How to Rank in Google AI Overviews and ChatGPT: The 2026 AEO Playbook

Google now shows an AI Overview – a synthesised, conversational answer sitting above the traditional blue links – on well over half of all searches, up from roughly a quarter of searches less than two years earlier. For a huge share of queries, that AI-generated paragraph is the only thing most users ever read. If your business is not one of the sources it draws from, you are invisible for that search regardless of where you rank in the links below it.

At the same time, a meaningful and growing slice of research now happens directly inside ChatGPT, Perplexity, Gemini and Claude rather than through Google at all. This is the practical shift behind Answer Engine Optimization (AEO): optimising not just to rank, but to be the source an AI system chooses to read, trust, and repeat.

This is a tactical playbook, not a theory piece – concrete changes you can make to existing pages, in priority order, to improve your odds of being cited.

How AI Overviews and AI Search Actually Choose Sources

Google’s AI Overviews are built on the same underlying index and ranking systems as traditional search, not a separate database – a page generally has to already be capable of ranking reasonably well in classic search to be eligible for citation in an AI Overview. From that eligible pool, the systems favour content that is:

• Structurally clear – organised under descriptive headings that map to real user questions, rather than clever or vague section titles.
• Directly answering, early. A page that states the actual answer within the first sentence or two of a section is far easier to extract cleanly than one that builds up to it through three paragraphs of preamble.
• Fact-dense. Specific numbers, named sources, dates and comparisons extract more cleanly than vague qualitative claims.
• Demonstrably experienced. First-hand detail – a specific project, a real number, a named case – signals genuine expertise over generic, could-be-anyone content.
• Fresh. Recently reviewed or updated content is measurably more likely to be pulled into a generated answer than a page untouched for years, even if the older page still ranks well in traditional results.

ChatGPT, Perplexity and other AI-native search tools work slightly differently – many draw more heavily on real-time web crawling plus their own training data – but the underlying content qualities they reward overlap heavily with the list above: clear structure, direct answers, and demonstrable authority.

Step 1: Restructure Your Highest-Opportunity Pages Around Questions

Open Google Search Console’s Performance report, filter to your highest-impression, lowest-click pages, and look at the actual queries driving those impressions. These are pages Google already considers relevant enough to show – often at a position too low to earn a click, but exactly the kind of page an AI Overview might still lift a paragraph from if the content answers the query cleanly enough.

For each target query, add or rewrite a section with an <h2> or <h3> phrased close to how a person would actually ask the question (“How much does a dental website redesign cost?” rather than “Pricing Considerations”), followed immediately by a direct, two-to-three sentence answer before any elaboration. This single change – question-shaped heading, immediate direct answer – is the highest-leverage, lowest-effort AEO improvement available on an existing page.

Step 2: Front-Load the Answer, Every Time

Traditional web writing often builds context before delivering a conclusion – fine for a human reading top to bottom, actively unhelpful for an AI system scanning for an extractable answer. Restructure key sections so the direct answer comes first, and the supporting explanation, caveats and nuance follow it. If a paragraph currently reads “There are several factors that affect X, including A, B and C, which means that generally Y happens,” rewrite it as “Generally, Y happens. The main factors driving this are A, B and C.”

Step 3: Add Genuine Fact Density

Vague, qualitative claims (“websites load faster today than they used to”) extract poorly. Specific, sourced or measurable claims extract well (“Google’s own research has repeatedly found that as page load time increases from one to five seconds, the probability of a visitor leaving increases sharply”). Go through your top pages and, wherever you have made a general claim, either attach a specific number, a named source, or a concrete example from your own work. Aim to introduce a verifiable fact roughly every 150-200 words rather than clustering all your evidence into one section – AI systems appear to weight fact density fairly evenly across a page rather than only rewarding a single statistics section.

Step 4: Implement or Repair FAQPage Schema

FAQPage structured data (covered in full in our Schema Markup Guide) remains one of the most direct technical levers for AEO specifically, because each question-and-answer pair is already packaged in exactly the shape an AI system needs to lift a self-contained answer. The requirement that trips people up: the schema content must match the visible page content exactly. Audit any FAQ schema you already have and confirm word-for-word alignment – mismatches risk suppression, not just wasted effort.

Step 5: Strengthen Entity Signals (Who Is Actually Saying This?)

AI systems weigh source credibility heavily, and credibility is built partly through consistent entity signals across the web: the same business name, description and details appearing consistently on your website, Google Business Profile, LinkedIn, and any industry directories or press mentions you have. Organization schema with a complete sameAs array (see the Schema Markup Guide) is the technical implementation of this – it explicitly links your website entity to your verified profiles elsewhere, rather than leaving an AI system to infer the connection.

Author identification matters too. A blog post attributed to a named person with visible credentials reads as more authoritative to both human readers and AI extraction systems than one attributed to “Admin” or left unsigned entirely.

