Answer Engine Optimisation, or AEO, is the practice of writing and structuring content so AI tools will find it, understand it, and cite it. Those tools include ChatGPT, Perplexity, Microsoft Copilot, and Google’s AI Overviews and AI Mode.
It’s a real shift in how people find information. But a lot of the advice being sold about it is either out of date or was never verified in the first place. This guide separates the two.
Start Here: AEO Is Mostly Just Good SEO
This is the part most AEO articles get wrong, and it’s worth being clear about before anything else.
In May 2026, Google published official documentation on optimising for its generative AI features. Its position is direct: from Google Search’s perspective, optimising for generative AI search is optimising for the search experience, and is therefore still SEO. AI Overviews and AI Mode are built on Google’s core Search ranking systems and pull from the same index. If your page isn’t indexed and isn’t eligible for a normal snippet, it can’t appear in an AI Overview either.
So for Google, this isn’t a separate discipline. It’s the same discipline, applied with a bit more attention to clarity and originality.
For ChatGPT and Perplexity, there’s more genuine difference, because they use their own crawlers and indexes rather than Google’s. But even there, the fundamentals overlap heavily: be crawlable, be clear, be worth citing.
The honest summary: AEO is about 80% good SEO and about 20% platform-specific technical work. Anyone selling it as a brand new discipline is overstating things.
Step 1: Find the Questions People Actually Ask
Query research still matters. The lens is just slightly different.
Instead of chasing broad, high-volume keywords, look for the specific questions people type or speak into AI tools. Useful sources:
Google’s “People Also Ask” boxes
Search autocomplete suggestions
Question-mapping tools like AlsoAsked and Answer the Public
Your own sales and support calls, which are usually the most commercially valuable and the most overlooked
Prioritise questions starting with what, how, why, when, who, and which.
One caution from Google’s own guidance: don’t create a separate page for every possible variation of a question. Doing that to manipulate rankings falls under Google’s scaled content abuse policy, and it doesn’t work anyway. Google’s systems can match a page to a query even when the wording is different.
Step 2: Answer the Question Early
This advice is old, sensible, and still holds.
Put the core answer in the first two to four sentences, then expand with context and evidence. It’s the inverted pyramid structure that journalists have used for a century. A page that buries its answer under three paragraphs of throat-clearing is harder for a reader to use and harder for a machine to extract.
For example, if your target question is “what is answer engine optimisation,” open with a definition, not a preamble about how the world is changing.
Just be honest about why this works. It isn’t because AI models have a documented preference for short opening paragraphs. It’s because clear writing is easier to extract, easier to verify, and better for humans. That’s a good enough reason.
Step 3: Use Clear Headings
Headings help. Google’s guidance says people appreciate pages organised into sections with headings that give a clear structure. Well-structured content is easier for both readers and machines to work through.
What to do:
Phrase headings as questions where it fits naturally
Match the language of your target query
Make each section self-contained enough to stand alone
Skip clever headings that hide meaning. “The Secret Sauce” tells nobody anything
What not to do: don’t obsess over perfect semantic HTML. Google says explicitly that perfectly semantic HTML isn’t required, though it’s a good idea anyway because it helps screen readers.
You’ll also see advice about “chunking” your content into small blocks for AI. Google says there’s no requirement to do this and no ideal page length. Write for your audience instead.
Step 4: Rethink Your Schema Plan
This is the section where most AEO advice is now badly out of date, so read carefully.
FAQ rich results are gone. On 7 May 2026, Google added a deprecation notice to its FAQ structured data documentation. FAQ rich results no longer appear in Google Search for any site, including the government and health sites that kept the feature after the 2023 restrictions. Search Console reporting and Rich Results Test support ended in June 2026, and Search Console API support ends in August 2026.
HowTo rich results are also gone. They were removed from mobile in August 2023 and desktop in September 2023.
And Google says schema isn’t an AI lever at all. Its 2026 guidance states that structured data isn’t required for generative AI search and there’s no special schema.org markup you need to add. It recommends continuing to use structured data as part of normal SEO, because it makes you eligible for rich results that still exist.
So what should you actually do?
If you already have FAQPage markup, leave it. FAQPage is still a valid Schema.org type, and Google has said unused structured data doesn’t cause problems. Other crawlers, including Bing’s and various AI crawlers, still parse it.
Stop recommending FAQ and HowTo schema as an AEO tactic. It isn’t one.
Do use schema types that still produce results: Organization, LocalBusiness, Product, Article, Event, Video, and Breadcrumb.
Validate with the Schema Markup Validator or Google’s Rich Results Test. Skip Google’s old Structured Data Markup Helper, which Google itself has described as outdated.
