SEO & GEO··8 min read

How to make your website visible to ChatGPT, Alice and Perplexity: GEO for business

More and more people skip the search results and ask an AI assistant instead: “recommend a studio that can build a medical equipment catalog site”. ChatGPT, Alice, Perplexity and Google reply with a ready-made text and name a couple of companies. The work of making sure you are among them is called GEO – Generative Engine Optimization.

Where AI assistants get information about companies

When an assistant answers a question about the current market, it usually does not rely on the model’s memory alone – it searches the web. For that it uses:

  • search indexes – Bing, Google, Yandex. If a site is poorly indexed by search engines, an AI assistant is unlikely to find it either;
  • its own crawlers – for example OAI-SearchBot from OpenAI, Claude-SearchBot from Anthropic, PerplexityBot from Perplexity;
  • third-party sources – directories, reviews, articles, discussions. AI assistants are more willing to recommend what is mentioned beyond the company’s own website.
01Your websiterobots.txt, facts, FAQ, markup, llms.txt
02Indexes and crawlersYandex, Bing, Google, OAI-SearchBot, PerplexityBot
03Other sourcesmaps, directories, articles, reviews
04The AI answer“Have a look at studio X – they have built similar catalogs”

Hence the main conclusion: GEO does not replace SEO, it builds on top of it.

Step 1. Let AI crawlers into the site

Check your robots.txt. Sometimes a hosting provider or a security plugin blocks “unknown” crawlers, and AI search along with them. For a business website it makes sense to open public pages and close only service sections (admin, API, user accounts).

CrawlerOwnerWhy it visits
OAI-SearchBotOpenAIsearch in ChatGPT
ChatGPT-UserOpenAIopens a page at a user’s request
GPTBotOpenAIcollecting data for model training
Claude-SearchBot, Claude-User, ClaudeBotAnthropicsearch, user requests, training
PerplexityBotPerplexitysearch and answers
Google-ExtendedGooglea token allowing use of the site in Gemini
YandexBotYandexthe index behind Alice and Neuro answers

Training crawlers can be blocked if you do not want your texts in training data – it does not affect search. On our own site we allow all of them: visibility matters for a studio.

Step 2. Write in facts that are easy to quote

An AI assistant builds its answer from short statements. “We are the best on the market” is useless: it cannot be checked and there is nothing to quote. Specific facts work:

  • who you are, where you work and for whom: “a web studio in Moscow building catalog sites and web services for B2B”;
  • what you have done: projects with names, links and a description of the task;
  • how you work: timelines, stages, warranty, who owns the result;
  • what you build with: technologies and platforms.

A good test: read the first paragraph of your home page aloud. If it does not answer “who are they and what do they do”, an AI assistant will not be able to answer either.

Step 3. Answer customers’ questions right on the site

People ask assistants the same questions they ask a sales manager: how much, how long, what is included, who owns the code. A “Questions and answers” block on the home page and service pages gives the AI ready-made wording. Each answer is 2–4 sentences, no filler, with specifics.

Step 4. Mark up pages for machines

schema.org structured data tells crawlers what is on a page: an organization, a service, questions and answers, an article, breadcrumbs. The minimum set for a company:

  • Organization or ProfessionalService – name, contacts, address, links to social profiles and directories;
  • Service – on service pages;
  • FAQPage – wherever there are questions and answers;
  • Article – for articles, with an author and a date.

Step 5. Add llms.txt

llms.txt is a text file at the root of the site with a short description of the company and links to the main pages in Markdown. It is a voluntary standard: major model developers have not officially said they take it into account. But it takes an hour to make, and AI agents and developer tools already read it. Example: gletas.ru/llms.txt.

Step 6. Get mentioned beyond your website

The strongest and the slowest step. AI assistants trust sources that confirm each other: a listing on Yandex Maps and 2GIS, profiles on industry platforms, articles and case studies on industry sites, client reviews. Use the same name, description and contacts everywhere.

Checklist

  1. robots.txt does not block AI search; sitemap.xml is submitted to Yandex Webmaster, Google Search Console and Bing Webmaster.
  2. The site loads fast on a phone and its text is visible without JavaScript.
  3. The first paragraph of the home page says who you are and what you do.
  4. There are service pages and case studies with facts.
  5. There are question-and-answer blocks with FAQPage markup.
  6. There is an llms.txt.
  7. The company is described the same way on maps, in directories and on social media.

If you would like us to check your site against this list, send a request – we will send a review within 1 business day.

§ – questions & answers

Related questions.

01

How is GEO different from SEO?

SEO competes for a place in the list of search results. GEO competes for an AI assistant mentioning and citing the company in its ready-made answer. The technical foundation is shared: the site must be accessible to crawlers, fast and clearly marked up. GEO adds an emphasis on clear facts, answers to questions and mentions on other websites.

02

Do I need an llms.txt file?

It is a voluntary format that major AI developers have not officially confirmed as a signal. It does no harm and is quick to make, and some AI tools and agents already read it. So we add it, but do not treat it as the main lever.

03

How soon will the site appear in AI answers?

There is no guaranteed timeline. AI search relies on search indexes (Bing, Google, Yandex) and its own crawlers, so the site first has to be indexed properly. The first changes usually show a few weeks after new pages are indexed.