Being Visible in AI Search (GEO)

Search behaviour is changing. Some users now type their question not into a search box but directly into an AI tool, and get a single answer instead of ten blue links. Being cited as a source inside that answer requires work different from classic ranking.

Why don't ChatGPT and similar tools show my site?

The three most common reasons: the site content is only generated after JavaScript runs, so the crawler sees an empty page; the answer to the question isn't on the page as a clear paragraph, the information is scattered across images and tables; or there is no machine-readable description of the site at all — who you are, what you do and what you are an authority on isn't written anywhere.

Where GEO differs from SEO

The foundation is the same: technical soundness, correct content, credibility. The difference is in form.

In classic search the goal is to get the user to click to your page. In AI search the goal is to be quoted inside the answer. To be quotable, the information must stand in short, verifiable pieces that are understandable on their own. Information that can't be understood without reading the whole page can't enter the answer.

The practical counterpart: under every main question, have a direct two- or three-sentence answer. Give the answer first, open up the detail afterwards. Introductions stretched out in marketing language don't work in this form.

The structure the machine reads

A few things make a measurable difference:

• The HTML from the server must be full. The page content must be visible even without JavaScript running. Testing this is easy: view the page source; if you can't see the text there, neither can the bot.
• Structured data. Organization, Article, FAQPage and HowTo are the most useful types. They leave nothing about what the page is to guesswork.
• Question-style subheadings. When the sentence people actually type is the subheading, the paragraph under it can be taken as a direct answer.
• Consistent identity information. Founding year, contact details, service description must be the same on the site and in external sources. Contradiction lowers trust.
• llms.txt. A file placed in the site's root directory that describes what you do and your important pages in plain text. Not mandatory and not every tool reads it, but its cost is a page of text.

Measuring is hard, but not impossible

AI tools don't yet provide a regular referral report. Still, there are three routes: tracking bot visits in server logs, asking the tools the typical questions in your service area and seeing who gets shown, and watching the change in the part of direct traffic that carries no source. These don't give exact numbers but they show direction.

The part that shouldn't be rushed

Most shortcuts sold under the name GEO are tactics abandoned years ago in classic SEO, presented under a new name. AI tools ultimately look at the same signals: is the content really helpful, is the source credible, is the information consistent. No adjustment made without a solid foundation gives lasting results.

Which content forms get quoted more

The following forms appear most often in AI tools' answers:

• Definition sentences. Explanations completed in a single sentence, in the form "X is ...".
• Numerical thresholds. Verifiable values like "under 2.5 seconds is considered good".
• Step lists. Numbered, short procedure steps written in the imperative.
• Question-answer blocks. Sections where the question is written in the user's own words with the direct answer right under it.

The common point: all of them keep their meaning when taken out of context. A paragraph that can't be understood without reading the whole page isn't suitable for quoting.

Avoid vague expressions

Phrases like "usually", "it depends on the situation", "it depends on many factors" can't be placed inside an answer. Of course everything has a context; but after explaining the context, give a concrete range. The sentence "a corporate site takes 3-6 weeks" gets quoted; "the timeline depends on the project" doesn't.

Consistency of brand information

When AI tools talk about a business they compare several sources. If your founding year is written one way on your site and another on your social media profile; if your service list is ten items on the site and three on a directory site, a picture of low credibility emerges. Settling this information once and keeping it the same everywhere is one of the most effective steps and needs no technical intervention.

How long until it pays off

As in classic search, there is no instant result here either. Structured data and content adjustments are processed in the tools' next crawl cycle; that is a period measured in weeks. On the other hand, once the foundation is set up correctly, every new page inherits the same structure, so the work isn't repeated.

Simple layout changes that work

Before moving to structured data, there are gains to be had from text layout alone:

• Put the answer first. If a section starts with a question, let the first paragraph be the direct answer. Leave the reasoning and exceptions to the following paragraphs.
• Write subheadings as questions. Make them exactly the sentence people type into the search box.
• Don't turn the table into text; turn the text into a table. Comparative information reads far more reliably as a table.
• Give dates and numbers. "Doubled since 2024" instead of "increased in recent years". Verifiable information is preferred.
• Let every page focus on one topic. A page covering five separate topics isn't counted as an authority on any of them.

Shortcuts to avoid

Stuffing keywords into the invisible part of the text, adding hidden paragraphs to the page that only the bot sees, or duplicating the same content at different addresses are reheated versions of old SEO tactics. AI tools also push such pages down in credibility; they don't pay off even in the short term.

How to run it alongside classic SEO

Thinking of GEO as a separate budget item would be a mistake. Most of the work is already part of good SEO; the difference emerges in prioritisation.

In classic SEO the order of priority is usually: technical foundation, keyword targeting, content, links. Seen from GEO the order shifts slightly: the technical foundation stays the same, but keyword targeting is replaced by question targeting, and structured data comes earlier than link work.

In practice a single content calendar is run. Two questions are asked when preparing every post: which search query does this page answer, and which paragraph from this page remains a correct answer when taken out of context.

Managing what you can't measure

Because AI tools' referral data hasn't matured yet, showing the return of this work with classic reports is hard. Even so, there are three indicators to watch:

• Bot visits. Server logs show which pages AI crawlers visit.
• Brand searches. As you get recommended in a tool, people start searching for your brand directly; this increase shows in the search console.
• Traffic of unclear origin. The change in the share of visits carrying no referral information is an indirect sign.

None gives an exact number. Be cautious of services promising exact numbers in this field.

Frequently asked questions

Should I block AI bots with robots.txt? Not if you want to be visible. Blocking also prevents your content from being cited as a source in answers. This is a matter of preference: some publishers block because they don't want their content used.

Is llms.txt mandatory? No, and not every tool reads it. Since its cost is a page of text, it's still worth adding.

Do social media accounts have an effect? Not directly, but they do in terms of the consistency of brand information. The information on the site and the information on the profiles must not contradict each other.

To see your site's technical foundation you can use the SEO Analysis Tool.