What is answer engine optimization?
Answer engine optimization is the work of making a page easy for a machine to read, check, and quote when it answers someone’s question. The machine might be a search feature such as Google’s AI Overviews, or an assistant such as ChatGPT, Claude, or Perplexity.
The page has to do three things for that reader. It has to state its answer in plain text, so the answer can be lifted out. It has to agree with every other source about the business, so the answer can be trusted. And it has to load without scripts, so the answer can be found at all.
None of this is new craft. Good writers have long put the answer first, and good developers have long shipped content in the HTML. AEO gives those habits a name and a checklist, because more people now meet a site through a summary than through a visit.
How is AEO different from SEO?
AEO builds on SEO and shares its foundation: a page must be crawlable, indexed, and useful before anything else applies. The difference is the target. SEO aims to rank a page in a list of links. AEO aims to have a passage from the page used inside an answer.
Google says this directly in its documentation on AI features in Search. To appear as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to show a snippet. Google states there are no additional requirements and no special optimizations needed. So for Google, AEO sits on top of ordinary SEO fundamentals and adds no separate technical track.
In practice, the two disciplines reward the same pages. A page that answers one question clearly, with supporting detail below, serves a searcher who scans a result and an assistant that quotes a sentence. The work that differs is mostly editorial: writing so that a single paragraph still makes sense when it appears alone.
How do AI assistants choose what to quote?
Only part of the selection logic is public: which crawlers visit, what Google requires for its AI features, and that structured data should match the visible page. Nobody outside the vendors knows the rest.
Here is what the public sources say:
- Google’s documentation, linked above, says a page needs to be indexed and snippet-eligible, and that no special schema is required for AI features.
- OpenAI documents three crawlers: OAI-SearchBot, which surfaces sites in ChatGPT search; GPTBot, which gathers content that may be used for training; and ChatGPT-User, which fetches a page when a user asks for it.
- Anthropic documents three crawlers: ClaudeBot for training, Claude-User for user-directed retrieval, and Claude-SearchBot for search relevance.
- Perplexity documents two: PerplexityBot, which surfaces sites in its search results and is not used for foundation model training, and Perplexity-User, which fetches pages on a user’s behalf and, per its documentation, generally ignores robots.txt rules.
Anything beyond that, such as a ranked list of “AI ranking factors”, is a guess. Treat claims of a secret formula with caution.
What should a small site change first?
A small site should make four changes, in this order: put the answer first, keep the text in the HTML, state one consistent set of facts, and add schema that matches the page. Each costs little, and each helps ordinary search too.
Answer the main question in the first 60 words
The opening of every page should answer the question a buyer would type, in about 40 to 60 words. If an assistant quoted only that paragraph, it should still be correct and useful.
Start with the subject and the verb. Put background and caveats after the answer. Questions in headings help, because each section can then open with its own short answer, and each answer can be quoted alone.
Keep every word in server-rendered HTML
Content that appears only after JavaScript runs is content some crawlers never see. Plain HTML sent from the server is read by every tool, so the safest build puts the words in the markup and uses scripts for enhancement.
You can test this in a minute. Open the page source, or turn JavaScript off in the browser, and check that the headline, the opening answer, and the contact details are still there. This is a build decision, which is why the studio treats it as part of website builds.
Write one set of facts about the business and reuse it everywhere
Write the name, what the business does, where it is based, and how to contact it once, then reuse the exact wording on the home page, the about page, social profiles, and directory listings. Machines compare sources, and disagreement lowers trust.
A short fact sheet works well. List the legal and trading names, a one-sentence definition, the location, the services, and the public profiles. Keep it in one file and copy from it. The same discipline applies to people: use one spelling of each name and one title for each role.
Add schema that matches the visible page
Structured data describes a page to machines in a vocabulary from schema.org. Add types such as Organization, Article, or FAQPage only when the page shows the same facts to a human reader.
Google’s AI features page makes the matching rule explicit: structured data should agree with the visible text. Schema does not create trust by itself, and Google says no special markup is needed for AI features. It removes ambiguity, which helps every system that reads it. Content structure is also a design problem, and it connects to the studio’s product and UX work.
Should you let AI crawlers in?
Yes for the crawlers that bring visitors, and it is a choice for the ones that train models. The crawlers listed above each have their own name in robots.txt, so you can treat search use and training use separately.
OpenAI’s documentation gives an example: a site can allow OAI-SearchBot to appear in ChatGPT search and disallow GPTBot to opt out of training. Anthropic separates training, search, and user fetches into three crawlers. Perplexity separates search from user fetches, and says PerplexityBot is not used for training. Decide on each purpose, then write one rule per crawler.
Two cautions. Blocking a search crawler removes you from that product’s answers, so a block should be deliberate. And Perplexity-User, the fetcher that acts on a user’s request, generally ignores robots.txt, as Perplexity’s documentation says. Robots.txt expresses a preference. It does not enforce one.
A newer addition is Content Signals, a policy published by Cloudflare under a CC0 license. It adds one line to robots.txt, such as Content-Signal: search=yes, ai-train=no, that states a preference for three uses: search (building a search index and showing results), ai-input (feeding content into an AI model, for example to ground an answer), and ai-train (training or fine-tuning a model). A yes permits the use, a no prohibits it, and an omitted signal states nothing. Like the rest of robots.txt, it signals a preference and does not block anyone technically. This site’s robots.txt sets all three to yes.
How do you know it is working?
Check four things: indexing, impressions, referrals, and repeated manual queries. No single number tells you whether an assistant is quoting you, so combine small signals.
- Indexing and impressions. Google Search Console shows whether pages are indexed and which queries show them. Bing Webmaster Tools does the same for Bing.
- Referrals. Look in analytics for visits from chatbot domains. The counts will be small and incomplete, because many answers never produce a click.
- Manual checks. Keep a list of ten questions a buyer would ask. Ask them in several assistants each month, on a fixed date, and record whether your site appears and what it says about you.
- Accuracy. When an assistant misstates a fact, find the page it may have drawn from and fix the source.
Keep the record in a spreadsheet. Trends across months mean something, and a single result does not.
Can anyone guarantee a citation?
No. Assistants decide what to quote, their behavior changes, and no outside party controls it. Anyone who sells a guaranteed citation or a guaranteed ranking is promising something they cannot deliver.
What a studio can do is make the page easy to read, accurate, and consistent, then measure the result honestly. That is what the SEO and AEO practice covers, and this site is built to those rules: answer-first pages, one fact sheet, text in the HTML, and schema that matches what you see. You can check any of it with View Source.
I wrote this guide from my own practice, drawing on computer science training and the information architecture side of UX design. If you want a site reviewed against these changes, email the studio. For the accessibility side of the same build standard, read the practical WCAG 2.2 note.