Answer engine optimisation is the practice of structuring your content so that ChatGPT with search, Perplexity and Google AI Overviews cite it when they answer a question. These systems quote a small handful of sources per query and ignore everything else, so the whole job is getting into that handful, and that includes the service pages you actually sell from. This guide covers how answer engines pick their sources and how to rebuild any page, including a brochure-style service page, so it can be quoted.
How answer engines choose their sources
These systems do not read the whole web at answer time. They run a retrieval step: the question is turned into one or more searches, a set of candidate pages comes back, and the model composes its answer from the passages it can extract cleanly. ChatGPT search leans on Bing’s index, Perplexity runs its own retrieval per query, and AI Overviews draw on Google’s ranking systems. Two consequences follow. First, ordinary search visibility still gates AI visibility, so a page that Google or Bing indexes poorly is invisible twice over. Second, the unit of competition is the passage, not the page: the model quotes the single paragraph that answers the question on its own, without needing the rest of the article for context.
Write passages that survive extraction
A quotable passage names its subject, gives the answer, and carries the key qualifier in one self-contained block. Compare “it depends on several things we will cover below” with “answer engine optimisation on a single service page usually shows its first citations within four to eight weeks of reindexing, because the engines recrawl and recompose their answers on their own cycle”. The second sentence can be lifted, attributed and quoted; the first cannot. Write definitions in the pattern “X is Y that does Z”, keep every number and threshold in the same sentence as the claim it supports, and never split an answer across a paragraph of preamble.
Structure pages so they chunk cleanly
Retrieval systems split a page into chunks, usually along its headings. Make each H2 a real question or a complete claim, keep one idea per section, and open every section with its conclusion rather than a wind-up. Tables and lists extract particularly well for comparative and “best X for Y” questions. Keep your dates honest too: answer engines prefer recently maintained pages for anything that changes, so publish a genuine last-updated date and revise the page when the facts move. A 2023 date on a 2026 question is a reason to pass you over. A page built this way is also easier for a person to scan, which is why answer engine optimisation rarely conflicts with classic SEO.
Rebuild service pages so answer engines can quote them
Most service pages are brochures: adjectives at the top, a process diagram in the middle, a form at the bottom. Paste yours into an assistant and ask what it costs, who it is for and how long it takes. If the honest reply is that the page does not say, you have found why it is never cited. Three changes fix that without weakening the sale.
First, replace label headings with question headings. “Our Approach” and “Why Choose Us” carry no retrievable meaning because nobody asks an assistant about your approach. Rewrite each as the question a buyer would type: “How much does local SEO management cost”, “What is included in the monthly retainer”, “How long before the work shows results”. Source those questions from the Google Search Console Performance report filtered on question words with a regex such as ^(who|what|how|when|where|why|can|is|do), from the People Also Ask boxes for your core terms, and from the questions your team fields on every sales call.
Second, open each section with a forty to sixty word answer block that names the subject explicitly instead of starting with “we” or “it”, and keeps any figure in the same sentence as the claim. That length survives extraction whole: it fits a featured snippet, a spoken answer and an AI citation at once. Expand below the block for human readers as far as the sale needs.
Third, do not strip the page while you restructure it. Keep the proof, the pricing signals, the calls to action and the testimonials. This is surgery on structure, not a content diet, and the question-led version usually converts better, because a buyer skimming on a phone reaches their answer sooner.
Add an FAQ and mark it up
Below the main sections, collect the five to eight smaller questions that do not deserve a block of their own and answer each in two or three sentences. Add FAQPage schema markup whose text matches the visible answers word for word, because the engines cross-check structured data against the rendered page and quietly discard anything that does not match. Google now shows FAQ rich results far less often than it once did, but the markup still feeds the retrieval systems behind AI answers, so it earns its place for answer engine optimisation even when it no longer earns a rich result. Validate it in the Rich Results Test and watch the enhancement reports in Search Console for errors.
