This guide is for owners who’ve handled the basics of AI search — the crawlers can reach your site, you’ve perhaps added an llms.txt file and some FAQs — but you’re still not being named when people ask the AI engines about your space. If that groundwork is new to you, start with our beginner guide on getting your site mentioned by AI search. Here we go after the harder, more valuable thing: actually winning the citation over your competitors. That means understanding the two separate gates a page must clear, writing content that gets extracted, building the credibility AI looks for, staying fresh, and measuring it properly across engines.
The single most useful shift in thinking at this level is realising that getting cited involves two distinct gates, and most owners only optimise one. First is source selection: when someone asks a question, the engine retrieves a set of candidate pages, and yours has to be among them. Second is answer absorption: from those candidates, the engine builds its answer and attributes specific claims — your actual wording has to be the bit it lifts. Passing only the first gate means you’re retrieved but never quoted; passing only the second is impossible if you’re never retrieved at all. You have to win both, and they reward different things — selection rewards authority, relevance and accessibility, while absorption rewards how cleanly your content answers the precise question. Everything below maps to one gate or the other.
To be selected as a candidate source, three things must hold. Your pages must be accessible — AI crawlers allowed, nothing blocking them, which is the prerequisite no other signal can compensate for; confirm it with the AEO Checker. They must be relevant to the query, which means having a page that genuinely addresses the specific question, not a generic page you hope covers it. And the source must carry enough authority that the engine trusts it as worth drawing from; higher-traffic, well-established sites are retrieved more often, which is why the trust work in Step 3 feeds directly back into selection. Practically: make sure each important customer question has a dedicated, findable, well-structured page, and that an llms.txt and clean site structure help the engines map your key content. The LLMs.txt Auditor helps here.
This is where most of the winnable advantage sits, because it’s the most neglected. AI engines build answers by lifting passages that stand on their own, so your job is to make the quotable version of every answer effortless to extract. The core moves:
Measure how clear and extractable your writing is with the Readability checker — dense, convoluted writing is harder for an engine to lift cleanly, just as it’s harder for a person to read.
AI engines don’t decide credibility from your page alone; they look for agreement about you across multiple independent sources before they’ll confidently cite you. This is why a thin, anonymous site struggles to be quoted even with perfect on-page structure — nothing corroborates it. Build that consensus deliberately. Make your business a clear, consistent entity across the web: the same name, description and details on your site, your social profiles, and any directories or listings, so the engines can connect them into one trusted entity. Strengthen the credibility signals Google calls E-E-A-T — real named authors with genuine expertise, clear about-and-contact information, real reviews and third-party mentions — and check where you stand with the E-E-A-T Checker. Earn mentions and coverage elsewhere, because each independent source that describes you the same way raises the odds an engine treats you as the authority on your topic. Authority built here feeds straight back into Step 1: trusted sources get selected more often.
Structured data makes your entity and your answers machine-readable, which supports both gates. Organization schema (with your consistent name, URL, logo and social profiles) helps engines resolve who you are and connect your entity across the web. FAQPage schema presents your questions and answers in the exact structured form that aids extraction. Article schema with a clear, real author reinforces authorship and freshness. You don’t need to code it — the AI Schema Generator produces it — and the deeper detail is in our guide on schema markup.
The engines differ, and freshness is where they differ most. ChatGPT leans on its training plus on-demand search; Perplexity performs a live web search for almost every query and openly favours recent content. The practical consequence is that stale pages quietly lose citations to fresher competitors, particularly on Perplexity and Google’s AI results. So treat your key pages as living: revisit and genuinely update them on a schedule, refresh facts and figures, and let your published and modified dates reflect real updates rather than leaving content to age. A page that’s accurate today beats one that was authoritative two years ago.
You can’t improve what you don’t track, and AI visibility has to be measured differently from rankings. Use the AEO Checker to gauge your overall AI visibility and what’s holding it back, and run a Site Audit to catch anything technical undermining you. Then test directly and regularly: type the real questions your customers would ask into ChatGPT and Perplexity and see whether you’re named, and whether it’s your wording being used. Note that the engines diverge — one may cite you while another doesn’t — so check the ones your audience actually uses. Treat it as an ongoing loop: test, see who’s being quoted instead of you and why, sharpen the relevant page, and re-test.
Treating “AI search” as one thing costs you, because the major engines select and cite differently, and tuning for those differences wins citations. ChatGPT answers partly from its training and reaches for live web search on demand; it tends to cite fewer sources per answer but lean heavily on the ones it does, so being the clear, authoritative answer to a specific question pays off. Perplexity runs a live search for almost every query and always shows its sources, weighting freshness and clear sourcing heavily — which makes it the fastest engine to reward updated, well-structured pages and often the first place you’ll see movement. Google’s AI results sit on top of its existing index, so traditional ranking strength and strong structured data both feed your chances there. The structural work in this guide helps across all of them at once, but when you’re deciding where to focus first, start with the engines your own audience actually uses, and expect Perplexity to respond quickest to freshness and structure changes.
An accountant is technically visible — crawlers allowed, an llms.txt in place — but ChatGPT keeps naming two rival firms when asked “who’s a good small-business accountant.” Testing reveals the rivals have crisp, self-contained answers to exactly that question, clear named authors, and consistent profiles across the web, while the accountant’s site buries its answer in a long “about our approach” essay with no author and inconsistent business details. They rewrite the key pages as direct question-and-answer blocks with the answer stated first, add real author bios and reviews, make the business name and details identical across their site and profiles, add Organization and FAQPage schema, and set a quarterly refresh on the main pages. Over the following weeks the firm starts being named — first on Perplexity, where freshness moved fastest, then ChatGPT. They passed both gates: retrievable and credible enough to select, clear enough to absorb.
The recurring ones at this level: optimising only one of the two gates — readable but not authoritative, or authoritative but written in unextractable waffle. Burying the answer instead of stating it first. Being an anonymous entity with no consistent identity for engines to trust. Letting key pages go stale and losing citations to fresher rivals. Assuming all engines behave the same and only checking one. And treating AEO as a one-off setup rather than a test-and-refine loop.
You’re likely passing only one of the two gates. Being crawlable gets you considered, but you also need answers written to be extracted (stated upfront, self-contained, specific) and enough cross-web authority for engines to trust and select you.
Source selection (your page is retrieved as a candidate for the query) and answer absorption (your specific wording is lifted into the answer). Selection rewards authority and relevance; absorption rewards clear, self-contained answers.
State the answer in the first sentence or two, keep each answer self-contained, be specific with facts and figures, use the real question as a heading, and structure pages as genuine question-and-answer blocks.
Engines that search live, especially Perplexity and Google’s AI results, favour recent content. Stale pages lose citations to fresher competitors, so update your key pages on a schedule with genuine changes.
No. ChatGPT leans on training plus on-demand search; Perplexity searches live for nearly every query and weights freshness more. Check the engines your audience actually uses, as one may cite you while another doesn’t.
Use the AEO Checker for an overall view and what’s holding you back, then test directly by asking the engines your customers’ real questions and seeing whether you’re named and quoted. Re-test as you improve.