Answer engine optimization: how to rank in AI search

What answer engine optimization is

Answer engine optimization is the practice of structuring a website so that AI answer engines can find its pages, understand them and cite them when they answer a question. An answer engine is a tool that replies with a written answer instead of a list of links. ChatGPT, Perplexity, Claude, Microsoft Copilot in Bing and Google AI Overviews all work this way. The same idea goes by other names, including agentic SEO, generative engine optimization and AI search optimization.

It builds on search engine optimization instead of replacing it. A page still has to be crawlable, fast and clear. What changes is the reader: an answer engine does not scan ten blue links, it pulls out a passage and quotes or paraphrases it. Your job is to make the right passage easy to pull out and easy to attribute to you.

Be wary of anyone who sells AEO as a set of secret tricks. No answer engine publishes its full source-selection rules, and they change. What follows is the set of practices that are sensible regardless of which engine is asking, because they also make the page better for people.

How answer engines pick their sources

Each engine works a little differently, but the broad pattern is shared. Understanding it explains why the practices later in this guide help.

• They need to reach the page. Many engines fetch pages live or draw on a search index, and they can only use what a crawler can read. Content that appears only after heavy client-side scripting, sits behind a login or is blocked in robots.txt is often invisible to them.
• They look for a passage that answers the question. The engine breaks the question into searches, retrieves candidate pages, and picks passages that answer it directly. A page that states the answer in its first sentences is easier to use than one that buries it under an introduction.
• They prefer sources that look reliable. Clear authorship, a named organization, consistent facts across pages and across the web, and links to primary sources can all make a page safer to cite. A claim that contradicts itself between two of your own pages can make a page less safe to cite.
• They rely on the usual search signals. Google AI Overviews and Bing Copilot draw on their own search indexes, so ordinary SEO still feeds them. A page that is not indexed has little chance of being quoted.
• They attribute selectively. Some answers cite sources with links and some do not. Treat a citation as a bonus, not a certainty, and judge success by the traffic and enquiries you can measure.

Seven practices that make a page easy to cite

These are the changes worth making first. None needs a new tool, and each one helps ordinary search and human readers as well.

1. Answer the question first. Open each page and each section with a direct, self-contained answer in one or two sentences, then add the detail. If a passage still makes sense when it is pulled out of the page, an engine can use it. Write headings as the questions people actually ask.
2. Add FAQ blocks. A short list of real questions with plain answers matches how people query an engine. Keep the answers visible on the page, not hidden behind a click that a crawler never makes, and make sure each answer repeats only facts the page already states.
3. Mark up the page with structured data. Use schema.org markup in JSON-LD: Organization for who you are, Article for guides, FAQPage for the question blocks and BreadcrumbList for the page trail. The markup must match the visible copy exactly. Markup that says something the page does not is worse than none.
4. Publish an llms.txt file. The llms.txt convention is a plain markdown file at the root of your site that lists your key pages for language models, often with a longer llms-full.txt holding the full text. Support is uneven and no engine promises to read it, but it is cheap, it may help coding agents and some crawlers, and keeping it generated from the same data as the pages stops the two from drifting apart.
5. Keep your entity facts consistent. Say who you are, where you are based, what you do and who does the work in the same words everywhere: your site, your profiles, your directory listings. Engines build a picture of a company from many places, and conflicting details weaken it.
6. Serve fast, crawlable pages. Render the important text on the server so it arrives in the first HTML response. Keep a sitemap with accurate last-modified dates, avoid blocking crawlers you want, and keep pages light. A page an engine times out on cannot be quoted.
7. Back claims with sources and keep them current. Cite primary sources for numbers, name who wrote a page, and show when it was last updated. Do not invent statistics: readers distrust claims that cannot be traced, and engines weigh how reliable a source looks. Review key pages on a schedule and update dates only when the content actually changes.

How to measure AEO

Measurement is the weak point of AEO, because engines rarely report how often they cite you. You can still build a useful picture from three sources.

• Search Console. Google Search Console shows impressions, clicks and queries for your pages. Watch whether question-style queries start appearing, and check that each new page is indexed. Bing Webmaster Tools gives the same view for Bing, which also feeds Copilot.
• Referrers in your analytics. In a product analytics tool such as PostHog, break traffic down by referring domain. Visits from chatgpt.com, perplexity.ai, copilot.microsoft.com, claude.ai and similar domains are people who followed a citation. Track which landing pages they reach and whether they go on to your contact form.
• Asking the engines yourself. Run a fixed list of the questions your buyers ask in each engine, on a regular schedule, and record whether you are mentioned, cited or missing. The answers vary from run to run, so look at the trend over weeks, not a single result.

Pair these with server logs if you have them: requests from the crawlers that answer engines operate show which pages they actually fetch. Treat all of it as directional. A small number of well-qualified visits from an answer engine can matter more than a large number of casual ones.

How Devino applies this on its own site

We hold our own site to the practices above, so this section lists what is live on devino.ca today instead of promising results.

• Answers first, in plain HTML. Service pages, guides and case studies are rendered on the server, open with a direct answer and carry a visible FAQ whose answers are in the HTML without any click.
• Structured data that matches the page. Guides carry Article, FAQPage and BreadcrumbList markup generated from the same data as the visible copy, so the two cannot drift apart.
• llms.txt and llms-full.txt. The site publishes an llms.txt index and an llms-full.txt that contains the full public text of the site, generated from the same data modules that render the pages.
• An honest sitemap. Every page in the sitemap has a last-modified date that is meant to change only when that page's copy or data changes.
• One set of facts. Who we are, where we are based and how we work are stated once and reused, and automated tests flag retired figures before they ship.

If you want a site built this way from the start, or an existing site audited against these practices, that is software work we do. Describe it through the intake form and an engineer replies within one business day.