Answer Engine Optimization
Answer engine optimization (AEO) makes pages easy for AI answer engines to read, understand, and cite. These best practices help your content show up accurately in AI-generated answers.
What is answer engine optimization?
Answer engine optimization is the practice of structuring web pages so AI tools such as ChatGPT, Claude, Perplexity, and Google's AI Overviews can find them, understand them, and quote them accurately in their answers.
It builds on search engine optimization (SEO) rather than replacing it. The difference is in how the content gets used. A search engine lists links to whole pages. An answer engine splits pages into passages, picks the passages that best match a question, and writes an answer from them, citing the source. So a page needs to be readable without running code, clearly labeled, and written in sections that still make sense when quoted on their own.
Can answer engines read the page?
Only if the content is in the HTML the server sends. Most AI crawlers, including OpenAI's GPTBot and Anthropic's ClaudeBot, download a page's HTML but don't run its JavaScript. In Vercel's study of AI crawler traffic, none of the major AI crawlers rendered JavaScript. Google's Gemini, which uses Googlebot, and Applebot were the exceptions.
Sites built with JavaScript frameworks often render their content in the browser, which leaves those crawlers with an empty page. To make the content readable:
- Serve the page's text, headings, and links in the initial HTML, using server-side rendering or pre-rendering.
- Use real links (
<a href>) for navigation, so crawlers can follow them from page to page. - Return a real 404 status for pages that don't exist, so error pages are not indexed as content.
- Keep drafts, templates, and internal pages out of search indexes.
How do engines know what a page is about?
They read the page's title, description, and canonical URL. Every page should have its own:
- Title — the page name followed by the site name, for example "What is UX? - UX Design Guides". Make each title unique.
- Description — a one- or two-sentence summary of the page, under about 160 characters. Search results and answer engines often show this text when they cite the page.
- Canonical URL — the one official address of the page. If the same content is published at more than one address, the canonical URL tells engines to treat the copies as one page instead of splitting credit between them.
A page's opening paragraph is a good source for its description. Write it as a direct summary of the page.
How do crawlers find every page?
A sitemap lists them. Publish a sitemap.xml with the canonical URL of every public page, and keep it up to date as pages are added or removed.
A robots.txt file tells crawlers what they may visit and where the sitemap is. If you want your content cited, make sure it doesn't block AI crawlers. Crawlers only read robots.txt at the root of a host, so a site that lives in a subfolder should also submit its sitemap directly, for example in Google Search Console and Bing Webmaster Tools.
Why add structured data?
Structured data tells engines who published a page, what kind of page it is, where it sits in the site, and when it last changed, in a standard format (schema.org JSON-LD) that doesn't depend on reading the layout. Engines use it to attribute and date what they cite.
Useful types for guides and documentation:
- Organization — the team or institution that publishes and maintains the content.
- TechArticle or Article — the page itself, with its headline, description, and canonical URL.
- BreadcrumbList — the page's place in the site, for example Guides › Process › Validate.
- Date modified — when the content last changed. Engines tend to favor content that is visibly current, so keep this date accurate rather than setting it on every publish.
Use FAQ markup (FAQPage) only for pages that really are a list of questions and answers. Since 2023, Google shows FAQ rich results only for well-known government and health websites.
Why phrase headings as questions?
Because people ask answer engines questions, and a heading that matches the question, followed directly by the answer, gives the engine a passage it can quote as is. This is a writing practice, not a published ranking factor, but it follows from how answer engines use pages.
How does a question heading help?
Answer engines usually split a page into passages at its headings, then match passages against the user's question. A heading helps in two ways:
- Matching — a passage under "How should you validate design decisions?" closely matches someone asking how to validate a design. A passage under "Recommended Approach" has to rely on its body text alone.
- Context — once a passage is cut out of the page, the heading is often all that says what it is about. Generic headings that repeat across pages, such as "Recommended Approach" or "Guiding Principles", don't tell a reader which page the passage came from.
Where does the answer go?
In the first sentence or two below the heading. Answer engines prefer passages that answer the question directly. If the answer comes three paragraphs later, after background, the passage is weaker even with a good heading. Only rephrase a heading as a question if the section opens with the answer.
What does a good question heading look like?
It names the topic and asks what a reader would ask. Some examples:
| Generic heading | Question heading |
|---|---|
| Recommended Approach | How should you validate design decisions? |
| Common Pitfalls | What are common pitfalls when iterating? |
| Guiding Principles | How do you choose the right page type? |
| Mobile Considerations | How should forms work on mobile? |
| Chunking content | What is content chunking? |
When should a heading stay a name?
When the section is a reference entry that people scan or look up by name, such as Resources, Purpose, Typical Features, Anatomy, Variants, or the name of a component or page type. Turning those into questions makes a page harder to scan and adds little, because people already search for them by name. A page where every heading is a question also reads like a marketing FAQ, so save questions for sections that answer something a person would type or say.
Why does placeholder text matter?
Because answer engines index it like any other content. A published page with "lorem ipsum" text, especially in its opening paragraph or description, looks like a low-quality page and can be shown as-is in search results. Replace placeholder text before publishing, or keep the page out of search indexes until it has real content, and keep a list of pages that still need writing.
What should content authors do?
- Open every page with a paragraph that summarizes it in one or two sentences, under 160 characters.
- Phrase guidance headings as the question a reader would ask, and answer it in the first sentence below.
- Keep each section focused on one topic, so it makes sense when quoted alone. See chunking content in the Information Design guide.
- Make headings specific to the page. Avoid generic headings that repeat across pages.
- Name and link the source of every quote and statistic.
- Write alt text for diagrams that explains what they show, since engines can't read the image.
- Replace placeholder text before a page is considered done.
References
- The rise of the AI crawler, Vercel (December 2024)
- Understand JavaScript SEO basics, Google Search Central
- How to specify a canonical URL, Google Search Central
- How to write meta descriptions, Google Search Central
- Changes to HowTo and FAQ rich results, Google Search Central Blog (August 2023)
- Sitemaps protocol, sitemaps.org
- TechArticle and BreadcrumbList, schema.org