In 2026, writing for Google and writing for AI assistants are surprisingly similar. Both reward text that answers quickly, is easy to scan and can be trusted. The difference is in the detail: an AI assistant doesn't take your whole article. It takes one sentence, table or paragraph that answers the user's question. In this article I show how to write content for AI search so it can be quoted, while still performing well in classic Google.
Content that AI assistants cite puts the answer right at the start (3 to 5 sentences with concrete facts), uses headings phrased as questions the way people actually ask them, short paragraphs with one idea each, tables for comparisons and numbers with a source. Add what AI can't find anywhere else: your own experience, your own data and a clearly named author. The update date matters too, because both AI and Google prefer current information.
How is writing for AI different from classic SEO?
| Classic SEO | AI search (GEO) | |
|---|---|---|
| What gets evaluated | the whole page and its ranking | individual blocks of text that answer a question |
| What the user sees | a link and description in the results | an answer assembled from several sources, with your site as a citation |
| What decides | relevance, links, technical health | clarity of the block, trustworthiness, agreement with other sources |
| How it's measured | rankings, clicks, impressions | mentions and citations in answers, visits from AI assistants |
The good news: the rules don't conflict. Text written the way this article describes helps in both channels. GEO (generative engine optimization) isn't a replacement for SEO, it's an extension of it.
Before you write: what question are you answering?
The most common mistake isn't style, it's the topic. An article that doesn't answer any specific question gives AI nothing to cite. So before you write, jot down one sentence: What question does this text answer, and who is asking it?
Pick the format based on the type of question:
| Question type | Example | Best format |
|---|---|---|
| How much does it cost | "how much does a logo cost" | a price table, a range, what affects the price |
| How to | "how to increase website traffic" | numbered steps |
| Which to choose | "Shoptet or Shopify" | a comparison table and a recommendation for each situation |
| What is it | "what is GEO" | a definition in the first sentence, then examples |
| Why isn't it working | "why does my website get no traffic" | a list of causes with a fix for each |
How do you structure an article that can be quoted?
A quick answer right below the intro
Three to five sentences that answer the main question of the whole article. No "it depends on many factors", just concrete facts. If the answer really does depend on circumstances, say which ones and what the typical values are. This block is the most likely candidate for a citation, and it also helps readers who don't have time to read the whole text.
Headings phrased as questions
A heading like "How much does website maintenance cost?" is a clear signal to both AI and Google that an answer follows. A heading like "The financial side of things" isn't. Write headings the way a customer would ask the question, and answer it in the first sentence right below.
One idea, one paragraph
Long paragraphs that mix three topics are hard to read and hard to quote. A short paragraph with one idea is a standalone block that can be used without the surrounding text. If the content has a clear axis (options, prices, before and after), it belongs in a table.
An FAQ section at the end
Question and answer pairs are ready-made material for AI assistants. Pick the questions your customers actually ask you and answer briefly, ideally in five sentences or fewer.
How do you write so that both AI and readers trust you?
- Numbers with a source. Back up every figure that isn't your own measurement with a link to the original source: a government office, documentation, an official price list. A made-up statistic is the fastest way to lose trust.
- Your own experience and your own data. This is the one thing AI won't find on a thousand other websites. What you've seen with clients, what you've measured, which mistakes keep repeating. Even a small measurement of your own is worth more than rewording generic advice.
- A named author. A real name, what the author does and why they understand the topic. An anonymous text from "the editorial team" is a weaker signal for both AI and Google.
- Consistent business information. When your name, address, services and prices differ between your website, your Google profile and Firmy.cz, AI doesn't know what to believe.
- An update date. For topics that change (prices, rules, tools), state when the text was last updated, and actually update it.
What should you do after publishing?
- Link to the article from at least two existing pages that get traffic, and link from the article to the service page you want to sell.
- Check that the text is directly in the HTML. If the content is only rendered by JavaScript, many AI crawlers won't read it. A simple test: view the page source and search for a sentence from the article.
- Don't treat the article as finished. After two or three months, check Google Search Console to see which queries it shows up for. You'll often find people are searching for something slightly different than you expected, and all the article needs is a section that answers it.
- Track visits from AI assistants in Google Analytics (sources like chatgpt.com or perplexity.ai). I explain how and why it's worth measuring in the article AI search is already sending customers.
Can I write content with AI?
You can, as long as you use AI as an assistant, not as the author. Let it help with structure, suggest FAQ questions, proofread or simplify a complicated sentence. But the substance, the experience and the numbers have to be yours. Text generated "on a topic" with no value of its own is exactly what the internet already has too much of, and Google explicitly fights mass-produced content. And an AI assistant has no reason to cite text that just rephrases what it already knows from elsewhere.
What to avoid
- A half-page intro before the reader gets the answer.
- Repeating the keyword in every sentence. Write for people; search engines have understood synonyms for a long time.
- Two articles on the same question. They compete with each other. Better to have one thorough, up-to-date article.
- Generic advice without concrete examples that could appear on any website.
- Outdated information with this year in the title.
Frequently asked questions
What is GEO?
GEO (generative engine optimization) is optimizing content and a brand so that AI assistants like ChatGPT, Perplexity or Google AI Overviews mention and cite it. It builds on the same foundation as SEO and adds an emphasis on clear, independently quotable blocks of text and on brand trust across the internet.
Do I have to write differently for Google and for ChatGPT?
No. Text with a quick answer, question headings, short paragraphs, tables and sourced numbers works in both channels. It's not two strategies, it's one well-written text.
How long should an article be?
Long enough to fully answer the question, and no longer. A short question like "what is GEO" needs a few paragraphs, while a guide to increasing website traffic can easily have ten steps. Length on its own guarantees neither rankings nor citations.
Does an FAQ section help?
It mainly helps readers and AI assistants, because question and answer pairs are easy to use. You generally won't see the special FAQ display in Google results anymore; the value is in the visible text.
How can I tell whether AI assistants cite me?
Ask AI assistants the questions your content is meant to answer and watch whether they mention your brand or link to your site. I describe the process in the article how to find out whether ChatGPT recommends you.
Want content that brings in customers?
If you'd like me to go through your website's content and suggest which pages to improve and which topics to add, take a look at SEO optimization. It covers both the technical and the content side, including preparation for AI search.