GEO optimisation: what it is and what to do first

GEO (Generative Engine Optimization) is work on your website and on what others write about you, so that generative search engines use you in their answers. A generative search engine is a tool that, instead of a list of links, puts together an answer: ChatGPT with web search, Perplexity or AI Overviews in Google. A customer asks “which company in Bratislava does X” and gets a few names. GEO is about whether you will be among them.

In practice GEO and AEO overlap so much that there is no point telling them apart. We explained the difference between AEO and SEO in the article What is AEO and why SEO is not enough. This article is practical: what to do first and in what order.

What is GEO optimisation in one sentence?

GEO optimisation means that an AI tool can read your website, understands from the code who you are and what you do, finds a direct answer to the customer’s question on the website, and sees the same data about you in other trustworthy places too. When one of these four things is missing, the model is more likely to reach for someone else, even if you rank high in Google.

How does GEO differ from SEO?

SEO competes for a position in a list of links. GEO competes for the model to use you as a source and mention you in its answer. The technical foundation is shared: a fast website, clean HTML, correct headings. On top of that, GEO puts emphasis on three things that traditional SEO often pushes aside:

• company data in machine-readable form (structured data),
• a form of content that can be cited as a standalone answer,
• mentions outside your own website: reviews, directories, articles, profiles.

What the difference looks like in practice is shown in the article Why ChatGPT does not recommend you, even if you rank first in Google. A good position in search is not the same as a recommendation in an answer.

What should you do first?

Below are five steps ordered by the ratio of effort to impact. The first two can be done in a day; the last two are ongoing work.

1. Let AI crawlers onto your website

Check the robots.txt file (you will find it at yourwebsite.sk/robots.txt). If it contains a block for GPTBot, PerplexityBot, ClaudeBot or Google-Extended, your customers’ tools cannot see your content. Someone often put the block there years ago, or a security plugin added it, and nobody knows about it.

The second part of this step: turn off JavaScript in your browser and open your homepage. What you cannot see, an AI crawler often cannot see either, because many of them read only bare HTML. A website that is assembled entirely in the browser is a blank page to them.

2. Tell machines who you are

Your website’s code should contain structured data in JSON-LD format: company name, type, address, phone, e-mail, company ID number (IČO) and links to profiles (LinkedIn, Google Business Profile, directories). Visitors do not see it, but search engines and AI tools use it to build a picture of the company.

What matters is that the data match to the letter everywhere there is something about you: on the website, in the Google profile, in the business register, in directories. When a company has three different addresses or names in three places, the model is not sure it is the same company and may rather leave it out.

3. Rewrite content as questions and answers

AI tools look for a sentence they can use as an answer. It helps them when:

• the heading is a question the customer asks (“How much does … cost?”, “How long does … take?”),
• right below it is a direct answer of 40–60 words with a specific number or fact,
• prices, deadlines and parameters are in the text or in a table, not in a PDF or an image,
• the article has a named author and a date, so it is clear who stands behind the claim and whether it is current.

There is no need to rewrite the whole website at once. Start with the service pages and the three most common questions you get by e-mail or over the phone.

4. Make sure others write about you

This is the slowest and at the same time the strongest step. AI tools recommend companies confirmed by several independent sources: Google reviews, profiles in company and agency directories, articles in trade media, case studies with the client’s name, talks.

What can be done straight away:

• ask satisfied clients for reviews and reply to every one,
• complete your Google Business Profile: the right category, services, photos,
• create a company page on LinkedIn and link it to your website,
• publish at least one case study with a concrete result.

Do not buy reviews or links. It is against the search engines’ rules and against common sense, because trustworthiness is exactly what GEO builds.

5. Measure it with your own set of questions

In AI tools there is no “first place”. What is measured is in how many answers from a fixed set of questions you get mentioned. Write down 10 to 20 questions your customer would ask (service + city, comparison, price) and once a month enter them into ChatGPT, Perplexity and Google. Note who was mentioned and whether the data about you are correct. Answers differ with every query, so one test is a sample, not statistics. A trend over three months already tells you something.

What does not belong to GEO?

• Hidden texts and “prompts” for AI on the website. Attempts to write an instruction for the model into the page are manipulation and will damage trustworthiness.
• Mass-generated articles without original content. The model is looking for a source that adds something new, not the hundredth version of the same thing.
• Promises of a guaranteed position in ChatGPT. Whoever makes them is promising something outside their control.

How do you find out where you stand now?

The quickest thing is to measure the technical part. Our free full website audit (in Slovak) also checks data for AI (GEO) on the homepage: structured company data, content without JavaScript, access for AI crawlers and the llms.txt file. On top of that it adds speed, SEO and a check of marked-up questions. You see the result within a minute, for free. The score is a technical indicator, not a measure of whether AI cites you, and neither llms.txt nor structured data guarantees a recommendation.

If you want us to do it with you, from testing questions in AI tools through adjusting content to measurement, you will find the description under the AI visibility service. On a short call we will also tell you honestly what you can manage on your own without us.