Give the model a reason to trust you
Models weigh corroboration before they cite. Resolve every page to a clear entity: use one exact service name in the H1, the answer blocks, the navigation label, the schema and your Google Business Profile services list, rather than three near-synonyms scattered across surfaces. Add Organization schema markup site-wide with sameAs links to your real profiles, name an author with genuine credentials, and keep your claims consistent between the page, your About page and your directory listings, because contradiction reads as risk. Unlinked mentions count as well: Reddit threads and comparison articles are retrieved heavily for “best” and “vs” questions, so being discussed accurately in those places is worth deliberate effort. You can monitor where your brand is mentioned with Ahrefs or Semrush, both of which now report AI Overview and assistant citations alongside classic rankings.
Track citations like a channel
Run a fixed set of prompts every month: the fifteen or twenty questions your buyers actually ask, phrased the way people type them into ChatGPT, such as “best technical SEO agency for Shopify sites” or “how do I fix pages crawled but not indexed”. Record which sources each engine cites, so citation share becomes a number you can move rather than a feeling. On the analytics side, watch referral traffic from chatgpt.com and perplexity.ai in GA4; if you export GA4 to BigQuery you can query those sessions at scale and pipe the result into a Looker Studio dashboard sitting next to your citation log. Note that Search Console folds AI Overview impressions into normal Performance data rather than reporting them separately, so read a sudden impression rise on informational queries as a prompt to check.
Start with a one-page pilot
Do not restructure the whole site at once. Take your five most commercially important questions, then pick the single service page with the most revenue behind it. Crawl it first with Screaming Frog or Sitebulb to capture its current headings and confirm it has no answer blocks and no FAQ schema, then record a baseline by asking the engines your top ten buyer questions and noting whether the page is cited anywhere. Rebuild it to the anatomy above, request reindexing in Search Console, and repeat the same prompts monthly. Once the citations start, turn that page into the template for every other service you sell, so your citation surface grows with the site instead of needing another retrofit later. Most competitors have not started any of this, so the early positions are still cheap.
If you would rather have this done for you, our Answer Engine Optimisation service rebuilds your highest-value pages to this anatomy and tracks the citations that follow. Questions are welcome at digitalstrategy@globalcharter.com.
Key takeaways
- Answer engines retrieve and quote self-contained passages, so the passage, not the page, is what competes for a citation.
- Search visibility gates AI visibility: a poorly indexed page cannot be cited, so classic SEO and answer engine optimisation are one programme.
- Rebuild brochure service pages with question headings and forty to sixty word answer blocks, and do not strip the proof, pricing or testimonials while you do it.
- Add FAQPage and Organization schema markup that matches the visible text, and keep one exact service name across the H1, navigation, schema and Google Business Profile.
- Measure with a monthly prompt set for citation share, plus GA4, BigQuery and Looker Studio for referral traffic from chatgpt.com and perplexity.ai.
- Pilot on one high-revenue page first, baseline it, rebuild, request reindexing, then template the win across the site.
Frequently asked questions
What is the difference between AEO and GEO?
Answer engine optimisation (AEO) is the broader practice of getting cited by any system that answers a question directly, from featured snippets to AI Overviews. Generative engine optimisation (GEO) is often used for the generative assistants specifically, such as ChatGPT and Perplexity. In practice the work overlaps almost entirely: self-contained passages, clean chunking, consistent entity signals and trustworthy corroboration serve both, which is why treating them as one effort is more productive than splitting the budget.
Do I need schema markup to get cited in AI answers?
It helps, but it is not the gatekeeper. Passage clarity and ordinary search visibility matter more, because the engines retrieve from indexes and quote plain text. Schema markup such as FAQPage, Organization and sameAs gives them a cleaner read of your entity and your answers, and it is cheap to add, so treat it as support for the writing rather than a substitute for it.
How long before a rebuilt page starts getting cited?
For a single page, expect roughly four to eight weeks after reindexing, because the engines recrawl and recompose their answers on their own cycle rather than on demand. Track it with a monthly prompt set rather than checking daily, and give the page a real last-updated date each time you revise it, since freshness is one of the signals these systems weigh.