First, separate the two meanings of AEO. If you are searching for the customs one, the "Hong Kong Authorized Economic Operator Programme" of the Customs and Excise Department also uses the abbreviation AEO. It has nothing to do with this article, so please go to the Customs website. This article is about Answer Engine Optimisation, and its subject is AI tools that give answers directly, such as ChatGPT, Gemini and Perplexity.
How Answer Engines Differ From Search Engines
A traditional search engine outputs a list of links, and the user picks one to click. An answer engine outputs a written answer, sometimes with a few source links and sometimes with no links at all. When a user asks "where in Central is there a print shop open all night", a search engine gives ten websites, while an answer engine names two or three shops directly.
This difference changes the goal of optimisation. SEO competes for "what position you rank in". Answer engine optimisation competes for "whether you are written into the answer". Ranking first but not being cited counts for nothing in the world of answer engines. Ranking eighth with content that is clear and citable may get you named instead. Google's AI Overviews and AI Mode, ChatGPT's search feature and Perplexity all belong to this category.
Where It Overlaps With SEO and Where It Does Not
The overlap is the foundation: a website that can be crawled, response speed, a clear heading hierarchy, and accurate content with depth. Google is direct about this. There are no additional requirements for a website to appear in AI Overviews or AI Mode, and no new files or special structured data are needed. In other words, once the fundamentals of SEO are done well, the admission ticket to the answer engine side is already in hand.
Three things do not overlap. The first is citability: answer engines extract statements that can stand on their own one sentence at a time, not the "atmosphere" of a whole article. The second is entity consistency: if the same company is written differently on different platforms, a machine may not recognise it as the same company. The third is crawler access: the crawlers used by answer engines are different from traditional search crawlers, and one company has several of them with different purposes. OpenAI states that OAI-SearchBot is responsible for ChatGPT search results, GPTBot is responsible for collecting training content, and ChatGPT-User visits in real time when a user asks a question. If OAI-SearchBot is shut out, the website will not appear in ChatGPT's search answers. Anthropic and Perplexity also each have several crawlers with a similar division of roles. This layer is something SEO has never had to handle.
Five Specific Things to Do
1. Check whether robots.txt has blocked answer engine crawlers by mistake. Suppose an air-conditioner cleaning company in Kwun Tong had someone add a rule two years ago to "block all AI crawlers", at the time only to prevent its content from being taken for training. As a result, even OAI-SearchBot, Claude-SearchBot and PerplexityBot, the crawlers responsible for search, were blocked together, and the website disappeared from three answer engines at once. The task is to decide crawler by crawler whether to allow or block: training crawlers can be blocked, and search crawlers should be allowed. For the list of crawlers and how to write the rules, see "AI Crawlers Explained".
2. Write core information as sentences that can be cited directly. A tutorial centre in Tsuen Wan writes on its fees page "We offer competitive tuition fees, enquiries welcome". An answer engine reads it and can extract nothing. Change it to "Form 4 to Form 6 mathematics small group, 90 minutes per lesson, four lessons a month, tuition fee HK$1,600". One sentence covers the subject, format, duration and price, and a machine can cite it as written. For a before and after comparison of a whole rewritten paragraph, see "How to Write Content That AI Search Engines Cite".
3. Add a set of real questions and answers to each service page. What an answer engine receives is mostly a complete question. If the page happens to have the same question with a direct one-sentence answer, it is the easiest to match. A veterinary clinic in Yuen Long writes the eight questions it is asked most often on the phone as "frequently asked questions", answering each one directly in one sentence first and then adding two or three sentences of detail. The questions should use the words customers really say, not industry jargon.
4. Standardise how the company entity is written. A renovation company uses a short Chinese name on its official website, the full name including "Limited" on its social media page, and only its English name in a business directory. To a machine these may be three different companies, and mentions are split between them. Decide on one standard form first, change all three places together, and then declare it once on the official website with Organization structured data. For how to write structured data, see "Structured Data Basics for Accurate AI Citations".
5. Prepare a summary of the website's key points for AI to read. An accounting firm has a sixty-page website, but the only pages it really wants cited are three: scope of services, how it charges and how to make contact. Put an llms.txt in the root directory that lists these pages with one sentence of description each in simple Markdown. The cost is very low. Bear in mind that it is still an industry proposal, not an official requirement of any answer engine. For how to write it, see "llms.txt Guide".
How to Measure: Three Checks You Can Do Yourself
Check 1: ask directly. Draft ten questions that customers would really ask, put them to ChatGPT search, Gemini and Perplexity separately, and record whether the answers mention you and whether they include your link. Only by retesting the same set of questions every month can you see a trend. For detailed steps and blind spots, see "How to Measure Your Brand's AI Visibility".
