What Is AEO? A Hands-On Guide to Answer Engine Optimization

Picture a resort owner who checks Google and finds their site still sitting at position 2, exactly where it has been for months. Yet direct bookings through the website are noticeably down. Search Console tells a more uncomfortable story: impressions are up, clicks are down.
In a scenario like this, the cause usually isn’t a competitor overtaking them. It’s that travelers have stopped typing “beachfront hotel Koh Lanta” and clicking through site after site. Instead they ask ChatGPT, “Which part of Koh Lanta is best to stay in with small kids?” and the AI writes a complete answer with three or four cited sources. If the resort’s site isn’t one of them, it never gets considered.
That is why AEO has become such a hot topic, and this guide walks you through doing it properly from start to finish.
What is AEO?
AEO (Answer Engine Optimization) is the practice of shaping your content and site structure so that answer engines such as ChatGPT, Google AI Overview, Perplexity, Gemini and Microsoft Copilot can pull your content into their responses and cite your site as the source. The goal isn’t just to rank. It’s to be chosen as the answer.
The key difference is the unit of success. Classic SEO is measured by position on a results page. AEO is measured at the sentence level. If the paragraph under your heading answers the question completely, clearly and verifiably, AI will use it. If you spend three paragraphs warming up before getting to the point, AI will move on to a site that answers faster.
Answer engines aren’t entirely new. The idea goes back to Featured Snippets and People Also Ask, where Google lifted a short answer to the top of the page. What has changed is that the answer is no longer quoted directly. It is synthesized from multiple sources, with citation links alongside. Sites that already win Featured Snippets often have an edge in AEO, because the content structure is very similar.
SEO vs AEO vs GEO: what’s the difference?
These three overlap enough to confuse almost everyone. The easiest way to think about them is as different destinations for your content.
| Aspect | SEO | AEO | GEO |
|---|---|---|---|
| Goal | Rank on the results page | Be chosen as “the answer” to a question | Be cited in answers AI generates |
| Destination | Google and Bing’s ten blue links | Featured Snippets, AI Overview, voice assistants | ChatGPT, Perplexity, Gemini, Copilot |
| Unit of competition | The page | The passage or answer block | Facts and your brand as an entity |
| Winning content format | Comprehensive, deep, well-linked | Short, tight, verifiable Q&A | Content with data, statistics and clear sourcing |
| Main metrics | Rankings, organic clicks | Snippet ownership, AI citation rate | Share of voice in AI answers, brand mentions |
| Key technical factors | Crawling, speed, internal links | Schema, question-style headings, answer blocks | AI-bot accessibility, llms.txt, clear entities |
| Hoped-for outcome | Clicks to your site | Clicks plus zero-click visibility | Brand credibility in the eyes of AI |
The short version we always give clients: SEO gets you found, AEO gets you used as the answer, and GEO gets you talked about credibly. All three rest on the same foundation of good content and a clean site. They are not three separate workstreams.
For brand-level strategy, read our guide to GEO (Generative Engine Optimization). For Google’s mechanics specifically, see Google AI Overview and SEO. This article focuses purely on getting your content picked up as the answer.
Why AEO matters now
The general pattern across the industry points one way, and it’s worth stating plainly that these are directional trends rather than hard numbers:
- Queries that trigger an AI Overview tend to see lower click-through rates than before, even when rankings stay the same.
- Referral traffic from ChatGPT and Perplexity is growing, starting from a small base for most sites.
- Visitors arriving from AI answers often convert well, because they’ve already done their comparing and arrive ready to decide.
Put simply: less traffic, but better traffic. A business that only tracks total sessions may panic at its graphs even though sales haven’t dropped at all. That’s why it’s worth separating out AI referral traffic in your Google Analytics 4 reports starting now.
How AI decides which answer to use
Before you start, it helps to understand that most of these systems work by retrieval first, then synthesis. The system fetches the most relevant chunks of content, then has the model write an answer from them. Whether you make the cut depends on:
- Does your chunk make sense on its own? If a paragraph starts with “Another option is…”, it becomes meaningless once it’s cut out of context.
- Does it contain concrete facts? Figures, steps, timeframes, conditions. AI prefers verifiable information to marketing language.
