GEO and AEO: How to Optimize Your Website for ChatGPT, Alice, and Perplexity in 2026

In 2024, ChatGPT added web search. In 2025, Yandex Alice started answering complex queries by citing sources. By 2026, the share of search sessions that end with an AI-generated answer — rather than a click through to a website — is estimated at between 15 and 30 percent. That means one in five users who would previously have found you through organic search now gets their answer directly from an AI.

The question for any business: is your website being cited in those answers? Or is AI describing your market in the voice of your competitors?

The difference: SEO, GEO, and AEO

The terminology is new, so here is a quick overview.

SEO (Search Engine Optimization) — what we have known for 20 years. The goal is to reach the first page of Yandex or Google results: 10 blue links the user can click.

AEO (Answer Engine Optimization) — optimization for direct answers. This covers "featured snippet" and "quick answer" blocks in Yandex, voice answers from Alice and Siri. The goal here is not a click — it is to have the search engine pull a short excerpt from your article and display it as the answer.

GEO (Generative Engine Optimization) — optimization for generative AI. This refers to answers from ChatGPT, Perplexity, Gemini, Claude — situations where an AI synthesizes an answer from multiple sources and cites them. The goal of GEO is to become one of those sources.

AEO came first — back in the era of voice assistants. GEO is a product of the last two years. There is significant overlap between them, but optimizing for Alice and optimizing for Perplexity are different tasks.

Why classical SEO does not cover these needs

A classical SEO page is designed for a user who will read it in full — or at least scroll to the section with the answer. The text has an introduction, development of the topic, and conclusions. The structure is built for a person reading from top to bottom.

AI search engines work differently. They do not "read" an article — they extract fragments from it. The easier it is to pull out a short, precise answer, the higher the chance that fragment will appear in an AI-generated response.

This changes how content should be written:

Direct answer in the first paragraph. If an article answers the question "what is X," the very first sentence must be a definition. Not "in this article we will explore X." Not "X is a complex topic." But — "X is [definition]."
Short, thesis-driven sections. Long paragraphs get compressed or ignored by AI. A 2–3 sentence paragraph that fully develops a single idea is more likely to be cited.
Lists and tables. Structured data is parsed more accurately than continuous prose. If you have three ways to solve a problem — format them as a list rather than embedding "first, second, third" inside a paragraph.
Numbers and facts with sources. AI favors specificity. "About 70% of users" is a weak signal. "According to Yandex data for Q1 2026, 73% of mobile queries in Russia" is a strong one.

What matters for appearing in AI answers

Each generative search engine works slightly differently, but the core principles are shared.

Domain authority. AI systems trust websites that are linked to by other authoritative resources. Here, GEO converges with classical SEO — a link profile matters in both directions. A fresh domain with no reputation rarely appears in AI answers.

Relevance to a specific question. Unlike SEO, where a single page can rank for a cluster of 100 related queries, GEO is precise. AI looks for the page that answers exactly the question the user asked. The more closely the title and subheadings match the phrasing of real queries, the higher the chance of being cited.

Structured data. Schema.org markup has become critical. FAQ Schema, Article, HowTo, Product — these are what AI search engines read first to quickly understand what a page is about. Without markup, competitors who have it will receive priority.

Freshness. AI avoids sources with outdated information. For fast-moving topics — technology, markets, legislation — an article from 2022 will almost never appear in a response, even if it is comprehensive. Regularly updating content with a visible "last updated" date is an essential practice.

E-E-A-T signals. Experience, Expertise, Authoritativeness, Trustworthiness — a Google concept that AI has inherited. An article with a named author, a stated title, years of experience, and links to a professional profile sends signals that improve citation rates. Anonymous text from "the editorial team" performs worse.

How GEO differs across platforms

ChatGPT (with web search). Uses Bing as its search backbone. This means that to appear in ChatGPT answers, you need to be indexed in Bing — not just in Yandex. For the Russian market, this is often overlooked, and that is a mistake: ChatGPT works in Russia and its audience is growing.

Perplexity. The most transparent of the AI search engines — it cites sources directly alongside the answer, with clickable links. This means that Perplexity traffic is real and measurable in the form of actual clicks. Its algorithm favors fresh sources and structured content.

Yandex Alice. Uses Yandex search as its backbone. This is good news for Russian-language websites — if you rank in the top 3 on Yandex for a query, your chances of being cited by Alice are high. Alice favors short answers pulled from FAQ blocks and tables.

Google Gemini / AI Overviews. The most unpredictable in terms of ranking logic — Google updates AI Overviews several times a year. Websites with a strong E-E-A-T profile and structured markup consistently perform well.

Can GEO traffic be measured?

AEO is easier to track — Yandex Metrica and Google Analytics 4 both have reports on "zero-click" sessions and featured snippet impressions. You can see how often your content is displayed without generating a visit.

GEO is harder to measure. Perplexity generates referral traffic (visible in analytics as traffic from perplexity.ai). ChatGPT, Claude, and Alice typically do not send direct traffic — they answer the user, and the session ends without a site visit. The effect of GEO can be estimated indirectly: growth in brand recognition (branded queries in Yandex), social media mentions, and contacts from users who say "I found out about you through AI."

This changes how SEO projects should be evaluated. In 2026, it is no longer enough to look only at "organic traffic" — invisible impressions within AI answers must also be counted as part of your reach.

What to do right now

If you already have an established website with an SEO history, the following steps are enough to start transitioning to GEO/AEO:

Rewrite the introduction of each key article — add a direct, thesis-style answer in the first paragraph.
Add FAQPage markup to any pages that contain question-and-answer sections.
Add Article + Author markup to your blog.
Structure your content — short paragraphs, lists, and tables wherever appropriate.
Display publication and last-updated dates, and refresh your materials at least once every six months.
Connect to Bing Webmaster (for ChatGPT) and Yandex Webmaster (for Alice).
Once a month, check what AI search engines answer for your key queries. Are they citing you or your competitors?

If you are building a new site — design it for both purposes from the start. A modern website in 2026 must perform equally well in Yandex search results and in a Perplexity answer. This is no longer an optional extra — it is a baseline requirement.

We build websites with GEO/AEO built in from the ground up, and we optimize existing ones — let's discuss your project.