GEO stands for Generative Engine Optimization: preparing a site so AI systems use it and cite it when they answer. It does not replace classic SEO, which still decides what gets crawled and indexed. It changes the target: instead of competing for a click in a list of ten results, you compete to be the source the AI names in its answer.

What is the difference between SEO and GEO?

SEO optimises for appearing in a list where the user chooses. GEO optimises so that a system writing the answer finds a specific, verifiable, attributable claim on your page and cites it. In practice they reward different things: direct self-contained answers, sourced figures, and technical access for AI crawlers, which are not the same as Google's.

What are AEO and GEO?

Two names for overlapping practices. AEO, Answer Engine Optimization, focuses on the page answering specific questions in an extractable way. GEO, Generative Engine Optimization, also covers being cited as a source by systems that generate text. Day to day you do both at once and the checks are nearly identical.

Blog · 2 September 2026

GEO SEO: how to check whether AI can cite your site

I searched for “GEO SEO” and read the top ten results. All of them explain what the acronym means. They are published by consultancies, business schools and the corporate blogs of a bank and a telecoms operator.

And above those ten, Google places its own generated answer. So the definition is already handled, free, without anyone clicking. Writing the eleventh explanation of what GEO is means competing for a position that no longer sends visits.

What I could not find anywhere: how to check whether your site qualifies. That part is measurable, and it is what follows.

What actually changes

In classic search you compete for a click inside a list. In a generated answer there is no list: there is a paragraph with two or three cited sources. Either you are among those sources or you do not exist for that query.

That rewards different things. It does not matter that your page is the most thorough if the sentence answering the question is buried in paragraph twelve, with no figure and no source.

The five things you can check

1. Are you letting AI crawlers in? They are not Google's. GPTBot, ClaudeBot, PerplexityBot and Google-Extended are separate agents, and plenty of sites block them without knowing because they copied a robots.txt from another project. It is the first thing I look at and where the problem usually is: nothing else matters if the door is shut.

2. Is the content in the HTML, or painted by JavaScript? Google renders JavaScript. Several AI crawlers do not, or do it worse. If your text only appears after the script runs, the page can look nearly empty to them. You check it by looking at the HTML as delivered, without executing anything.

3. Are there extractable answers? A question as a heading, and below it a direct answer of forty to sixty words that stands on its own without depending on the previous paragraph. That block is what a system can lift and cite. Excellent prose that only makes sense read end to end is hard to quote.

4. Do your claims carry a figure and a source? “Meta stopped charging for service messages on 1 November 2024” is citable. “Costs have changed recently” is not. The difference is a date and a link to the original documentation.

5. Is it clear who signs it? An identifiable author, structured data saying who they are and what they do, and consistency between what the page claims and what can be verified elsewhere. Systems that cite tend to prefer attributable sources.

How to check it without tools

Open yourdomain.com/robots.txt and look for GPTBot, ClaudeBot or PerplexityBot in any blocking rule. It takes a minute and settles the most important question.

Then view the source of your main page, without running scripts, and check the text is there. If you see an empty shell, you have a problem no content strategy will fix.

Finally, ask ChatGPT or Perplexity directly about your topic and see who they cite. If the same three competitors always come up, you know who you are up against and what format is being rewarded.

What does not work

Padding the page with artificial questions and answers to “look citable”. Question blocks help when they answer real doubts; as filler they lengthen the page and give nothing worth quoting.

Vague text written to cover many keywords does not work either. It is the opposite of what a generative system needs: specific claims it can attribute without getting it wrong.

And a warning worth hearing from someone not selling you the service: this does not replace SEO. If Google does not crawl or index you, the rest is moot. What matters about building systems that actually work applies here too, and I cover that in AI agents: what you learn once they start selling.

Where I would start

With robots.txt, which is free and may be blocking everything else. Then with the three questions your clients ask most: each one as a heading, with a direct answer below, carrying a figure and a source.

That is worth more than rebuilding the whole site, and it shows within weeks.

I post audits and examples on @WhatsMarketing_es.

If you want me to audit your site and tell you which of the five you are failing, get in touch. I work from Buenos Aires, originally from Málaga, with clients in Mexico City, Argentina, the rest of Latin America and Spain.

← See all articles