
From running citation audits across enough sites to see the pattern, five things consistently decide whether a page enters the citation pool for a given question. These are the same five the citation gap audit measures.
One. Crawl access. The AI engines crawl with their own bots. GPTBot, ClaudeBot, PerplexityBot, Google-Extended. A surprising number of WordPress sites have these blocked, usually by a security plugin doing what it thinks is its job. If the engine cannot crawl the page, the page cannot be cited. This is the first thing to check and the most common cause of zero visibility.
Two. Speed. AI engine crawlers have shorter timeouts than Googlebot. A page that takes more than two seconds to render meaningful content gets partially indexed or skipped. Sites that depend on heavy JavaScript to render their primary content are particularly exposed. The engine sees an empty shell and moves on. This is the technical tax applied to AI visibility rather than to Quality Score.
Three. Structure. The engine extracts claims, not paragraphs. Clear H2 sections, a definition in the first sentence of each section, FAQ blocks with question and answer pairs. Information buried in flowing prose is read but rarely cited. There is nothing to grab cleanly.
Four. Schema. JSON-LD gives the engine a machine-readable version of the page. Article, FAQPage, Organization, DefinedTerm. FAQPage in particular makes each question a self-contained citable unit. Most sites have no schema. The cited ones do.
Five. Specificity. Vague pages do not get cited. "We serve clients across the country" is invisible. "Personal injury cases in California" is citable for California-bound queries. Geography, jurisdiction, specialisation, named subcategories. The more concretely a page commits to a scope, the more cleanly it can be matched to a question.
What the work actually looks like
The infrastructure layer comes first. Robots.txt audit, crawl test for the AI engines specifically, render check to confirm the page paints something useful before the timeout. For most WordPress sites with the kind of accumulated plugin debt that builds up over years, this is the layer that requires the most work. Sometimes the only viable path is to move the primary landing pages to an edge-deployed sovereign node that lives outside the plugin stack.
Then the structural rewrite. Sections with explicit subjects, definitions in the first paragraph, FAQ blocks at the bottom of every primary page. The goal is that every claim has a place where the engine can grab it cleanly.
Then the schema. Article, FAQPage, Organization, DefinedTerm where the page introduces a concept. Schema is the layer most often skipped because it has no visible effect for human readers. It has a large effect for engines.
Then specificity. Replace generic phrasing with concrete commitments. Name the categories you serve, the locations you work in, the credentials behind the author. Make the page answerable to "why is this source authoritative for this question."
How long it takes
Crawl and speed fixes show up first. Two to four weeks before the engine starts including a previously-blocked or previously-slow page in its citation pool. Structural and schema work takes four to eight weeks. Specificity and authority signals compound over three to six months as the engine accumulates evidence that the source is reliable for the topic.
Citation share is not won in a single push. It is accumulated.
Frequently asked questions
Is this the same as AEO or GEO?
Close enough that the terms are used interchangeably across the industry. Answer engine optimisation, generative engine optimisation, citation engine optimisation. Same territory, different labels. The focus here is the citation event specifically. Did the engine name the URL or did it not.
How do I know if it is working?
Run a representative set of buyer-intent queries across ChatGPT, Gemini, Perplexity, and Claude on a weekly cadence. Record whether your domain appears in the cited sources. Compare to your top three competitors. The pattern shows up within a few weeks of starting infrastructure work and tracks consistently after that. A formal version of this measurement is the citation gap audit.
Do I have to abandon SEO?
No. Most of the technical foundation overlaps. Fast pages, crawlable, well-structured, schema-marked. A well-engineered page wins both at once. The mistake is assuming that strong SEO automatically delivers citation share. It does not. The two systems measure different things.
Why do two pages on the same topic get different citation outcomes?
Usually structural clarity. The cited page has cleaner sections, an explicit definition near the top, FAQ pairs that make individual claims extractable. The uncited page has the same information buried in flowing prose. The engine read both. It could only grab cleanly from one.
Does this matter for a small local business?
Yes. Local-intent queries to AI engines have grown through 2025 and 2026. "Best dentist in Austin" now returns a synthesised answer with two or three named practices. For a small practice with a limited ad budget, winning that local citation pool is more valuable than competing for the same query on Google Ads, because the AI engine's recommendation arrives with implicit endorsement.
Can I do this without hiring an engineer?
The content and structural work can be done in-house by someone competent at writing. The infrastructure work — robots.txt, page speed, server-rendering, schema — usually needs technical capability beyond a marketing team's normal stack, especially for WordPress sites with the kind of plugin accumulation that builds up over years. The two halves complement each other. Doing only one tends not to be enough.