Your website has one shot at getting cited by AI. Here’s what happens in those few seconds.

Not long ago, ranking on Google meant showing up in a list of ten blue links. The user picked one, clicked through, and read your page. You had time to make your case.

That’s not how most people find information now.

In 2026, a growing share of searches never reach a list of links at all. Someone types a question into ChatGPT, Perplexity, or Gemini. An answer comes back — already written, already synthesised — with a handful of source links tucked underneath. If your website isn’t one of those sources, you don’t exist for that query.

The question worth asking is: how does the AI decide who gets cited?

It’s not one process — it’s several

Different AI platforms handle retrieval differently, but the broad shape is consistent.

When a user asks a question, the AI doesn’t just pull from its training data. Most modern AI search tools — Perplexity, ChatGPT with search enabled, Gemini — run a live web retrieval process alongside that training knowledge. They’re doing something closer to a search engine than a library lookup.

Here’s roughly what happens:

Stage one: index filtering. The AI scans web index data and filters candidate pages by title, meta description, and topic relevance. Around fifty pages might enter this initial pool. Half are cut immediately — wrong topic, thin metadata, or low authority signals.

Stage two: server speed. The AI fetches the remaining candidates directly. If your server takes more than two seconds to respond, your page is out. This isn’t a soft penalty — it’s a hard cut. Slow hosting costs you the citation regardless of how good your content is.

Stage three: content scoring. The pages that survive get scored against the query. The AI is looking for content that directly answers the question — clearly, specifically, and with some signal of authority. Vague introductory paragraphs, keyword-stuffed copy, and walls of text without structure all work against you here.

Stage four: generation. The top three to five pages are used to generate the answer. Those pages get cited. Everything else is invisible.

The whole process takes a few seconds. Your window to influence it is much longer — but only if you know what to optimise for.

What actually gets a page cited

A direct answer near the top of the page. AI retrieval tools scan for answers, not for content quality in the abstract. If someone asks “how much does live-in care cost in Somerset,” your page needs to answer that question plainly and early — not in paragraph seven after three sections of general background. A clear, specific answer in the first two hundred words dramatically improves your chances of being selected.

Page speed. We’ve mentioned the two-second window. In practical terms, this means fast hosting, no unnecessary scripts loading on page render, and images that are properly compressed. A DebugBear audit will show you where you stand. If your Largest Contentful Paint is above 2.5 seconds, you’re losing AI citations you’d otherwise earn.

Structured content. AI tools parse structure. Pages with clear headings, short paragraphs, and FAQ sections written in question-and-answer format perform better in retrieval than long unbroken prose. Schema markup — particularly FAQ schema and Article schema — helps AI systems understand what your page is about and what questions it answers.

Authority signals. E-E-A-T (experience, expertise, authoritativeness, trustworthiness) matters here as much as it does for traditional SEO. Named authors with credentials, links from relevant third-party sites, and consistent NAP data for local businesses all contribute. These signals are baked into the index filtering stage — pages with thin authority rarely make it through.

Specific, verifiable claims. Generic statements (“we provide excellent care”) carry no weight. Specific, verifiable ones (“CQC-rated Good, serving Somerset since 2018”) do. AI systems favour precision. The more concrete your content, the more useful it is as a citation source.

What’s changed in 2026

A few things have shifted since AI search first became mainstream.

First, the platforms have multiplied. In 2024, most attention was on ChatGPT. Now you’re optimising for ChatGPT, Perplexity, Gemini, Claude, and a growing range of AI-assisted search tools built into browsers and apps. Each has slightly different weighting — Perplexity leans heavily on recency, Gemini integrates tightly with Google’s own index — but the fundamentals hold across all of them.

Second, local and niche queries are being answered by AI at a higher rate than before. If you’re a regional business or a specialist provider, your content is more likely to be directly cited in an AI answer than a generic national competitor’s. AI tools pull from the most relevant source — and “most relevant” often means most geographically or topically specific.

Third, real-time retrieval is now the default for most platforms. Earlier AI tools relied primarily on training data. Now, live web search is standard. This means publishing new, well-structured content on relevant topics has a much faster path to AI visibility than it did two years ago.

The practical checklist

If you want your website to show up in AI-generated answers, these are the changes worth making:

Write content that answers real questions directly and early. Don’t bury the answer.

Structure pages with clear H2 and H3 headings that reflect actual search queries. “How much does X cost?” is a better heading than “Pricing information.”

Add FAQ sections to service and product pages. Use question-answer format that an AI can extract cleanly.

Implement FAQ schema and Article schema markup. Test it with Google’s Rich Results Test tool.

Audit your page speed. Get your server response time below one second and your LCP below 2.5 seconds.

Build topical authority through consistent, specific content in your subject area. Broad, generic content ranks nowhere.

For local businesses: include specific location names, verifiable details, and third-party validation throughout your content.

The bigger picture

Traditional SEO got your page into a list. AEO (answer engine optimisation) gets your content into the answer itself — before the user ever sees a link.

The technical requirements overlap significantly with what good SEO has always demanded: fast pages, clear content, genuine authority. What’s different is the format. AI retrieval rewards directness and structure in a way that search engines, with their tolerance for long-form padding, never quite enforced.

The businesses that figure this out early will be the ones showing up in AI answers a year from now. The ones that don’t will be wondering why their traffic has quietly flattened.

Frequently asked questions

What is AEO and how is it different from SEO? SEO (search engine optimisation) helps your pages rank in traditional search results. AEO (answer engine optimisation) helps your content get selected as a source by AI-generated answers. The principles overlap — speed, authority, clear content — but AEO puts more emphasis on direct answers, structured formatting, and schema markup.

Does page speed really affect AI citation? Yes. AI retrieval tools fetch pages in real time. If your server doesn’t respond quickly, your page is skipped. A response time above two seconds significantly reduces your chances of being cited, regardless of content quality.

Do I need to optimise separately for ChatGPT, Perplexity, and Gemini? The same fundamentals apply across platforms: fast pages, specific content, clear structure, and authority signals. Perplexity weighs recency more heavily, so keeping content up to date matters more there. Gemini draws from Google’s index, so traditional SEO signals carry more weight.

How quickly can changes affect AI visibility? Faster than traditional SEO. AI retrieval tools index recent content more readily than older ranking algorithms. Well-structured new content on a credible domain can show up in AI citations within days.

Is this relevant for local businesses? Particularly relevant. AI tools often pull local-specific answers from local sources. A Somerset-based care provider or a regional trades business that publishes specific, well-structured content about local topics can outperform national competitors in AI answers for local queries.