AEO and GEO: A Simple Guide to How AI Search Works

I first came across the term Generative Engine Optimization while working at RemotePass. Our search impressions were going up, our clicks were going in the opposite direction, and some of our content had started appearing inside Google’s AI Overviews. 

I wanted to understand what the hell was happening, so I went down the GEO and AEO rabbit hole.

After RemotePass, I joined Xofu, an AI visibility platform by Citation Labs, where I led product marketing. That forced me to understand AI search beyond the surface-level advice you see floating around LinkedIn, and even after leaving, I kept digging into the discipline. I’ve since co-authored the first State of AI Search in Nigeria report and taught an AI Visibility Sprint at Treford, with a course coming soon.

I’m still learning because this field is moving ridiculously fast, but one thing I’ve noticed is how unnecessarily confusing people make it sound. Even marketers who understand SEO can hear terms like GEO, AEO, LLM visibility and retrieval and immediately feel like they’ve walked into a computer science lecture they didn’t sign up for.

So this is my attempt to explain the whole thing in a way my 15-year-old self would understand.

Let’s start with how people search

Imagine you want to know the best savings app.

A few years ago, you would probably open Google and type something like best savings apps. Google would return a list of websites, you would open a few of them, compare what they said and eventually make up your mind.

SEO grew around that behaviour. Companies created pages, Google organised and ranked them, and marketers competed for those rankings because appearing near the top usually meant more people visiting your website.

Now imagine asking ChatGPT:

What savings app would you recommend for someone who struggles to save consistently?

Instead of giving you ten websites to inspect yourself, it can compare the options, mention specific features and recommend something based on what you told it.

The website still has a role, but there is now another layer sitting between the search and the click.

Pew Research Center analysed almost 69,000 Google searches from 900 U.S. adults and found that users clicked a traditional result on 15% of visits without an AI summary, compared with 8% when one appeared. Links inside the AI summary itself were clicked on just 1% of visits. 

Ahrefs found that the presence of an AI Overview correlated with a 58% lower average click-through rate for the top-ranking page in its analysis of 300,000 keywords. 

I saw a version of this at RemotePass, where we became more visible in search while receiving fewer clicks, which meant rankings and traffic were no longer telling the entire story.

So what the hell are AEO and GEO?

AEO means Answer Engine Optimization.

The idea is simple: structure and communicate your information in a way that makes it easier for search and answer systems to use when somebody asks a question. Featured snippets, voice answers and some AI-generated responses all fall somewhere inside that world.

GEO means Generative Engine Optimization.

It focuses more specifically on increasing the chances that your brand, content or information appears inside responses generated by tools like ChatGPT, Gemini, Google AI Mode or Perplexity.

Researchers from Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi helped popularise the term through the original GEO research paper. The original GEO researchers tested several ways of improving content visibility inside generative-engine responses and reported gains of up to 40%. The useful part is that the effectiveness varied by domain, which is another reason I wouldn't turn the findings into some stupid universal checklist. 

Don’t become too attached to the terminology, though. Even the industry cannot agree on what to call the discipline. Depending on who you speak to, you’ll hear GEO, AEO, AI Search Optimization, LLM Optimization or simply SEO.

I find it easier to think about the job rather than the acronym.

SEO helps people find your page. AEO makes your information easier to use when answering a question, while GEO focuses on increasing the chances that your brand, information or point of view becomes part of an AI-generated response.

There is a huge amount of overlap between them.

How does an AI answer your question?

Imagine an extremely well-read student sitting an open-book exam.

The student knows a lot from everything they studied before the exam. When the lecturer asks something they need more information about, they can also open books, search for additional sources, compare what those sources say and then write an answer in their words.

AI search works in a similar way, although the technical process behind it is obviously much more complicated.

A model may know something about your company from its training data. Depending on the product and question, it may also search the web for newer or more specific information before generating the answer.

OpenAI describes ChatGPT Search as using web sources and search providers to find relevant information when responding to queries. Google has also explained that AI Mode can use something called query fan-out, where a complicated question gets broken into several smaller searches that are investigated before the system combines the findings.

Now, ask: 

What is the best payroll platform for a 50-person startup hiring across Nigeria, Saudi Arabia and the UK?

An AI system may effectively investigate several things at once: international payroll software, Nigerian employment, Saudi compliance, UK payroll, product pricing, company reviews, available features and other relevant questions.

Think of GEO as building an evidence trail

Go back to our well-read student.

Imagine ten credible websites say that your company supports payroll in Saudi Arabia. Your own website explains exactly how it works, customers discuss their experience on review platforms, journalists have covered the company and you have published useful research about Saudi employment.

The student has plenty to work with.

Now imagine your entire website says things like:

Unlocking tomorrow’s workforce through transformative human-first innovation.

Fantastic. Nobody knows what the fuck you do.

If there is very little useful information elsewhere on the web, the AI has even less evidence to rely on.

A large part of GEO therefore comes down to making your company easy to understand and creating enough credible evidence around it that an AI system can confidently use your information when answering relevant questions.

That evidence can come from your website, but it can also come from research, news coverage, reviews, Reddit discussions, YouTube videos, comparison pages, documentation and other third-party sources.

I increasingly think about GEO as managing an evidence ecosystem around your brand.

Being mentioned, cited and recommended are different things

This is one of the most useful concepts to understand.

If somebody asks ChatGPT about savings apps in Nigeria and it says PiggyVest is one of the available options. PiggyVest has been mentioned.