Check 2: read the server logs. Search the logs for names such as OAI-SearchBot, Claude-SearchBot and PerplexityBot. Records mean answer engine crawlers have visited. No records mean you should first go back to the first item and check access.
Check 3: run a free scan. Enter the domain into an AI visibility checker, and in about 30 seconds it lists results in three areas, crawler access, structured data and citability, which saves checking each item by hand.
What Search Console can and cannot show. Google explains that the traffic a website receives from appearing in AI Overviews and AI Mode is already counted in overall search traffic in Search Console, under the "Web" search type. That means you can see the total, but you cannot see which part comes from AI features, and there is no separate filter. To tell whether AI answers cite you, you still need the three checks above.
Three Common Misunderstandings
Answer engine optimisation will replace SEO. It will not. The two share a foundation, and Google also explains that traffic from AI features is counted within overall search traffic. If SEO is done badly, answer engine optimisation has nothing to build on.
Adding structured data means you will be cited. It does not. Google states that even when the markup is completely correct, there is no guarantee it will be shown. The marked-up content must also match the content visible on the page, and you must not mark up things readers cannot see. Structured data helps machines understand a page. It is not a pass.
Blocking all AI crawlers is safer. These need to be looked at separately. Blocking a training crawler (such as GPTBot) is a copyright position, and OpenAI explains that this does not affect search results. Blocking a search crawler (such as OAI-SearchBot), however, means opting out of ChatGPT search directly. Google's Google-Extended likewise governs only model training and grounding for Gemini-related products, and does not affect a website's inclusion or ranking in Google Search. A blanket block means giving up exposure that could have been obtained for free.
First See Which Step You Are At
For how to tell the three terms SEO, AEO and GEO apart, this series has another article, "SEO, AEO and GEO: What Is the Difference?". Before you start work, spend 30 seconds to see whether your website has its door open or closed to answer engines.
Frequently Asked Questions
Are AEO and GEO the Same Thing?
The market often uses them interchangeably. This series uses AEO for page-level work on "whether a single question can be cited by an answer engine", and GEO for a brand's overall visibility in generative AI answers. For the distinction, see the fifth article in this series, "SEO, AEO and GEO: What Is the Difference?".
How Long Does Answer Engine Optimisation Take to Show Results?
There is no official timetable. Google's statement on general search changes is that some take effect in a few hours and others take several months, and answer engines also have to wait for crawlers to fetch pages again and update their sources, so results are generally counted in months.
Is It Worth Doing for a Small Website?
Yes, and it is easier too. With few pages, the five things can be finished within one or two weeks, and most competitors have still not dealt with crawler access and citability.
Sources
All sources were accessed on 17 October 2026.
- Customs and Excise Department, Hong Kong, 《香港認可經濟營運商計劃》 (Chinese-language page), https://www.customs.gov.hk/tc/service-enforcement-information/trade-facilitation/aeo/index.html
- Google Search Central, "AI features and your website" ("There are no additional requirements to appear in AI Overviews or AI Mode"; traffic from AI features is counted in overall search traffic in Search Console under the "Web" search type; no new files or special schema.org markup needed; page last updated 2025-12-10), https://developers.google.com/search/docs/appearance/ai-features
- Google Search Central, "Google's common crawlers" (Google-Extended manages Gemini model training and grounding; "Google-Extended does not impact a site's inclusion in Google Search nor is it used as a ranking signal"), https://developers.google.com/crawling/docs/crawlers-fetchers/google-common-crawlers
- Google Search Central, "General structured data guidelines" ("Google does not guarantee that your structured data will show up in search results"; content that readers cannot see must not be marked up), https://developers.google.com/search/docs/appearance/structured-data/sd-policies
- Google Search Central, "SEO Starter Guide" ("Some changes might take effect in a few hours, others could take several months"), https://developers.google.com/search/docs/fundamentals/seo-starter-guide
- OpenAI, "Overview of OpenAI crawlers" (OAI-SearchBot "used to surface websites in search results in ChatGPT's search features"; "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers"; GPTBot is used for training and does not affect search; ChatGPT-User is a user action), https://developers.openai.com/api/docs/bots
- Anthropic, "Does Anthropic crawl data from the web, and how can site owners block the crawler?" (ClaudeBot collects content that may be used for training; Claude-SearchBot improves search result quality; Claude-User supports visits made when a user asks a question), https://support.claude.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler
- Perplexity, "Perplexity Crawlers" (PerplexityBot is used to show and link websites in Perplexity search results and is not used for training; Perplexity-User supports visits made when a user asks a question), https://docs.perplexity.ai/docs/resources/perplexity-crawlers