GEO website promotion: what it is, how we promoted a website

What is GEO promotion in simple words

GEO promotion (Generative Engine Optimization) is the process of ensuring a website appears in the answers of generative AI models: Alice AI, YandexGPT, ChatGPT, Perplexity, Google AI Overviews, and AI search modes. When a user asks a question, the model does not invent an answer out of nowhere: it looks for relevant sources, extracts specific fragments from them, and compiles an answer. The goal of GEO is to be among these sources, rather than just reaching the top of classic search results.

GEO is not a separate discipline that replaces search engine optimization. It is an add-on to it, not a replacement: without indexing in regular search, a page has no chance of getting into the pool of sources from which the model selects fragments for an answer.

Why GEO does not work without SEO and a technical foundation

AI models do not build their own independent index from scratch. Alice AI generates answers based on Yandex Search results, and Google's AI mode relies on the same index as regular search results. If a website does not rank in classic search, it has almost no chance of getting into the model's answer: the model simply has nothing to use as a source.

Hence a simple conclusion: without technical optimization and search engine trust, it is too early to start GEO promotion. An AI model has no reason to cite a website that a search engine does not trust. First the technical foundation and SEO, then the add-on for generative search, not the other way around. We regularly see websites ordering "promotion in AI models" even though their basic technical issues are not resolved, and the first thing we do is return them to this stage, even if the client came for GEO, not SEO.

How we promote websites in GEO

Let's break it down using our promotion case study, without abstractions.

Technical audit

We start with an in-depth technical audit using our own crawler. It checks the entire technical part of the website: errors in schema markup, missing SEO meta tags, duplicate pages, and loading speed. We specifically look at how articles and authors are marked up: disconnected Article and Person blocks hinder not only SEO but also GEO, because the model does not see the connection between the text and the person behind it.

The first months of work are spent resolving technical issues shown by Yandex.Webmaster: broken links, indexing errors, problems with canonical page addresses, and duplicates due to URL parameters. Moving on to content before the technical part is resolved makes no sense: even an excellent article will not help a website that a search engine cannot crawl properly.

Semantics and content

In parallel with technical fixes, we build the semantic core for the website using Wordstat and search volume analysis. Based on it, we write articles, glossary terms, and FAQ pages. This is the content that gives the AI model specific facts to cite: short, precise answers to questions, rather than general discussions on the topic.

Glossaries and FAQs work better for GEO than long reads: the model extracts a short fragment from the text rather than retelling the entire material, which means the fragment must be formulated so that it can be cited without losing its meaning. We specifically ensure that new terms and questions do not duplicate already published materials: we check them through our own keyword cannibalization framework before adding a topic to the plan.

Indexing and recrawling

We immediately submit every published page for recrawling in Яндекс.Вебмастер and Google Search Console. The sooner a crawler visits a page, the faster it gets into search results and starts gaining traffic. Only after this stage can an AI model pick up the page: first classic search, then generative search results on top of it.

How we measure results

We track progress across multiple sources at once because a single tool does not provide the full picture. In Яндекс.Вебмастер, we look at "Website visibility in Алиса AI".

In Google Search Console, a separate report on impressions in generative search features has been available since June 2026. For platforms like ChatGPT and Perplexity, there is no webmaster data at all, so we additionally analyze server logs for requests from AI crawlers.

Case study: Фотон, top 3 in Алиса AI search results

On the Фотон website (foton.ru) we went through the entire cycle. We collected a semantic core for the niche, published new FAQ questions every month, filled the glossary with terms, and wrote articles based on the collected query clusters. Each article had an author with a dedicated page on the website, which separately strengthened E-E-A-T signals for search engines, as we will detail in the next article.

Result: the website entered the top 3 in Алиса AI search results for its topic. The journey took more than a week: first the technical part, then systematic content publication month after month, and only after that did it become noticeable that the model started regularly citing the website as a source.

How much does GEO promotion cost

The GEO services market is still forming, so prices vary greatly. A one-time AI visibility audit from different contractors costs from 15 000 рублей and up, depending on the depth of the check and the size of the website. Retainer support starts from 30 000-50 000 рублей per month and grows depending on the content volume and the number of AI models targeted. For projects entering international markets, the range is wider, from 500 to 10 000 долларов per month. At SEO Eleven GEO promotion is provided as a separate service: we calculate the estimate for a specific website after an audit, with fixed KPIs and a breakdown of work, rather than a single total amount.

Who GEO promotion is suitable for, and who it is too early for

GEO is effective where users ask questions like "how to choose", "what is better", "what is the difference": online stores with a large assortment, service companies, complex B2B products where a purchasing decision requires comparing options rather than a single click on a direct link.

If a website lacks a basic SEO history and has accumulated many technical issues, you should start with the foundation, not GEO. Otherwise, the budget for generative optimization will be wasted: the AI model will simply have nothing to cite, and the money spent on content and markup will not pay off. This does not mean you have to postpone GEO for years; the technical foundation and content strategy can be built in parallel if the process is set up correctly from day one.

The technical foundation and content cover only part of the task. A separate factor that directly affects whether an AI model will trust a source is who is behind the text. In the next article, we analyze how author pages and the principles of E-E-A-T enhance website visibility in AI answers.