Answer engine optimisation (AEO): snippets, voice and AI

Answer engine optimisation (AEO) means shaping your pages so a machine can lift a direct answer from them. It is for any brand whose customers ask questions, which is almost every brand. It matters because the answer now often appears on the results page, in a voice reply or inside an AI summary, and only the page that supplied it gets named.

This article is part of a series. For how AEO sits next to SEO, GEO and the other labels, read engine optimisation explained. For getting cited by ChatGPT, Perplexity and Copilot, read the guide to generative engine optimisation.

What is answer engine optimisation?

AEO is the practice of writing and structuring content so that an "answer engine" chooses one passage of it as the reply to a question. The answer engines are Google's featured snippets and People Also Ask boxes, voice assistants, and AI features such as AI Overviews and AI Mode.

It is narrower than SEO. SEO tries to get a page ranked. AEO tries to get a specific paragraph, list or table quoted.

Where did the term AEO come from?

Nobody owns the term. No search engine, standards body or founding paper defines it, unlike GEO, which comes from a 2023 Princeton and IIT Delhi paper. AEO grew out of practice in the featured snippet era.

That era has a date. In a January 2018 post, Google's Danny Sullivan said featured snippets launched in January 2014, and that for spoken results Google cites the source page in the reply. Once a single answer could be read aloud, marketers began optimising for being that answer, and the label followed.

Why does AEO matter now?

Because the answer has moved onto Google's own page, and fewer people click through to yours. The scale of the shift comes from Google itself.

• 14 May 2024: Google began rolling out AI Overviews to everyone in the US, aiming for more than a billion people by the end of 2024, according to Google's announcement.
• 5 March 2025: Google introduced AI Mode as an opt-in Labs experiment, using "query fan-out" to run several related searches at once.
• 24 June 2025: AI Mode reached India as an English Labs experiment, with voice and photo input. Google said AI Overviews had driven a more than 10% increase in usage for the queries where they appear in markets including the US and India, in its India launch post.
• 8 September 2025: AI Mode added Hindi and four other languages, as TechCrunch reported.
• 19 May 2026: at I/O, Sundar Pichai said AI Overviews had over 2.5 billion monthly active users and AI Mode had passed 1 billion within about a year.

What do AI answers do to clicks?

Every serious study finds fewer clicks when an AI answer appears. The studies use different methods, so read the numbers with their caveats.

• Pew Research Center, July 2025. Pew tracked the real browsing of 900 US adults in March 2025. People clicked a traditional result in 8% of visits when an AI summary appeared, against 15% when it did not, and clicked a link inside the summary in about 1%. Question searches produced a summary 60% of the time. This is observed behaviour, but US-only and one month.
• Ahrefs, February 2026. Ahrefs compared desktop Search Console data for 300,000 keywords between December 2023 and December 2025. An AI Overview correlated with a 58% lower click-through rate for the top-ranking page. The figure depends on a forecast of what clicks would have been, and Ahrefs says it is correlation, not proof.
• Seer Interactive, November 2025. Seer studied 3,119 informational queries across 42 client organisations from June 2024 to September 2025. Organic click-through on queries with AI Overviews fell from 1.76% to 0.61%, but brands cited in the overview earned 35% more organic clicks than brands that were not. Seer notes its sample is its own clients and that citation may reflect brand strength rather than cause clicks.

The pattern is consistent: fewer clicks overall, and the clicks that remain go to the sources that are named.

When the answer is shown on the results page, the only traffic left is for the page that supplied it.

How do featured snippets, People Also Ask and voice answers work?

They all pick one passage from an indexed page and show it as the answer. Google decides which page; you cannot apply for it.

Featured snippets. Google's featured snippets documentation (last updated 10 December 2025) says its systems determine whether a page would make a good featured snippet for a query, and that there is nothing you can do to request one. It lists no required markup. Clicking a snippet can take the reader straight to the matching passage on the page.

The controls. The same page gives you two levers. nosnippet blocks all snippets for a page, featured and regular, and data-nosnippet hides chosen text. A low max-snippet value keeps normal snippets while making featured snippets less likely, though Google says it is not guaranteed. Google's AI features documentation confirms the same controls, plus noindex, limit what appears in AI Overviews and AI Mode. It also says there are "no additional requirements" to appear in those features beyond being indexed and eligible for a snippet.