- Is it credible enough? Who wrote it, what it’s based on, when it was updated. This is where E-E-A-T comes fully into play.
- Can the bot actually reach it? If content is rendered by JavaScript after load, many bots won’t see it at all.
How to do AEO, step by step
1. Write a 40–60 word answer block under every heading
This is the technique that delivers results fastest and that the fewest people use. Under every H2 or H3 that poses a question, answer it fully in 40–60 words first, then expand in the next paragraph.
A formula that works is definition + reason or context + a figure or condition. Instead of writing “SEO for restaurants involves many aspects to consider…”, write “Restaurant SEO means optimizing your Google Business Profile, menu photos and reviews so you appear in ‘near me’ searches. Restaurants that update their profile regularly are better placed to win calls from nearby diners.”
The second version can be used as an answer without changing a word.
2. Turn headings into the questions people really ask
AI matches user questions to page headings quite literally. Use H2s and H3s phrased as “What is…”, “How is X different from Y”, “How long does it take”, “Can I do it myself” rather than clever marketing headlines.
The best sources of questions aren’t expensive tools. They’re People Also Ask, the queries in Search Console, the questions customers repeat in chat, and comments on your social pages. Turn those directly into your page outline, and keep your heading hierarchy clean: don’t jump from H2 to H4.
3. Implement the core schema types
Schema doesn’t directly improve rankings, but it tells machines what your content means, which helps a lot with being retrieved. Every site should have:
- Article / BlogPosting: author, publish date, update date
- FAQPage: for Q&A blocks at the end of articles, using only questions that actually appear on the page
- HowTo: for clearly ordered, step-by-step content
- Organization and Person: tie your brand and authors to real entities, with
sameAspointing to social profiles and your Google Business Profile - BreadcrumbList: helps machines understand your site structure
For full JSON-LD examples, see our Schema Markup guide. One warning: never put information in schema that isn’t visible on the page. That counts as spammy structured data and can cost you rich results.
4. Create llms.txt and let AI bots in
An llms.txt file at the root of your domain works like a short table of contents for language models: who you are, which pages matter, and what to read first. It isn’t a standard every provider supports yet, but it’s cheap to make and has no downside.
At the same time, check robots.txt to make sure you aren’t blocking the bots you want, such as GPTBot, PerplexityBot, ClaudeBot and Google-Extended. Many sites blocked them in 2024 during the anti-AI backlash and forgot to undo it, quietly locking themselves out.
5. Don’t hide content behind JavaScript
Googlebot renders JavaScript reasonably well, but many AI bots don’t render it at all. They fetch the raw HTML and read what’s there. If your main content loads through a client-side fetch, or sits inside tabs or accordions that only inject text into the DOM after a click, then as far as those bots are concerned you have no content.
A simple check that needs no tools: press Ctrl+U to view the page source and search for a key sentence from the page. If you can’t find it, neither can the bot. The fix is server-side or static rendering.
6. Build your entity and brand mentions off-site
The most common AEO mistake is assuming everything happens on your own site. Models also learn who your brand is from elsewhere: media articles, forum threads, Google Maps reviews, LinkedIn profiles, and Wikipedia or Wikidata if you qualify.
Being mentioned in the right context, for example appearing in a list of recommended SEO providers in Phuket, carries more weight with AI than writing that phrase ten times on your own site. The only sustainable way to get there is the real thing: doing work that customers talk about, not posting about yourself.
How to tell whether your content is being cited
This is the question clients ask most, and the honest answer is that no single tool measures it completely. You can build a picture from several sources, though.
| Method | What you can see | Limitations |
|---|---|---|
| GA4 referral segments | Sessions from chatgpt.com, perplexity.ai, copilot.microsoft.com | Only shows people who actually clicked |
| Search Console | High impressions with low clicks may mean use in AI Overview | Doesn’t break out AI data directly |
| Server logs | Which pages GPTBot and PerplexityBot crawl most | Requires log access |
| Structured manual testing | Ask AI a set of 20–30 questions each month and record who gets cited | Time-consuming; results vary between runs |
| AI visibility tools | Share of voice and brand mentions in AI answers | Paid; coverage of some languages is still limited |
For small businesses we recommend a simple routine: write 20 monthly prompts that reflect your customers’ real questions, run them in ChatGPT, Gemini and Perplexity, and log the results in a spreadsheet with three columns: were you mentioned, were you linked, and which competitors were mentioned. After three months you’ll see a clear trend without spending anything.
