How to Rank in AI Overviews in 2026: AEO and GEO for Business

The question of how to rank in AI Overviews makes the most sense as a scene. A sales manager in Manchester is hunting for a firm to outsource the bookkeeping of a small limited company. A year ago he’d have opened Google, skimmed a dozen tabs, and stitched them together in his head. In 2026 he just asks ChatGPT: “reliable bookkeeping outsourcing for a small UK business, what to look for.” Seconds later he reads a tidy paragraph — what to check, the red flags, and a couple of firms it named on its own. He opens one first, not because it’s cheaper, but because the machine decided it could be trusted.

In the office block next door sits a firm that’s just as good — same services, same credentials, prices even gentler. But its website is a nameless “team of professionals,” a services page without one direct answer to a client’s question, a blog two years stale. The assistant never named it — there was nothing for the machine to grab onto.

The good news: the answer is far closer to familiar SEO than the “death of search” headlines suggest. Answer engines don’t conjure sources from thin air — they assemble a response from the same web SEO has always optimised. The foundation hasn’t changed; only the shop window where you’re displayed has.

How to rank in AI Overviews: the rules machines use to pick a source

Google’s AI Overviews, ChatGPT with web search, Perplexity and Gemini all work the same way: they take a query, find a handful of pages they trust, and assemble one answer from them — sometimes with source links, sometimes naming a company in the text.

The key word is trust. The machine pulls from pages that look authoritative, are cleanly structured, and give a self-contained answer to exactly the question asked — the same qualities that rank you in classic search. Two terms:

• AEO — answer engine optimisation. Becoming the source of a direct answer: AI Overviews, voice assistants, “quick answer” boxes.
• GEO — generative engine optimisation. Optimising for generative engines like ChatGPT and Perplexity, which assemble a longer piece from many sources — so you get quoted inside it.

Content that answers immediately: the core AEO move

Most pages bury the point around the fourth paragraph. A human tolerates that; a machine won’t. When an AI looks for the answer to “how much does a website cost,” it wants a sentence it can lift, not prose to wade through.

Hence the core answer-first move: give the direct answer in the first line or two under the heading, then expand. The heading frames the question the way a person asks it; the first sentence delivers the answer.

What makes a paragraph “quotable”:

• Self-contained. It makes sense without the text around it — the machine yanks it out and it doesn’t fall apart.
• A clear definition. One plain sentence in the “X is…” format. Definitions are an AI favourite: they slot in verbatim.
• Specifics over fluff. Ranges, steps, lists, honest numbers. “It depends on many factors” can’t be quoted — and won’t be.
• A question in the heading. H2s and H3s phrased as real queries (“People Also Ask” style) match what people ask an assistant.

That is how we frame every page when we build a multilingual site — a question heading, a direct answer in the audience’s language, then depth.

Structured data: the language you speak to machines in

Structured data markup (Schema.org) tells the search engine and the AI exactly what is on the page. Three types matter most for AEO and GEO:

• FAQPage — marks up a block of questions and answers. The most direct way to tell a machine “here is a ready question-answer pair, take it.” The FAQ below is wrapped this way.
• HowTo — marks up a step-by-step guide and helps an assistant pull the steps out in order.
• Article and Organization with an author — tie the text to a named person and a company with credentials, a signal about whom to trust.

An honest caveat: markup on its own won’t drag you into an answer. It removes the guesswork and raises the odds you’re quoted accurately.

Entities and topical authority: be the topic, not the page

Search engines and language models think in entities — people, companies, services, places with their own properties and connections. To a machine, “Webtor” is an entity: a web agency, Warsaw, services, case studies, mentions elsewhere. The clearer you’re described as one, the more confidently the AI names you on topic.

One strong article makes you a source on a single question. A dozen related, interlinked pieces that cover a topic from every angle make you a source on the whole topic. That is why we keep separate deep-dives on the cost of a website, on SEO pricing, and on local search: together they tell the machine that Webtor is about web and search in depth.

The practical minimum: collect the questions clients genuinely ask — from “how much does it cost” to “what if something goes wrong” — answer each on its own page, and interlink them using one consistent set of terms.

E-E-A-T and freshness: why AI won’t cite a nameless site

E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness. Google formalised these signals, and models look the same way when deciding whom to believe. A nameless “team of professionals” loses on all four counts. What strengthens trust in the eyes of an AI:

• A real author with credentials. Not “Admin,” but a person with a name, a title and experience. For sensitive topics — finance, law, medicine (YMYL, “your money or your life”) — both engines and models look hard at who stands behind the text.
• Signs of genuine experience. Specifics you can’t fake from generalities: numbers from real work, honest caveats, “here’s how we got burned” — the very “E” Google added on purpose.
• Freshness. A real dateModified field and content that is actually updated. An article with 2026 prices and a recent date beats the same article with figures from three years ago.
• Honesty over promises. Ranges, not invented precise statistics. One reason to understand why SEO matters before chasing AI: thin, nameless and outdated content is equally invisible to classic search and to answer engines.

None of this is about tricking the algorithm; it is about being genuinely credible. The AI has simply made that credibility mandatory.

