The SEO/GEO/AIO Definitive Guide

After working on AI-search optimisation and researching how websites appear in AI-generated answers, my view is somewhere in the middle.

There is a shift in focus, but SEO has not suddenly changed overnight. The fundamentals still matter: your website needs to be accessible, your content needs to be useful, and your brand needs to have authority. What changes is how we structure and present that information.

Traditional search gives people a list of results. AI search tries to give people an answer. That difference is what SEOs need to pay attention to.

Machine-readable Website

Before worrying about AI-specific optimisation, make sure the fundamentals are in place.

Your important content should be available on the page and accessible to search engines and AI systems. Do not make important information unnecessarily dependent on JavaScript or hide it behind interactions that a crawler may not be able to process.

The same basic SEO principle applies here: if a system cannot access your content, it cannot use it.

It is also worth checking your robots.txt file. Make sure you understand which crawlers you are allowing or blocking and why. A study by Search Engine Journal explains how bots can encounter access problems when crawling websites.

Backlinks and Citations

Backlinks are still important. AEO does not suddenly make traditional SEO authority irrelevant.

However, AI search adds another interesting layer: your brand can be mentioned even when another website does not link to you.

A SEMrush study reveals the link differences between traditional SEO and AI-search visibility, including the role of different types of links. An Ahrefs study also found a relationship between brand mentions across the web and visibility in AI Overviews.

The important point is not that links no longer matter. It is that AI visibility can involve more than links alone. Being consistently associated with a topic across relevant websites can also help an AI system understand that your brand belongs to that topic.

Schema Markup: Is It Effective?

We have been using schema markup in SEO for years. It is useful, but it is not a magic button for AI visibility.

Schema gives search engines and other systems additional information about what a page is about. It can help clarify things such as an organisation, product, article, person or other information on a page.

But schema should support the actual content on the page. It should not be treated as a replacement for useful information.

A fun duck-related experiment by Mark Williams-Cook demonstrated how an AI system could pick up information placed in structured data, even when that information was not visible in the main page content.

Does llms.txt Work?

This is an area that I am afraid to speak about.

I tested llms.txt for a client and accidentally added incorrect URLs. A few weeks later, when checking AI visibility data in SEMrush, I noticed those URLs appearing among the cited pages.

That experience convinced me that the file was being picked up somewhere in the process.

However, it does not prove that llms.txt directly caused those citations. There are too many other factors involved to make that conclusion from one test.

So my current position is simple: it is worth testing, but I would not treat llms.txt as an established ranking or AI-visibility signal.

Content Structuring in SEO vs AEO

This is where I think the biggest practical change happens.

Traditional SEO content often starts with a keyword. We find a search term, understand the intent behind it, use the keyword in important parts of the page and build content around the topic.

That approach still works. But when people use AI search, they often ask full questions rather than entering a short keyword.

For example, if someone is searching for "optimize for llms" might instead ask an AI system: "what is the workflow to appear in llms?"

This means we should think beyond the keyword and consider the questions and problems behind it.

For AEO, I would focus on making important information easy to find and understand:

• Answer the question directly below the heading
• Use bullet points where they make information easier to scan
• Use tables for comparisons
• Break processes into clear steps
• Add FAQs where they genuinely help

Clearly explain important products, services, companies and other entities.

Clarity is the keyword here

A SearchPilot case study found a positive organic traffic impact after changing relevant page sections into Q&A formats and adding FAQ structured data.

The test doesn’t prove that question headings alone caused the improvement, but it does support the idea that content structure can affect search performance.

Does Keyword Research Still Matter?

Yes. But I don’t think keyword research should automatically lead to creating a new page for every keyword.

Keyword research tells us where search demand exists. It does not always tell us everything a person wants to know.

This is where query fan-out becomes useful.

Query fan-out is taking a query and fanning it out, as the name suggests.

Instead of taking one keyword, take the topic and look at the questions and concerns around it. Look at PAA’s, your GSC data, forums, and questions coming from your sales team.

For example, instead of simply targeting "technical SEO", you might discover related questions such as what technical SEO includes, how to conduct a technical SEO audit, etc.

Those questions can then shape the content.

Is AEO Actually Different From SEO?

Yes, but not in the way some of the hype around AEO suggests.

You still need the same SEO foundations: crawlable pages, useful content, relevant topics, good internal linking, authority and a website that works well.

2 Cents

I don’t think SEOs need to throw away everything they have been doing for the last decade because AI search has arrived.

Instead, we need to understand how the search experience is changing and adjust where necessary.

Keywords still matter. Links still matter. Technical SEO still matters. Content still matters.

But if a user asks a question and an AI system is expected to provide the answer, the content needs to make that answer easy to find.