Step 6: Add Genuinely Original Data, Images or Examples

Content built entirely from paraphrasing what already exists elsewhere on the web gives an AI system no reason to prefer citing you over the dozen other pages saying the same thing. A specific case study, an original before-and-after example, a genuine number from your own client work, or an original diagram or screenshot all create content an AI system cannot get from a competitor’s page – which is precisely the situation where it is more likely to cite the original source.

Step 7: Refresh on a Real Schedule

Set a recurring quarterly review of your highest-value pages: update any statistic that has aged, replace an outdated example, and update the visible “last updated” date alongside the dateModified field in your schema. Content that is demonstrably current is measurably more likely to be pulled into AI Overviews than content that has visibly not been touched in years, even when the older content still ranks well in traditional search.

Step 8: Track What Is Actually Happening

Google Search Console has begun rolling out AI Overview visibility as a filterable dimension inside the Performance report’s Search Appearance breakdown for properties where it is enabled – check whether it has reached your property and, if so, monitor it the same way you monitor traditional impressions and clicks. Separately, set up a segment in GA4 for referral traffic from chat.openai.com, perplexity.ai, gemini.google.com and similar domains – small numbers today, but a trend worth watching as these tools capture a growing share of research traffic.

What Does Not Work

• Keyword stuffing. AI extraction systems parse meaning, not keyword density – stuffing degrades readability without improving citation odds.
• Thin AI-generated filler published purely for volume. More pages of generic content dilute your site’s overall authority signal rather than strengthening it; a smaller number of genuinely deep, well-structured pages consistently outperforms a large volume of shallow ones.
• Schema without substance. Marking up thin or generic content does not make it authoritative – schema accelerates trust that has already been earned by the underlying content.
• Chasing AI visibility while ignoring traditional SEO fundamentals. AI Overviews draw from the same eligible pool as traditional rankings – a page with poor technical SEO, weak backlink profile or thin content is unlikely to be AI-Overview-eligible in the first place, no matter how well the individual sections are structured.

A Realistic Timeline

Restructuring existing high-impression pages (Steps 1-4) typically shows measurable movement in AI Overview appearances within four to eight weeks, since you are working with content Google already considers relevant enough to surface. Entity and authority signals (Steps 5-6) compound more slowly, closer to the three-to-six-month timeframe of traditional authority-building, because they depend partly on factors outside any single page – overall site authority, backlink profile, and consistency across the web.

Frequently Asked Questions

What is the difference between SEO, AEO and GEO?

SEO (Search Engine Optimization) is the broad discipline of ranking in traditional search results. AEO (Answer Engine Optimization) focuses specifically on being selected as the direct answer inside features like Google’s AI Overviews and voice assistants. GEO (Generative Engine Optimization) is the closely related practice of getting cited and recommended by generative AI tools like ChatGPT, Perplexity and Gemini. In practice the three overlap heavily and most of the tactical work – clear structure, fact density, strong entity signals, genuine expertise – serves all three at once.

Do I need special schema markup to appear in AI Overviews?

Google has stated no special schema is strictly required. Clear, well-structured, semantic content matters more, though complete, accurate schema does make it easier for Google’s systems to verify and extract your content confidently – see our full Schema Markup Guide.

Can a small business realistically compete for AI Overview citations against larger competitors?

Yes, more realistically than for competitive traditional keyword rankings in many cases. AI systems favour clear, direct, well-structured answers over sheer domain authority for a specific factual query – a smaller site with one genuinely excellent, precisely structured page on a specific topic can out-compete a larger site’s more generic coverage of the same question.

How do I know if my content is currently appearing in AI Overviews?

Check whether Google Search Console has enabled the AI Overview filter under Performance and Search Appearance for your property. Beyond that, manually searching your target queries (from a logged-out browser, since results vary) and noting whether an AI Overview appears and cites you remains the most direct check available today.

Will optimising for AI Overviews hurt my traditional SEO rankings?

No – the practices involved (clearer structure, direct answers, fact density, better schema, fresher content) are the same practices that support strong traditional SEO. There is no meaningful tradeoff between the two.

How much content do I need to rewrite to see results?

Start narrow. Identify your five to ten highest-impression, lowest-position pages in Search Console and apply the eight steps above to those specifically, rather than attempting a sitewide rewrite at once. Concentrated effort on pages Google already considers relevant produces faster, more measurable results than spreading the same effort thin across an entire site.

Does page speed affect AI Overview eligibility?

Indirectly but meaningfully – since AI Overviews draw from the same eligible pool as traditional rankings, and page speed remains a documented ranking and user-experience factor in that pool, a slow site reduces your odds of being in the eligible set to begin with. See our Core Web Vitals guide if this has not been audited recently.