}
Emina Demiri-Watson
Massive thanks to the amazing Emina Demiri who is back this week to demystify GEO, AEO, AI Search Optimisation, and whether there’s any difference at all to good quality SEO.
A client recently emailed me a pitch deck by a competitor agency that had a whole section on Generative Engine Optimisation. Most of it was “pork pies”, but some of it was solid advice.
The problem was the solid advice was not really any different from what most great SEOs have been advising for ages already. It just had a fancy new name pinned to it.
I might end up eating my hat in a few months but at the moment, optimising for LLMs is basically no different than optimising for search engines. Here’s why…
Contents:
- What even is GEO, AEO & AI Search Optimisation?
- The emperor has no clothes
- Semantic search & user journey mapping have been around for a while
- GEO is SEO if done right—for now
What even is GEO, AEO and AI Search Optimisation?
Let’s first define what we mean by GEO, AEO or AI Search Optimisation—and, in fact SEO, in the first place.
According to the Google SEO Starter Guide, SEO is a process of helping search engines understand your content, and users to find your site and decide whether they want to visit the content.
The latter is interesting, since it points to SEOs as having the responsibility of not just getting your product into the search results, but understanding what will make those users visit the website in the first place. Remember this point!
Now let’s look at a different definition: Wikipedia.
Search engine optimisation (SEO) is the process of improving the quality and quantity of website traffic to a website or a web page from search engines.
Here, the emphasis is actually on the goal: traffic. But what happens when the traffic starts to dwindle?
There is no doubt that AI Overviews have decreased click-through rate (CTR) for traditional organic listings.
According to Ahrefs, websites are experiencing a whopping 34.5% drop in position 1 CTR when AI Overviews were present, based on an analysis of 300,000 keywords.
Amsive also reports an average 15.49% CTR drop (with much larger losses when combined with featured snippets), based on an analysis of 700,000 keywords.
If traffic is your KPI, it’s no wonder the industry is reaching for new terms to prove we still bring value!
Enter Generative Engine Optimisation (GEO), Answer Engine Optimisation (AEO) and AI Search Optimisation—all three are heralded as the new, fresh, rebrand of (or a complementary strategy to) search engine optimisation.
Since I’m short for time and an SEO, I had a quick look at which one of the three is currently leading in terms of interest. It’s GEO.
There is no difference in GEO, AEO or AI Search Optimisation, so let's stay on trend and use GEO only. One mystery solved: GEO, AEO or AI Search Optimisation are basically the same.
According to SEL, GEO is the process of improving your website to boost visibility in AI-driven search engines.
It’s obvious that both GEO and SEO aim to increase the visibility of content for users searching online.
Personally, I don’t have a problem with the use of GEO or any other term that’s rising in popularity. In fact, I see a lot of sense in owning this space before it starts being saturated and it’s an opportunity to improve bad SEO. The problem is not the use of the new term. It’s the claim that the tactics of GEO vs SEO are different, when they are not.
The emperor has no clothes
Let’s call it: The emperor has no clothes.
GEO isn’t a revolution. It’s a rebrand. And here’s what’s really happening…
How LLMs work: schema galore or…?
Current LLM responses work on the basis of probabilistic patterns. They output what seems likely to be correct, not what is verifiably correct. This is true whether you’re talking about OpenAI’s ChatGPT, Perplexity, or Google’s new AI Overviews (AIOs).
You’ve likely heard the term hallucinations—when LLMs generate plausible but nonfactual information.
Most LLMs, including Google’s AI Overviews, don’t produce factual responses by default. They generate what’s likely to be true, which is why hallucinations happen.
To fix this, AIOs are grounded in search data. They pull from ranked pages—pages SEOs already work hard to optimise.
How this grounding happens in detail is still unknown, at least to me. For example, AIO might mimic the behaviour of a RAG (Retrieval-Augmented Generation) architecture, but I doubt they use it strictly.
Thanks to recent insights shared at Google Search Central Live NYC, we have some confirmation on how it all works. Google described the system as using Predictive Summaries – generated responses that are produced ahead of time for likely queries – and Grounding Links, which are chosen from ranked content to support those summaries.
These aren't hallucinations; they're algorithmically generated summaries of search-vetted pages.
The links shown in AIOs are not random—they’re grounded in Google’s core ranking systems, just surfaced in a new generative format. So, while the interface may feel new, AIOs are still anchored in the same search ecosystem SEOs have been optimising for all along.
|
RAG is a specific architecture: the model retrieves relevant documents at the time of query, then generates a response grounded in those documents only! Kind of a: “Here are 5 sources. Read only these. Now answer the question.” AIO on the other hand is more like: “Here are some relevant search results. Now generate an answer that seems correct, based on these and what you already know.” |
This is why claims such as ‘adding schema markup is important’ are both pork pies and true. Yes, it is important to add schema markup, but this is only because LLM ground their response and schema is important for search.
