Answer engine optimisation (AEO) in 2026: the guide I'd write today

TL;DR

• Answer engine optimisation (AEO) gets a specific answer from your page into an AI-generated response, so the thing you optimise is the answer itself, a definition, a number, a comparison, or a step
• Measurement finally exists: Google rolled out generative AI performance reports in Search Console to every site on 31 August 2026, and Bing Webmaster Tools now reports AI citations and citation share
• Google's own May 2026 guide says unique, non-commodity content will shape your presence in AI search more than anything else it recommends, and warns that spinning out pages for every fan-out query breaks its spam policy
• Clicks keep shrinking, with 68.01% of US Google searches ending without a click in early 2026
• There is still no special schema, AI markup, or llms.txt requirement for Google's AI features
• Put the answer in the first third of the page, back it with something only you know, and track citations alongside conversions

In May 2025 I wrote the comprehensive guide to answer engine optimisation for CXL. CXL refreshed it in August 2026 and a fair bit of it holds up. Some of it is already showing its age, though. A couple of the numbers have been overtaken, and two of the sections I gave the most space to, voice search and featured snippets, read like a different era now.

So this is the version I'd write today, for the people Contengi is built for: the solo founder, the one-person marketing team, the creator running everything from one laptop with too few hours in the week. Same core idea, fresher evidence, and a workflow you can run without a team of 10.

What answer engine optimisation is

Answer engine optimisation is the practice of making the useful parts of your content easy for AI systems to find, understand, and justify citing. The answer engines that count in 2026 are Google's AI Overviews and AI Mode, ChatGPT, Perplexity, Claude, Gemini, and Microsoft Copilot.

One idea from the CXL guide I'd defend today: the unit of AEO is the individual answer, a definition, a price, a comparison, one step in a process. An engine lifts the paragraph that settles the question and leaves the rest of your page where it found it.

AEO sits on top of SEO. Google's AI features run on its core ranking systems, so a page that isn't indexed or can't rank won't get cited either. If you want the small business version of the basics, start with our guide to getting your business mentioned by LLMs.

What changed since I wrote the CXL guide

You can finally measure it, at least partly

When I wrote the original, AI visibility was guesswork plus an expensive tracker. That changed this year. Google introduced generative AI performance reports in Search Console on 3 June 2026 and rolled them out to all sites worldwide on 31 August. The report counts how often links to your site appeared in AI Overviews and AI Mode, broken down by page, country, device, and date. It shows impressions only, so you see how often you were shown and nothing about clicks.

Microsoft got there first. Bing Webmaster Tools launched an AI Performance report in February 2026 that shows how often Copilot and Bing's AI summaries cite your pages, plus the grounding queries the AI used to find them. In June it added Intents, Topics, Compare, and Citation Share. Citation Share is the percentage of all citations for a grounding query that went to your site. Bing is clear that it isn't a ranking or a competitor scoreboard, and it doesn't show competitor domains. For a small team it's still the most useful free AI metric on the table right now.

Google wrote its own AEO playbook

In May 2026 Google published a guide to optimising for generative AI features in Search. Two parts of it are worth pinning above your desk.

The first explains the mechanics. AI Overviews and AI Mode use retrieval-augmented generation, which pulls pages from the Search index to ground the answer, and query fan-out, which fires off related searches around the original question. Google's own example takes "how to fix a lawn that's full of weeds" and fans it out into searches about herbicides, removing weeds without chemicals, and preventing them in the first place. Your page can earn a citation by answering one of those side questions well, even if it never ranks for the main one.

The second tells you what not to do with that knowledge. Creating separate pages for every fan-out variation, mainly to manipulate AI responses, falls under Google's scaled content abuse policy. The same guide says content people find unique, compelling, and useful will likely shape your presence in generative AI search more than any of its other suggestions. Google even uses the phrase non-commodity content, an idea we've been banging on about for a while in our guide to making non-commodity content.

The audience got bigger and the clicks got smaller

ChatGPT now has 1.2 billion weekly users, according to OpenAI on 29 September 2026. The CXL guide quotes 900 million, so that number has already moved twice. Google says AI Overviews has more than 2.5 billion monthly users and AI Mode passed 1 billion within a year of launch.

Clicks went the other way. SparkToro, working with Similarweb's clickstream panel, found that 68.01% of US Google searches ended without a click between January and April 2026, up from 60.45% in 2024. Ahrefs re-ran its study with December 2025 data and found that an AI Overview now correlates with a 58% lower click-through rate for the top-ranking page. Pew's browsing data shows the same pattern from the searcher's side: people clicked a traditional result on 8% of visits when an AI summary appeared and 15% when it didn't, and clicked a link inside the summary on 1% of visits.