Off-site mentions: you get cited when others talk about you

Models weigh what others say about you. When several independent sources name your company in the right context, the assistant takes you into the answer more confidently — corroboration, not just your own word.

It isn’t about buying links by the sackful, but about being present where your topic is discussed: a complete Google Business Profile, accurate directory listings, on-topic mentions, live reviews. For a local business, maps and directories carry extra weight — the machine often leans on them for “best [service] in [city].” The more consistently a company is named across the web, one name, address and phone everywhere, the sharper you are as a trusted entity.

llms.txt and access for AI crawlers: technical hygiene

llms.txt is an optional text file in your site root (by analogy with robots.txt) where you list your most important pages and describe them briefly, so AI systems can orient faster. It is a suggestion, not a command: models need not follow it, and on its own it won’t drag you into an answer. But as a tidy map of your key pages it does no harm.

Access for AI crawlers. Models roam the web with their own bots (GPTBot, ClaudeBot, PerplexityBot and others). Accidentally block them in robots.txt and you can’t end up in their answers — you simply won’t be read. Decide who gets in deliberately.

And separately — the technical base, without which nothing works. A page that is slow or breaks on a phone loses in normal search and in the race to be cited alike. Core Web Vitals in 2026 means LCP (how fast the content appears), INP (how fast the page responds) and CLS (how much the layout jumps). Foundation first, then citability — the reason Core Web Vitals sit under every other ranking decision.

AEO versus classic SEO: what changes, and what doesn’t

Where AEO and GEO shift the emphasis, and where everything stays as it was:

Aspect	Classic SEO	AEO / GEO in 2026
Goal	Rank higher in a list of links	Land inside the finished answer, get cited
Where you’re seen	The results page, blue links	AI Overviews, ChatGPT, Perplexity, Gemini
Content format	Long-form text for a query	Direct answer first, depth below it
Role of markup	Helps with rich snippets	Helps the machine lift a ready answer
Main signal	Relevance and authority	The same relevance and authority, plus citability
Foundation	Trusted content from the live web	The same trusted content from the same web

The takeaway is the bottom row: the foundation is shared. Anyone who already did honest SEO isn’t starting from scratch — they’re tuning structure and delivery. Anyone who ignored SEO won’t pick up AI traffic on a side road, because the machine has nowhere to take you from.

Zero-click is rising — but hot queries still convert

The owner’s worry is fair: if AI answers right in the results, won’t it skim off the traffic? Some of it, yes — by industry estimates the “zero-click” share in 2026 is noticeably higher than a few years ago. But traffic didn’t vanish, it split. Informational queries like “what is INP” increasingly close on a summary. Queries with intent to act — “build me a website,” “urgent SEO audit” — are the ones a short summary handles worst. For a decision that costs money, people go to a website to compare and to fill in a lead form — nobody hires a contractor straight from a summary paragraph.

And a mention inside the answer works like advertising. When an assistant names your company for “reliable web developer in Warsaw,” it is recommending you to people who’d never have found you otherwise. The goal for 2026 isn’t to hold onto every click, but to be the one who gets named when your topic comes up.

How to measure clicks and mentions from AI

There is no single counter yet, but an honest picture is achievable:

1. AI referral traffic. In your analytics, isolate visits from domains like chatgpt.com, perplexity.ai and gemini.google.com into a separate segment — the most direct signal that you’re not just quoted but actually opened.
2. Search Console. Watch impressions and clicks on queries where AI Overviews appear. If impressions hold but clicks dip, a summary took part of the answer — a cue to strengthen the “hot” transactional pages.
3. A direct test in the assistants. Ask ChatGPT, Perplexity and Gemini your key queries and see whether they name you, your competitors, or nobody. Repeat regularly to track movement.
4. Brand mentions. Track how often and in what context your company is named off-site — a rise in mentions usually runs ahead of a rise in AI citations.

Where to start this week

You can’t rewrite a whole site in a week, but you can shift the needle. In descending order of payoff:

1. Take your five most important pages and rewrite each opening on the answer-first principle: a question heading, a direct self-contained answer in the first lines, depth below.
2. Add a FAQ with FAQPage markup to key pages — five real client questions with short, honest answers.
3. Sign the content with a real author with a name and credentials, and set dateModified where the content is fresh.
4. Check robots.txt so you haven’t blocked AI crawlers by oversight, and add a simple llms.txt with your main pages.
5. Run your key queries through ChatGPT and Perplexity — record where you’re named and where you aren’t, for a baseline next month.

AI visibility arrives page by page. The same logic carries into a website redesign: structure and answers first, polish second.

Who wins in the end

The manager in Manchester opened first the firm the machine named — and the machine named it for direct answers, a real author, fresh figures and off-site mentions, the same things that have long pushed you up in ordinary search.

That is the whole plot of 2026. AI did not abolish SEO; it raised the stakes on credibility and added a second shop window where credibility now gets collected twice — once by people scrolling the results, once by machines deciding whom to quote. How to rank in AI Overviews comes down to the same old thing: be a source people trust. Only now both humans and algorithms feed from one foundation — honest, useful, well-structured content.