If the answer links to a PiggyVest page while explaining one of its features, PiggyVest has been cited.

If the response says PiggyVest may be a good choice for someone who wants automated savings, it has been recommended.

Those three outcomes do not always happen together.

An AI system could recommend your company while using Reddit, G2 and TechCrunch as supporting evidence. It could also cite a research report from your company without recommending your product at all.

That is why measuring AI visibility purely by asking, “Does ChatGPT mention us?” leaves a lot unanswered.

I also want to know what the AI says about the company, what sources influence that description, whether competitors appear more frequently, whether the information is accurate and which brands are actually being recommended when buying questions come up.

Zach Chahalis from iPullRank summed up part of the shift well when he said:

“We’ve moved from, ‘Do we rank?’ to, ‘Are we cited?’”

I would extend that question further: are people seeing you, understanding you correctly and being given a reason to consider you?

Does this mean SEO is dead?

No.

Please don’t fire your SEO person because ChatGPT exists.

Google itself says the same SEO fundamentals used for normal Search continue to apply to its AI experiences. Pages generally still need to be crawlable, indexed and eligible for normal search results before they can appear as supporting links inside AI Overviews or AI Mode.

Google has even warned marketers to focus on effective SEO practices instead of chasing random “AEO/GEO hacks.”

The basics still do a lot of heavy lifting. Your website needs to work properly, search engines need to understand it, your content should answer real questions, your information should be accurate and your company should have some level of authority.

GEO simply stretches the job further into areas marketers may previously have treated separately, including digital PR, brand reputation, third-party mentions, original research, reviews and how clearly information about your company exists across the web.

What can marketers actually do?

The first tip is painfully simple: say what you do clearly.

If your software helps startups pay contractors across multiple countries, write that somewhere machines and humans can easily understand. You do not need to turn every product page into corporate poetry about “unlocking limitless possibilities across a borderless future.”

Clear information gives machines something useful to work with.

The second thing is creating material worth retrieving.

Original research, useful statistics, clear product information, expert commentary, comparisons, examples, definitions and firsthand experience all give an AI system more substance than another generic article repeating what the rest of the internet already said.

The original GEO researchers tested different ways of improving content visibility inside generative engines and found that approaches involving citations, quotations and statistics could improve visibility significantly in parts of their benchmark. The results varied depending on the type of query, which is important because it means there is no universal checklist you can mindlessly apply to every page.

Please do not read that study and start throwing random statistics into every paragraph.

The third thing is paying attention to what other people say about you.

Pew’s analysis of Google AI summaries found that platforms including Wikipedia, YouTube and Reddit appeared regularly among cited sources. Other research and industry experiments have repeatedly shown AI systems drawing heavily from reviews, forums, publications and other third-party websites when explaining or comparing brands.

Your website tells the internet what you claim about yourself.

The rest of the internet gives AI systems additional evidence about whether those claims make sense.

There is no secret button for ranking in ChatGPT

This is where the conversation gets less sexy.

Nobody outside these companies knows the complete recipe behind every ChatGPT recommendation, Gemini answer or Google AI Overview.

Models change. Retrieval systems change. Sources change. Prompts change. The same person can ask a question twice and receive slightly different answers.

Google has openly said that its AI experiences can use different models and techniques depending on the query. Research from tools such as Ahrefs has also shown considerable volatility in the sources appearing inside AI Overviews.

So when somebody shows you one screenshot where their company appears in ChatGPT and declares that they have cracked GEO, please relax.

One appearance is not a strategy.

The better approach is continuous experimentation. Identify the questions customers actually ask, measure how your company appears across those questions, study the sources influencing the answers, improve the information around your brand and repeat the process over time.

Measuring GEO requires more than traffic

Traffic still matters.

I would personally rather have 10,000 people visit a website and buy something than receive an imaginary trophy because ChatGPT mentioned the company 400 times.

But traffic cannot tell you everything when people can now research your category, compare your competitors and develop an opinion before clicking through to anybody’s website.

When I analyse AI visibility, I care about questions like:

• How frequently does the brand appear across relevant prompts?

• How often does the AI recommend it?

• Which sources are being cited?

• How visible are competitors?

• Is the description of the company accurate?

• Does the brand appear for high-intent questions or only informational ones?

• Are visitors coming from AI tools eventually converting?

Tools and methodologies for measuring some of these things are still developing, so I would be suspicious of anyone pretending the industry has already settled on a perfect measurement framework.

The goal remains the same as most good marketing: understand whether the work is increasing the chances that the right people discover, trust and eventually choose you.

The easiest way I’ve found to understand the whole thing

When somebody uses traditional search, your website is competing to become one of the places they visit while looking for an answer.

AEO makes your information easier for search and answer systems to extract when responding to a quetion, while GEO expands the focus to how your brand, content and reputation appear inside AI-generated responses.

You have to think beyond a single webpage.

The machine may be learning about you from your website, somebody’s Reddit comment, a review platform, a journalist, your documentation, your original research or a combination of all of them.

Your job is to make sure the information surrounding your brand is clear, useful, credible and available in the places these systems look.

There is still a lot we do not know about how AI search will evolve, and anyone claiming otherwise is probably selling something.

I’m continuing to test how brands appear across ChatGPT, Gemini, Google AI and Perplexity, and I’m co-authoring the State of AI Search in Nigeria to understand what this shift looks like in our own market. You can sign up here to get the report when it drops, and I’ll keep sharing what I learn as this entire industry inevitably changes again next week.