SparkToro also found that only 0.34% of searches went to AI Mode in those four months, a number that deserves more airtime than it got. AI Mode is growing fast, but if you're a small team deciding where to spend effort this quarter, AI Overviews and ChatGPT are where people are reading your answers today.

What I'd keep from the original

Put the answer where it can be found. CXL's analysis of 100 Google AI Overview citations found 55% of cited passages sat in the first 30% of the source page. Lead with the answer, then give the context.

Target questions, in full sentences. Pew found AI summaries on 60% of searches that started with who, what, when, or why, and on 53% of searches with 10 words or more. One and two-word searches triggered a summary 8% of the time. The longer and more conversational the query, the more likely an AI answer sits on top of it.

Skip the hunt for AI markup. Google's documentation on AI features and your website says you don't need new machine-readable files, AI text files, or special schema to appear in these features. Structured data is still useful for normal SEO, so keep the markup you have and stop looking for a secret AI version. Treat llms.txt as optional.

Evidence beats adjectives. Bing's own guidance on its AI Performance report says examples, data, and cited sources build trust when content is reused in AI answers. A sourced number gives an engine something to justify the citation with.

What I'd cut or rewrite

Voice search got a full section in the CXL guide. I'd cut it to one line. Voice assistants still read answers aloud, but the work is the same work you'd do for AI Overviews, and I wouldn't build a separate strategy around it.

I'd also stop treating featured snippet counts as a headline AEO metric. Snippets still exist. Your Search Console AI impressions, your Bing citation share, and the answers you check by hand tell you far more about whether your content is being used.

The zero-click section needs a sharper conclusion too. Traffic is a lagging and partial signal now. SparkToro's advice is to replace traffic as a KPI with a correlation dashboard. For a small team I'd make it simpler again: pick the 10 questions you want to own, and measure those.

An AEO workflow for a team of one

This is the loop I'd run with five hours a week and no team. Each step feeds the next, and none of it needs enterprise tooling.

1. Mine your own questions

Your best AEO targets are sitting in sales calls, support emails, podcast recordings, and LinkedIn comments. Transcribe them, pull out every question a customer asked in their own words, and group them by topic. Those messy, full-sentence questions are the exact query shape Pew found triggers AI summaries most often. Our guide to using transcripts for content covers the extraction side.

2. Write the answer block first

For each question, write a two to four sentence answer that would make sense if it were the only thing someone read. Put it straight under a heading that mirrors the question. Then build the section around it with context, caveats, and examples. If the answer only becomes clear in paragraph six, it's in the wrong place.

3. Add one thing nobody else has

A number from your own customers, a test you ran, a price you charged, a mistake you made and fixed. Google's guide uses a first-hand review as its example of a unique point of view, against a summary that restates what's already out there. An answer engine looking at 10 near-identical pages needs a reason to pick yours, and recycled advice gives it none.

4. Get cited off your own site

Pew found Wikipedia, YouTube, and Reddit together made up 15% of the sources linked in Google's AI summaries. Answer engines lean on the wider web to decide whether you're credible. For a solo founder that means LinkedIn content with a clear point of view, useful Reddit answers in your niche, podcast guest spots, and reviews wherever your category lives. I'd take one mention in a respected niche newsletter over another page on my own blog any day.

5. Check the doors are open

None of this works if the crawlers can't get in. Google needs a page to be indexed and eligible to show with a snippet, and Search Console now has a control to include or exclude your site from Search generative AI features, so check it's set to include. OpenAI, Anthropic, and Perplexity each run separate bots for training, search, and user-requested fetches, which means a blanket block can lock you out of the answers you want to be in. Our robots.txt guide for AI crawlers walks through which bots to allow.

6. Measure the answers as well as the visits

Once a month, open the Search Console generative AI report and Bing's AI Performance dashboard, note which pages earn impressions and citations, and compare against the previous month. Then run your 10 priority questions through ChatGPT, Perplexity, Gemini, and Claude and write down who gets named. It takes about half an hour, and it turns tracking into an editorial plan. Our guide to LLM tracking shows how to set up that loop without a paid tool.

When a page keeps losing, fix the structure before you give up on the topic. Move the answer higher, add your own data, and restate the claim in the words your customer used when they asked.

Where AEO goes from here

Google says the list of features in its generative AI report will grow, and Bing's new citation tools are still in preview. AI Mode queries have more than doubled every quarter since launch, so the 0.34% figure won't stay small for long.

The fundamentals from the original guide hasn't moved much since May 2025: a clear answer, near the top, backed by evidence, on a page crawlers can reach. What's different in 2026 is that a solo founder can see whether it's working, for free, in the same dashboards the big teams use. That turns AEO into a monthly habit a one-person team can keep.