
GEO stands for generative engine optimisation, and at its core it is nothing more than SEO aimed at ChatGPT, Perplexity and the AI overviews in Google rather than at the traditional blue links. Where SEO is about a high position in the search results, GEO is about something subtler, namely whether you are mentioned or cited when an AI system composes an answer. That distinction matters, because a page outside Google's top ten can still show up prominently in an AI Overview as long as the content is clear and well structured. In this article we explain how that works and what you can do about it this week.
What GEO actually involves
Traditional SEO optimises for an algorithm that ranks pages by relevance and authority. GEO optimises for a language model that summarises an answer and selects sources to cite along the way. The model is not looking for the best position, it is looking for the clearest and most reliable text to build its own answer on.
That changes what "good content" means. A long, comprehensive article with plenty of keywords works fine for Google, but a language model prefers a short, sharply worded answer it can lift almost verbatim. Question-led headings such as "what does ... cost" or "how does ... work", followed by a direct answer in three to four sentences, demonstrably perform better than a long introduction before you get to the point.
Why this matters for your business now
More and more people put their orientation questions directly to ChatGPT or Perplexity instead of Google. For many small and medium-sized businesses that means a new channel with, quite simply, little competition yet, because most websites are still optimised purely for Google. What we often see with business owners is that they score well on classic keywords but are never mentioned when someone asks ChatGPT to name three providers in their industry. That is exactly the gap GEO tries to close.
This is not a completely new discipline you have to learn from scratch. Most of the things AI systems value are the same things a visitor values: a clear answer, a trustworthy sender and information that is correct. So GEO is less a technical exercise than a reason to finally clean up those outdated content pages.
In concrete terms, your site is competing not only with other websites but also with the language model's own answer. If someone asks a simple factual question, ChatGPT often generates a complete answer itself without citing a source. There is little to win on those questions. It is different for questions that ask for a specific provider, experience or opinion, and that is also where most small businesses will find the biggest opportunity.
Five things you can do this week
You do not need to turn this into a technical project. Most steps can be done in an afternoon.
- Check whether AI bots can visit your site. Look in your robots.txt to make sure crawlers such as OAI-SearchBot are not blocked by accident, and remove unnecessary noindex tags on important pages.
- Put the answer first. Start a section with the concrete answer in three to four sentences, and only then build up the context. Readers and language models both appreciate that.
- Write about specific customer questions. AI systems already summarise generic topics perfectly well on their own. A concrete problem your customer has, together with your approach to it, is far harder to replace with a standard answer.
- Show who the author is. A name with relevant experience, backed up with a real-world example or screenshot, builds the kind of credibility language models are sensitive to.
- Refresh your oldest content. A three-year-old article with a visible update date is seen as more reliable than the same article without any sign of maintenance.
Structured data such as FAQ or Article schema can help, but only if it matches what is actually on the page. Adding an FAQ block that adds nothing to the content itself is pointless.
A myth that can cost you time
Since the start of this year a lot has been written about the llms.txt file, a kind of robots.txt specifically for language models in which you would indicate which content matters. It sounds appealing as a quick fix, but at the moment the major AI search systems largely ignore this file. Spending time on llms.txt before you have basics such as crawlability and direct answers in order is getting the sequence backwards. Foundations first, extras later.
A page outside Google's top ten can still appear prominently in an AI answer. Position in Google and visibility in ChatGPT are two different contests.
How to measure whether it works
The tricky thing about GEO is that there is no Search Console for it that shows you exactly when you are mentioned. There are now specialised monitoring tools, such as Otterly.ai, that track how often your brand appears in answers from ChatGPT, Perplexity, Google AI Overviews and Copilot. Otterly starts at around 29 dollars per month, which comes to roughly 26 euros. For a small business that is often not the first investment to make, but manually asking ChatGPT and Perplexity the same questions a few times a month and noting whether you are mentioned already gives a decent picture.
Make it a fixed moment, for example the first Monday of the month, and ask the same three to five questions a potential customer would realistically ask each time. That way you build your own rough trend line of how often you are mentioned, without any tool.
Read also: How to get AI search engines to cite your LinkedIn content, the same principles applied to LinkedIn.
Frequently asked questions
Do I need to rewrite all my content for GEO now?
No, start with your best-performing or most-visited pages. Restructure the opening there into a direct answer and only then look further. Tackling everything at once is usually unnecessary and takes far more time than it needs to.
Does GEO rule out SEO, or can you do both at once?
They largely overlap. A page that is technically sound, clearly structured and written with clear authority usually performs better on both fronts. GEO is more a shift in emphasis than a completely different discipline.
Does GEO also work for a local business without a content-heavy website?
Yes, although the gains are usually smaller there. At the very least, make sure your basic information, such as opening hours, location and offering, is accurate and up to date on your site. That is exactly the kind of fact language models like to cite, even if you do nothing else about GEO.