Make sure your structured data matches the visible content on the page.
One requirement people miss: to appear in Google’s generative AI features, your site must be indexed, eligible for a normal search snippet, and not opted out of AI features in Search Console. Google added that opt-out control in June 2026, so it’s worth checking nobody on your team switched it on.
Step 5: Write Things Worth Citing
This is where the real work is, and Google’s guidance is unusually specific about it.
Google draws a line between commodity content and non-commodity content. Its own example: “7 Tips for First-Time Homebuyers” is commodity content, built from common knowledge that anyone could have written. “Why We Waived the Inspection and Saved Money: A Look Inside the Sewer Line” is non-commodity content, built on direct experience.
Google says creating unique, useful content will influence your visibility in generative AI search more than anything else in its guide.
Practically:
Write from first-hand experience, not from summarising other articles
Be specific. “Many people find this useful” is worth nothing. A number, a date, or a named example is worth something
Link to primary sources when you make a factual claim
Never invent data. Fabricated statistics get cross-referenced and will damage your credibility
Add relevant images and video, which give you extra ways to appear
A warning worth taking seriously: Google explicitly calls out “seeking inauthentic mentions” as something that doesn’t work. Its ranking systems focus on quality content and its spam systems block manipulation, and its AI features depend on both. If an agency is selling you bulk mentions across low-quality sites, that’s the tactic Google just named.
Step 6: Measure It Properly
Tracking has changed a lot recently, and most AEO advice hasn’t caught up.
Search Console now has a dedicated report. On 3 June 2026, Google launched Generative AI performance reports showing impressions from AI Overviews, AI Mode, and generative features in Discover. You can break it down by page, country, device, and date.
Two limits to know. There’s no click, CTR, or query data yet, so you can see visibility but not what it’s worth. And the rollout started with a subset of UK site owners, so it may not have reached your account.
The old advice was to infer AI visibility from rising impressions and flat clicks in the main Performance report. You don’t need to guess like that anymore, at least once the report reaches you.
For Bing and Copilot, Bing Webmaster Tools added its own AI performance reporting earlier in 2026.
For ChatGPT referrals, filter your analytics for utm_source=chatgpt.com. OpenAI adds it to links automatically.
For manual checks, prompting ChatGPT, Perplexity, and Gemini with your target questions is worth doing, but treat it as a rough signal. These systems aren’t deterministic, so the same prompt can return different sources on different days and different accounts. One check proves nothing. A pattern across many checks means something.
On third-party tools: general brand-monitoring tools like Brand24 and Mention track the web and social media. They don’t see inside AI answers. If you want AI citation tracking, use a tool built for it. And note Google’s own caution here: no third-party tool has access to Google’s internal ranking or AI systems, so be sceptical of any tool claiming to use internal Google metrics.
Mistakes to Avoid
Vague, non-committal answers. Content that says “it depends” without saying what it depends on is hard to use and hard to cite. Commit to an answer, then qualify it.
Chasing deprecated tactics. FAQ schema for rich results. HowTo schema. These stopped working and plenty of published advice still recommends them.
Buying tactics Google has publicly said don’t work. Google’s guidance names llms.txt files, content chunking, AI-specific rewriting, special markup, and inauthentic mentions as things you can ignore for Google Search. Some of these may still help with other platforms, and creating an llms.txt file does no harm, but nobody should be charging you for them as a Google strategy.
Treating AEO as completely separate from SEO. This is the mistake I’d flag hardest, because it’s the one the industry has an incentive to make. Google’s documentation says the opposite. If a proposal asks you to rebuild your content strategy from scratch, ask what specifically it does that good SEO doesn’t.
What’s Coming Next
One area genuinely worth watching is agentic browsing. AI agents that book, compare, and buy on a user’s behalf read your site differently from a search crawler. They look at how the page renders, at the DOM after JavaScript runs, and at the accessibility tree.
That means clean, accessible markup with proper ARIA labels is starting to matter for commercial reasons, not just compliance ones. If your site is a pile of unlabelled div tags, agents will struggle with it.
Where to Go From Here
For the foundations, read our full breakdown of what Answer Engine Optimisation is.
For the ChatGPT-specific technical side, including OpenAI’s crawlers and how to check you aren’t blocking them, read our guide to showing up in ChatGPT answers.
If you’d rather have someone do the work, Digileads’ content strategy services are built around this. We help B2B businesses produce content that’s clear, original, and technically visible to the tools their buyers are using.
We won’t promise you citations, because nobody can. What we will do is make sure the fundamentals are right and you’re not spending money on tactics that stopped working.
