What actually changed
For twenty years the goal was a position. Now a growing share of commercial questions never reach a results page at all: someone asks an assistant, gets a synthesised answer, and clicks one or two of the sources it names. The competition is no longer ten slots. It is three or four citations.
That shifts the work. Comprehensiveness used to win because it matched more queries. Quotability wins now, because a model has to lift a clean, self-contained claim out of your page and stand behind it. If your best answer is spread across four paragraphs and a video, you are harder to cite than a competitor who states it in two sentences.
1 to 5
Sources typically cited in a single AI answer, versus ten organic positions
Days
How quickly a retrieval-based assistant can pick up a newly crawled page
40 to 60
Words in the ideal direct answer under a question-shaped heading
Unit of success
- AI citation
- Being named or linked inside one answer
- Classic organic
- Ranking in a list of ten blue links
Winner takes
- AI citation
- Usually 1 to 5 cited sources
- Classic organic
- Ten organic positions plus features
Speed to show up
- AI citation
- Days after a crawl
- Classic organic
- Weeks to months
Best page shape
- AI citation
- One question, one clean answer, evidence below
- Classic organic
- Comprehensive topic coverage
Biggest blocker
- AI citation
- Crawler blocked or answer buried
- Classic organic
- Weak authority and thin content
Reporting
- AI citation
- Prompt logs, assistant referrals, bot hits
- Classic organic
- Rankings, impressions, clicks
Be an entity, not just a website
Models reason about things, not URLs. Before an assistant will name you, it needs to be confident about what you are, where you operate and what you sell. That confidence comes from consistency, repeated across your own site and the wider web.
- One canonical name, spelled identically everywhere, including your schema, your footer and your directory listings.
- One Organization node with sameAs links to your real profiles, and one WebSite node that references it.
- Named humans with job titles and areas of expertise, not an anonymous "team".
- Explicit service definitions with published prices, so the model can answer a commercial question without guessing.
- A stable address, service area and contact route that match everywhere they appear.
This is the least glamorous part of the job and the part most sites skip. It is also why a small specialist can outrank a large agency in an answer: the specialist is unambiguous.
If a model cannot restate what you do in one sentence, it will not risk citing you in one either.
Write answers that can be lifted verbatim
Structure your pages the way an assistant reads them. Every heading that asks a question should be followed immediately by a direct answer of roughly 40 to 60 words, written so it makes sense with no surrounding context. Then add the nuance, the caveats and the examples underneath.
The pattern that gets quoted
- Question-shaped H2, phrased the way a person would actually ask it out loud.
- A self-contained answer paragraph directly beneath it, with the specific number, price or timeframe included.
- Evidence next: the method, the range, the exception.
- No pronouns doing heavy lifting. Name the subject in the answer sentence.
Specificity is what separates a citable claim from marketing noise. “Affordable monthly SEO” is unquotable. “SEO and AI visibility from £1,250 a month plus VAT, with a fixed number of backlinks and content assets each month” is a fact a model can hand to someone asking about price.
Structured data that earns citations
Schema does not force a citation, but it removes ambiguity, and ambiguity is what gets you dropped. The set worth maintaining is small: Organization and WebSite sitewide, Service or Product with real Offers on commercial pages, FAQPage where you genuinely answer questions, Article with a named Person author on guides, and BreadcrumbList on anything nested.
- Give every node a stable @id and reference it, rather than repeating the same organisation ten times.
- Include sameAs links to authoritative identifiers such as Wikidata entries for your topics and tools.
- Mark prices as they are actually sold, including currency and whether tax is included.
- Keep schema and visible copy identical. A mismatch is worse than no schema.
Let the right crawlers in
The most common cause of zero AI visibility is the dullest one: the assistant cannot fetch the page. Check your robots.txt for blanket disallows on GPTBot, OAI-SearchBot, Google-Extended, PerplexityBot, ClaudeBot and Bingbot. Check that key content is not rendered only after a client-side fetch, and that your sitemap is accurate.
A plain-text summary file at /llms.txt is a cheap addition: a short, factual description of who you are, what you sell, what it costs and where the canonical pages live. It will not carry you on its own, but it gives a retrieval system an unambiguous starting point.
Off-site corroboration still decides ties
When two sources say the same thing, the model prefers the one corroborated elsewhere. That means real mentions: client case studies with named brands and numbers, directory and trade listings with matching details, guest pieces, review platforms, and press where it is warranted. Volume matters far less than agreement.
How to measure it without guessing
Pick 20 to 30 prompts a real buyer would type, covering the whole funnel from “what does X cost” to “best X in Norwich”. Run them monthly across ChatGPT, Gemini and Perplexity, and log three things: were you mentioned, were you linked, and who was cited instead. Pair that with referral traffic from assistant domains and crawler hits from your server logs.
That gives you a share-of-answer trend you can actually act on, rather than a vague sense that AI is eating your traffic. If you want this run for you, it is included in our SEO & AI Visibility subscription, with the prompt set and the results in your monthly report.
Questions we get asked
What is generative engine optimisation?
Generative engine optimisation, or GEO, is the practice of making a website easy for AI assistants such as ChatGPT, Google Gemini, Perplexity and Copilot to understand, trust and quote. It overlaps heavily with technical SEO, but it optimises for being cited in an answer rather than ranked in a list of links.
Do AI assistants use Google rankings to decide what to cite?
Partly. Most assistants either use a search index directly or run their own retrieval over crawled pages, so strong organic visibility still helps. But citation also depends on how cleanly a page answers one specific question, whether the claim is attributable to a named author or organisation, and whether the site allows the assistant's crawler at all.
Should I block AI crawlers in robots.txt?
Not if you want to be cited. Blocking GPTBot, Google-Extended, PerplexityBot, ClaudeBot and similar agents removes you from the pool of sources those assistants can quote. Block them only if you have a commercial reason to keep content out of AI answers entirely, and accept the visibility cost.
How long does it take to appear in AI answers?
Faster than classic SEO. Retrieval-based assistants can pick up a new page within days of it being crawled, whereas an organic ranking usually takes weeks to months. The slower part is entity trust: consistent naming, structured data and third-party mentions build up over a quarter or two.
How do you measure AI visibility?
Three ways. Track referral traffic from assistant domains in analytics, run a fixed set of prompts monthly and log whether your brand is named or linked, and monitor AI crawler hits in server logs to confirm your pages are being fetched at all. We report all three inside our SEO & AI Visibility subscription.
Does long content help AI visibility?
Length is not the lever. Structure is. A 900-word page that answers one question with a clean 40-word definition near the top gets quoted more often than a 3,000-word guide that buries the answer in paragraph nine. Write the answer first, then the evidence.
If clicks have already dropped while rankings held, read the AI Overviews guide next, then check the Growth Bundles if you would rather this was somebody else’s monthly job.
Written by
Sam, co-founder, growth and paid media
Runs growth and paid media at Mesmerise Marketing. Ten years of paid and organic campaigns for challenger brands across DTC, B2B services and hospitality.
Meet the team