When Fabrice Canel, Principal Product Manager at Microsoft Bing, said at SMX (Search Marketing Expo) in Munich, “Schema Markup helps Microsoft’s LLMs understand content,” what he likely was referring to was grounded search.
LLMs process everything as tokenised text, and structured data is turned into “statistical soup” during the training process—at the moment, this doesn't mean forever. If you want to dig deeper into the possible future of AI, dive into the area of RAGraph.
‘Til then, we are stuck with tokenisations, and schema remains important as a tactic in SEO and not specific to GEO.
Brand is NOT a new thing
At the start of this piece, I asked you to remember something about the definition of SEO in the Google SEO Starter Guide.
The Guide describes SEO as helping search engines understand your content, helping users find your site, and – critically – helping users decide whether they want to visit it.
That final step is about more than just relevance. It’s about trust, resonance, and recognition—in other words, brand.
Ranking in search has never been just about clicks. It’s about visibility, credibility, and the repeated exposure that builds mindshare of a well-defined brand.
Whether it’s a blue link, a featured snippet, or an AIO summary, every appearance in the SERP reinforces brand perception. Brand is not something we invented. It comes before SEO. Brand power helps SEO, not the other way around.
As leaked data, DOJ trial documents, and exploits have shown, Google uses brand-like signals in its quality score:
- Brand visibility (e.g., branded searches).
- User interactions (e.g., clicks).
- Anchor text relevance around the web.
And, no one will be looking for your company if your brand is wishy washy!
Great SEOs have understood this for decades. GEO doesn't change this, it just offers another surface to do the same job.
Semantic search and user journey mapping have been around for a while
Semantic search has been around for a while—this isn’t new either. What has shifted is the interface.
Search engines have long moved beyond basic keyword matching, using natural language processing and entity understanding to interpret meaning, context, and intent.
Google's Hummingbird update in 2013, and BERT in 2019, were both major steps toward semantic search—helping the engine understand what a user meant, not just what they typed. Even earlier, in 2012, things changed with the now famous ‘Things not strings’ and the introduction of the Knowledge Graph.
So when people talk about GEO as a whole new paradigm because it “understands context,” it’s worth remembering: we’ve been here before. What’s changing now is how that semantic understanding is delivered—not that it exists.
The same goes for user journey simulation. Even the upcoming AI Mode – and importantly, the query fan-out technique where Google expands your query into multiple sub-queries to deepen coverage – isn’t a dramatic departure from past behaviour. It’s just a new way of surfacing what search engines have been modelling for years through clickstream data.
In traditional SEO, we’ve long talked about intent mapping, topic clusters, and creating content that supports every stage of the user journey. That process relies on anticipating how people move through search-based journeys:
- What do they search first
- What they look for next
- What they click on
- What satisfies them (or doesn’t)
Clickstream data captures those journeys at scale. It reveals:
- How users reformulate queries
- How they move across related content types
- What keeps them engaged (or doesn’t)
- And how their goals evolve across a session
So when Google uses query fan-out in AI Mode, it’s essentially using behavioural intelligence – built from clickstream patterns – to simulate a multi-step search journey in a single generative step. Instead of waiting for a user to refine their query, Google anticipates and answers the next three questions upfront.
If you've already been creating content that maps to search journeys and intent shifts, you're not behind.
GEO is SEO if done right—for now
Personally, I don’t want to minimise what is happening in search at the moment. The chickens have finally come to roost for many SEOs out there and if using a new term will help them do their job better, then that’s great.
But ultimately, if you have been doing SEO correctly, GEO for now brings nothing new.
I agree with Michael King when he claims that the emergence of AI-driven search interfaces like AIOs signifies a transformative phase in search marketing.
But integrating semantic understanding and brand perception has been part of great SEO for a while. Yes, conversational search models prioritise semantic relationships over keyword matching. But so has Google within the traditional SERPs, for a while now.
Ultimately, for now, my strategies have not changed significantly. That doesn't mean I have not been diving deep into LLM rabbit holes, monitoring closely and speaking to clients about the future of SEO. It just means I don't feel the need to rebrand what I am already doing. So far at least.
You might also like:
Sitebulb is a proud partner of Women in Tech SEO! This author is part of the WTS community. Discover all our Women in Tech SEO articles.
Emina is the Head of Digital Marketing at Vixen Digital, a Brighton UK-based digital marketing agency. She has over 10 years of experience in SEO and digital marketing. Her special connection to publishers also comes from her BA in Journalism. Emina’s marketing passions include technical/on-page SEO, analytics, channel alignment and automation.