People Also Ask. Google says featured snippets can also appear inside the "People Also Ask" group of related questions. Each expanded question is effectively another featured snippet, so every question heading on your page is another chance to be picked.

Voice assistants. Voice is the strictest answer engine, because only one answer is read out. A 2018 Backlinko study of 10,000 Google Home results found that 40.7% of spoken answers came from a featured snippet and the average answer was 29 words. It is old and covers one device, so treat it as a signal, not a law. Google's speakable markup is still in beta and limited to English news content on US Google Home devices. The assistants themselves are changing: Google has been moving Android users from Google Assistant to Gemini, a switch delayed into 2026, so voice answers increasingly come from the same AI systems as AI Mode.

How do you structure answer-first content?

Ask the question, answer it at once, then prove it. These are the steps I would follow on any page that should answer a question.

1. Use real questions as headings

Write H2s and H3s as the questions people actually type or say: "How long does a PAN card take to arrive?", not "Processing times". Pull them from Search Console queries, People Also Ask boxes and your sales or support team.

2. Answer in 40 to 60 words, straight away

Put a complete answer in the first one or two sentences under each heading, roughly 40 to 60 words. This is a working convention, not a Google rule, but it matches the 29-word average for voice answers above. The answer must make sense if it is lifted out of the page on its own.

3. Use lists for steps and tables for comparisons

Featured snippets come as paragraphs, lists and tables. Number the steps of a process and put comparisons in an HTML table, not an image.

4. Add evidence under the answer

Follow the answer with the detail: a named, dated, linked source for every number. That is what the GEO research rewards, and it is why a page is trusted enough to quote.

5. Keep FAQs for readers, not for rich results

Google's August 2023 post on changes to HowTo and FAQ rich results limited FAQ rich results to well-known government and health sites and restricted HowTo results to desktop. HowTo results were then dropped on 13 September 2023. Google's structured data documentation records that FAQ rich results stopped appearing in Search from 7 May 2026, and the documentation was removed on 15 June 2026. A visible FAQ section still answers real follow-up questions, so keep one. Just do not expect a special result from the markup.

6. Decide what you do not want quoted

If an answer is your product, such as a paid report or a price calculator, use data-nosnippet on that part. Most brands should not block snippets at all.

When should you do this?

In week one, list the 20 questions your customers ask most and match each to a page. In weeks two to four, rewrite those pages answer-first. From month two, check monthly which questions show your page in snippets or AI answers, and rewrite the ones that do not. Recrawling can take days to months, so judge changes over a quarter.

Where does AEO go wrong?

It goes wrong when teams chase the box instead of the reader. Four common mistakes:

• Answers buried under introductions. If the answer is in paragraph four, a machine will quote someone else.
• Teaser answers. Holding back the answer to force a click usually means losing the snippet altogether.
• Adding schema in place of answers. FAQ and HowTo markup no longer earn rich results in Google.
• Judging success by traffic alone. A page can win the answer and still lose clicks. Track citations and enquiries as well.

In India there is a further gap. AI Mode now works in Hindi and accepts voice and photo questions, yet many brands selling beyond the metros publish answers only as images, videos or English PDFs. I would publish the key answers as plain HTML text, in the languages customers ask in, including Hindi.

Who should own AEO?

The SEO or content lead should own AEO, because it is mostly a writing and structure job on top of SEO. Subject experts, such as doctors, lawyers or product managers, supply the answers and evidence. Developers own templates, page speed and the snippet controls. Analytics owns one monthly report that puts search clicks, AI citations and enquiries side by side.

A small business does not need an AEO specialist. A founder who writes clear answers to customers' real questions is already doing most of the work.

My view

My view is that AEO is the most useful of the new labels, because it describes a writing habit rather than a tool. I managed a ₹1.2 Cr monthly performance budget across Google, Meta and affiliate networks, and the platforms changed their layouts constantly. What lasted was being clearly relevant to the question in front of the customer.

The trade-off is uncomfortable. Writing the full answer at the top can mean the reader never visits your page, and the studies above suggest that is now common. I would still do it, because the alternative is that a competitor is quoted instead. Measure the work by citations, enquiries and sales, not by sessions alone.

Start with your 20 most-asked questions

Write down the 20 questions your customers ask most. For each, find the page that should answer it, turn the question into a heading, and put a 40 to 60 word answer directly under it. Then search each question on Google and in AI Mode and note who is quoted. That list is your AEO plan.