Jesús Marrone

Index

What is SEO / GEO?
How to start: Text-first strategy
Good practices for GEO optimisation
Prompt examples

What is SEO / GEO?

SEO stands for Search Engine Optimisation. The meaning is to make your content easy to find on search engines like Google. When users perform a search they include keywords, and your goal is that your page reach position nº 1, the most clicked.

Now GEO stands for Generative Engine Optimisation, making your content easy for AI tools to understand and reuse. The term (GEO) was formalised in an academic paper on November 2023. They share a fictitious example, which is the image you can see, explaining what happens when a user prompts: “Things to do in New York?”.

In response A, before GEO, the output tells us to first visit Central Park, then the Statue of Liberty and finally to try the New York-style pizza.

That seems to be a great plan, but if I am the owner of the Pizzeria, maybe I expect the output goes for response B: “Start by tasting the city’s famous New York-style pizza for breakfast”. Honestly, I can’t think of a better way to start the day!

So nowadays users expect direct answers generated by AI. Moreover, they want to iterate with the output to keep exploring the specific topic they are interested in, to get ideas, to personalise the results… and our goal is that those outputs use information created by JRC, in this case.

Even now Google has been introducing AI Overviews in its results, mixing SEO and GEO. In the image you can see the output in blue, and the regular Google results in pink. So both SEO and GEO are mixing, melting!

We need to master both of them. Not to focus only on Google’s algorithm, but to be relevant enough to be included in AI answers.

	SEO	GEO
Goal	Appear in Google top results	Appear in AI-generated answers
Focus	Keywords, links, and ranking	Clarity, structure, factual accuracy, and brand reputation
Audience	Human users	AI systems
Main platforms	Google and other search engines	LLMs (ChatGPT, Gemini, Claude…)

Let’s see in a blink the main differences between SEO and GEO. The goal in SEO is to be in the top results in Google, while in GEO is to appear in the AI-generated answers. How?

For SEO:

Include relevant keywords and terms in your content
Try to be linked from relevant web sites, this way Google’s algorithm thinks your information is valuable
So you climb positions in the ranking

For GEO:

Your content needs clarity, a good structure, relevant and trustful data
Your organization is reliable

So the challenge is not only to rank at the top of search results, but to create structured and reliable content that AI can highlight.

In this video from the film ShortCircuit 2, the robot reads a book in 10 seconds to know who is the murderer (and surprisingly the butler didn’t do it!):

SEO would give you the link to the book, maybe to the page, and then you’ll need to read it.
GEO gives you directly the answer, the specific piece of information you are looking for.

How to start: Text-first strategy

So, how to start? I’ll suggest a text-first strategy, because it is easy to implement. What I mean is this: Always provide a text-based alternative to media (for LLMs).

Videos, images, sounds… should have the text version, so the information can be retrieved. And you can say, OK, but some LLMs read images. That’s right, but there are 2 approaches regarding AI:

1.- Triggered: if you manually upload an image, the GenAI can extract text (but can’t check a video, just transciption or subtitles).

2.- Passive: Don’t expect an AI that goes to you web and crawls it (it won’t scan it) to watch your video, listen to your audio or interpret your image. Ensure all media has a text equivalent.

This is I blog post I made, which includes an infographic. The text highlighted in red is not readable, it is an image. What we can do is to transcript the infographic and paste the text (curated if needed) below the image, so that information can be read for searching tools and LLMs.

For audio, video… we do the same. It is good to have subtitles, but better if we add the time-coded transcript. In tools like Teams, Clipchamp… you can download the automatic transcription and review it before using it.

Users usually prompt in their mother tongue, so AI prefers to give data in the same language. This page is a good example, it displays the whole information in section “Text”, readable by AI.

One advice: avoid relying on a PDF uploaded for core information. AI reads them, but PDFs lack the structured signals that HTML provides (like headings and metadata). If you want to check what’s written in a page, right-click with your mouse and choose: “View page source”, then you’ll see the code but also the text included.

Bad for GEO

Good for GEO

The climate is changing in ways that are really noticeable in a lot of places, and you can see trees falling down and animals moving around in ways that are not what they used to do. And this has been happening for a while now, so it’s really important that we pay attention and do something about it because if we don’t things could get worse. And everyone talks about it all the time, but not much action seems to happen, which is kind of a problem. The situation is complicated.

According to the 2025 JRC Land Use report, deforestation in the Mediterranean region increased by 12% over the last decade, causing shifts in local fauna habitats and contributing to soil degradation.

Targeted reforestation and sustainable land management are essential to mitigate these effects.

One fictious example. AI Slop is the term that’s been used for low-quality or messy outputs. For instance inaccurate, inconsistent, poorly structured, or just long without saying anything, as shown in the example on the left. On the right we have better text, with data and clear information.

