Solutions
Services
Company Cases About Blog Partners Careers Free session
RUEN

SEO & AI · Practical guide

GEO/AEO in 2026: how to get cited by ChatGPT and Perplexity

15 June 2026~8 min readIMPULSE
What an AI answer with a cited source looks like — and why not everyone gets in.
What an AI answer with a cited source looks like — and why not everyone gets in.

Ask any customer how they last chose a supplier or product. Increasingly the answer is: "I asked ChatGPT and it suggested three." At that moment your site is either in the answer or it is not. There is no third option — and the click has nothing to do with it.

SEO is not dead, but it has a neighbour taking a growing share of attention. Some queries now end directly in the assistant's answer, with no visit to a site. That changes the goal: we used to compete for a position and a click, now we compete to be quoted inside the answer. Here is what that takes.

252,000trials in a factorial experiment across six language models — establishing that relevance and position in context are what decide
3 of 54method-and-domain combinations produced a significant gain in the C-SEO benchmark. In question-answering tasks: none
150–200opening tokens of a page carry disproportionate weight when a source is selected

SEO, GEO, AEO — the difference in plain terms

SEO gets your site into classic Google results. A person sees ten blue links and chooses. Nothing new here.

GEO, Generative Engine Optimization, is preparing content so you get included in generative answers: ChatGPT, Perplexity, Google AI Overviews. There is no position #1 here, only a fact — you were mentioned or you were not.

AEO, Answer Engine Optimization, is about phrasing. People query assistants and voice devices with whole questions rather than keywords: "who does server-side tracking well in Poland", "what does it cost to launch ads for a D2C brand". Content built as a clear answer to that kind of question gets into an assistant's output far more often than a page stuffed with keywords.

We used to compete for a ranking and a click. Now we compete for a sentence in your text that can be lifted into an answer without rewriting.

Why your site is currently invisible to AI

A typical corporate page is written for someone already on the site: heavy on adjectives, light on facts, structured as "we are the best, here are our advantages". An assistant has nothing to quote from that. It needs a clear fact it can extract and place in an answer: what you do, for whom, with what result, at what price.

We ran into exactly this on the TripLegend project — a strong travel platform that simply did not exist as far as ChatGPT, Perplexity and voice assistants were concerned. The product was there; the AI answers were empty. That is not "bad SEO", it is content that was never built for a machine to retell.

+500%
Organic traffic growth for TripLegend over 8 months. The platform started appearing in AI recommendations for 30% of target queries, and sign-ups doubled.

What to do specifically

GEO/AEO is not magic. It is fairly down-to-earth engineering work. Here is what moves the needle.

Rewrite key pages in question-and-answer form

Not "our tracking services", but a direct question as the heading and a clear answer in the first paragraph. Assistants favour pages where the answer is visible immediately rather than buried on the third screen after the company history.

Give the machine structure through markup

Structured data — FAQ, HowTo, Organization, Service — is the language in which you tell a search engine and an assistant what is written here. Without markup the machine guesses; with it, it knows. This is one reason we implement schema.org even where "a human can obviously tell".

Write long, factual and sourced

Assistants prefer content that can be trusted: detailed material with specifics, tables and links to research. Ten deep guides work better than a hundred thin keyword pages. A fact, a number, a source — that is what gets quoted.

Don't forget Bing

ChatGPT leans heavily on the Bing index. If you are absent from Bing Webmaster Tools and everything there is neglected, you have cut off one of the main routes into answers. It is boring and it works.

In classic search you compete for a click. In generative search you compete to become the source an answer cites. Different sports.
What the evidence does and does not support

Two things are established with high confidence: relevance to the query, and the position of the source in the model's context. A factorial experiment across 252,000 trials in six models showed that moving a source higher in context has more effect than most rewrites of the text itself.

Medium confidence: extractable facts and clear structure make a text easier for a model to use. Also — as everyone adopts the same techniques, the advantage to any one player erodes.

Low confidence: that white-hat optimisation reliably improves visibility across different assistants. Very low: any link between AI citations and clicks, conversions or revenue. If someone promises you a specific percentage, they are ahead of the research.

How to measure something with no clicks

There is no analytics dashboard for this yet, so the measurement is manual and it is fine that way.

Pick ten questions your customer would actually type. Ask them monthly, in a signed-out or temporary chat so the assistant's memory of you does not colour the answer. Record whether you appear, in what position, and which of your competitors appear alongside. Save screenshots. After three months you have a trend, and a trend is a metric.

Do the same in Perplexity, and in Google with AI answers turned on. Different systems draw on different sources, and being present in one says nothing about the others.

If you want this done properly, we run visibility audits across search and AI answers and rebuild pages so there is something to quote.

FAQ

What is the difference between SEO, GEO and AEO?

SEO gets your site ranked in classic search results. GEO (Generative Engine Optimization) prepares content so you are included in generative answers from ChatGPT, Perplexity and Google AI Overviews. AEO (Answer Engine Optimization) is about phrasing content as a direct answer to the whole question a person actually asks.

Is SEO dead?

No. Classic search still drives the majority of traffic for most businesses. But a growing share of queries now end inside an assistant's answer with no click, and that share is not addressed by classic SEO.

What actually gets a page into an AI answer?

On the strongest evidence: relevance to the query, and the position of the source within the model's context. A factorial experiment across 252,000 trials in six models found that moving a source higher in context matters more than most text rewrites.

Does schema.org markup help?

It helps a model understand what is on the page. The evidence here is medium: extractable facts and clear structure make text easier to use. It is not a magic switch, but it costs little and removes ambiguity.

How do I measure something that produces no clicks?

By repeat testing. Ask the same ten customer questions monthly in a signed-out or temporary chat, record whether you appear and in what position, and track the trend. That is the metric — there is no dashboard for it yet.

Why does Bing matter for ChatGPT?

ChatGPT leans heavily on the Bing index for web results. If you are absent from Bing Webmaster Tools and your setup there is neglected, you cut off one of the main routes into answers.

Can an agency promise a specific increase in AI visibility?

It cannot honestly. In the C-SEO benchmark only three of 54 method-and-domain combinations were significantly positive, and none in question-answering tasks. Specific percentage promises are not supported by the research.

How long does it take to see results?

There is no reliable published figure. In our own work, changes to opening paragraphs and structure show up in assistant answers over weeks rather than days, but this varies by topic and by how often the model's sources are refreshed.

Check your visibility in AI answers

We'll run the questions your customers ask through ChatGPT and Perplexity, show where you appear and where you don't, and what to change. Free, within 48 hours.

Get the audit

Sources. Evidence levels for optimisation techniques — a critical review of the field for 2023–2026 published on arXiv, including a factorial experiment across six language models (252,000 trials) and the C-SEO benchmark. The weight of the opening 150–200 tokens — a study by Carnegie Mellon University. TripLegend figures come from our own project data and are specific to that engagement.