A hypothetical example
Consider a clinic with a well-written service page that reads nicely as marketing copy but is never picked up by AI. A sensible AEO fix would involve three changes:
- Add question headings based on what patients actually ask in chat, each followed by a 45–55 word answer block.
- Add FAQPage and Article schema, naming the doctor who reviewed the content and their professional credentials.
- Move key information that loads via JavaScript, such as a treatment comparison table, into plain HTML.
The point of an example like this is that AEO is a separate layer from rankings. A page can become a cited source in AI answers even if its Google position barely moves, which is why you need to track it separately.
Common AEO mistakes
- Writing long without answering. 3,000 words without a single sentence that answers the question directly will lose to an 800-word article that answers clearly in its first two lines.
- Adding FAQ schema for questions not on the page. You risk losing rich-result eligibility and credibility.
- Copying answers from AI. Content with nothing new won’t be chosen, because the model already knows it. What gets you cited is primary information: real processes, real cases, figures from your own work.
- Leaving out the author and update date. Both cost nothing and matter a lot for credibility.
- Blocking AI bots in robots.txt and forgetting about it. Check at least once a year.
- Neglecting SEO fundamentals. A slow site with broken structure and no internal links gives AI nothing to work with. Technical SEO is still the foundation.
- Treating it as a one-off. AI answers change every month. Review at least once a quarter.
At Southern Whale, we know most business owners don’t have time to comb through schema line by line or run a monthly prompt set. That’s why AEO work is built into our SEO service: planning the question structure, writing answer blocks, implementing structured data, and reporting which AI tools mentioned your brand each month. If you’d like to know where your site stands, get in touch.
AEO action checklist
- Every important heading opens with a 40–60 word answer block
- At least half of your H2/H3s are phrased as questions people really ask
- Article and FAQPage schema are in place and pass the Rich Results Test
- Author name, role and last-updated date appear on the page
- Each article has at least three concrete facts, steps or figures
- View-source shows the full main content without relying on JavaScript
- robots.txt doesn’t block GPTBot, PerplexityBot, ClaudeBot or Google-Extended
- llms.txt and an automatically updated sitemap.xml are in place
- Internal links connect articles within the same topic cluster
- A 20-question prompt set is measured monthly
Frequently asked questions
Do AEO and SEO need to be done separately? No. AEO is a layer on top of good SEO. If your site is slow, poorly structured or thin on content, AEO won’t achieve anything. Get your on-page SEO solid first, then add the AEO layer.
How long does AEO take to show results? It varies with how often bots crawl your site and how much authority your domain already has. Restructured content commonly takes a couple of months or more to start being picked up, and new sites take noticeably longer. Nobody can guarantee a timeline.
Can a small site compete with big sites in AI answers? Yes, and sometimes it has the advantage, because AI tends to pick the most specific source. A small hotel that writes in detail about getting to its own island can beat a giant OTA writing in general terms.
Do I need FAQ schema on every page? No. Only use it on pages that genuinely contain questions and answers, and they should be questions people actually ask, not ones invented for SEO.
Does zero-click mean websites are dead? Not at all, but you need to measure differently. Being cited in an AI answer is a new form of brand awareness. People may not click today, but they remember your name and search for it later. Track branded search trends alongside everything else.
Summary
AEO isn’t a new trick that requires tearing everything down. It’s writing that genuinely answers people’s questions, structure that machines can understand, and a site that bots can get into. Good SEO should already be doing all three. The difference is that they’ve gone from nice-to-have to essential.
Start with something you can do tomorrow: pick the three pages with the highest impressions, add a 40–60 word answer block under every heading, show the author and update date, and check that robots.txt isn’t shutting bots out. That alone puts you ahead of most competitors.
Want us to handle this for you?
We audit your site for free first. Send the URL and we will tell you where it is losing ground and what to fix in what order.