Good practices for GEO optimisation

One important consideration is to know that GenAI companies have not disclosed much about GEO, so we aren’t sure which aspects their LLMs value most in content. Well, this is the same thing as Google doesn’t disclose what elements are more valuable for its algorithm.

“Success starts with content that is fresh, authoritative, structured, and semantically clear”

Krishna Madhavan (Principal Product Manager, Microsoft Bing)

Microsoft published an article offering some useful tips (October 2025, Krishna Madhavan): ‘Optimizing Your Content for Inclusion in AI Search Answers’. I’ll share some good practices for GEO optimisation, but remember keep things readable for humans as well!

It is good to add quotations from recognised experts in your organisation, with the name and position, as we can see in the example. And the picture improves your content’s credibility in the eyes of AI systems.

One advise, when naming files it is good to provide keywords, in this case the name of the colleague and his position, so Google and other tools can match that picture with him in the results.

Remember that GEO values when your site is a relevant source. You need to be the Fountainhead, the primary source of information.

You need to structure your content for AI search visibility:

Index
Use different heading sizes, to clearly set which is a section, subsection, etc.
An introduction on the topic is good for the machine to understand what’s going to be said, as well as a conclusion with key points and insights.
It is preferable to set 1 idea in each paragraph, so we provide a usable chunk of information
Avoid long walls of text, it is boring for human reader and also not clear for AI.
Bullet points, tables (comparing) and summaries also work well.
Give one final conclusion

When you prompt:

The GenAI analyses the information it knows and it can retrieve, to make structure pieces of information
It evaluates each data and checks the relevance of the web page
Then it provides an output which seems to be coherent. The goal is that our institution and information are in those responses.

AI search is less about ordering entire pages and more about which pieces of content earn a place in the final answer.

We can say that semantic clarity boosts your content visibility

It’s important to use a precise language, with consistent terminology. If you use a term with a meaning once, don’t confuse the machine with other meanings. You can use synonyms and related terms, this helps AI connect concepts.

As I said before, avoid vague language and overloaded sentences. Clear writing is a win!

Remember to add context so each paragraph is snippable (so it makes sense even when pulled out of the article). AI systems don’t just scan for keywords; they look for clear meaning and consistent context.

Furthermore, measurable data like figures, percentages, locations, dates… are important to set: when and where was shared the information, and what specific insights it brings.

Some advices to avoid common AI-generation pitfalls

I’m sure you are aware of hallucinations, so it is really important to verify facts and all informations.
Please double-check it, because maybe the information is outdated and you need to check for new data.
As we saw in the example of AI slop, the output could produce superficial content, so it’s good that you add your original value!
Finally, it could happen that the response lacks strong opinions. You can add your impressions sustained on the information you have gathered. Don’t hesitate to use human intuition.

I don’t want to overwhelm you with many acronyms, I just share this one but it’s better to retain that questions and answers work really well for GEO. This structure goes fine with what I’ve told you before about creating chunks of information, snippable content, reusable segments that AI can copy.

You can see that even Google started using it some years ago. Because direct questions with clear answers reflect what people want. With this, we try that GenAI outputs the same answer we have provided in our page.

Prompt examples

Here’s a long prompt template you can copy and paste, with some placeholders in gray, so you can adapt it to your case.

✨ GEO long prompt template 💾

You are an expert in Generative Engine Optimisation (GEO), focused on making content visible, understandable, and accurately represented within AI-generated answers.
Your task is to create an article designed to be cited by LLMs when generating responses.
Check the document attached.
More context: [add who the target audience is, more information, what to focus on, the length…]
Structure the content in a way that is easy for LLMs to extract as a direct response.
Write in a natural way avoiding excessive use of technical language.
Use an index, headings, an introduction and bullet points to make the content easier to read and more structured.
Add relevant quotations if possible
Add a FAQs section.
Do not explain what you’re doing, just give me a publish-ready piece.

I used this pdf and ChatGPT produced the output you can see in the image. Remember that these screenshots are illustrative examples. It may vary between users, even when using the same prompt.

We can see that it has created an index, different sections, so you can get inspired and of course, keep iterating to get other ideas or insights. Please, don’t forget to oversight the information, don’t just copy/paste

Another example-

You can add your own ideas to this prompt, depending on what you expect in the reply.

✨ GEO Q&A prompt template 💾

Use the attached document or this information/link: xxx
Write 8-12 questions for GEO, it is fine to repeat key terms in each one
Cover: key findings, implications, insights, risks, limitations, and next steps.
Provide concise, plain-language answers grounded strictly in the report.
Avoid speculation unless the report explicitly supports it.
Include sections when needed

In this case I want to create a Questions and answers section from this link, a format known to perform well with LLMs. I emphasised that it is fine to repeat key terms in each question, so AI has complete context when snipping.