LLMO , GAIO, GEO, SGE: The Future of Search in the Age of AI

I've looked up synonyms, had words and entire paragraphs translated, sought inspiration for this very article, asked for a raspberry sponge roll recipe... and probably forgot many other things my little helper assisted me with last week. So, roughly speaking, it's 20:0 for ChatGPT vs. Google. A heavy blow for the search engine giant, as well as for other platforms, because I've consequently seen about 60 fewer ads and clicked on 10 fewer websites. Is nothing less than the future of the internet as we know it at stake here? And how can brands remain relevant in this uncertain future, i.e., get their content into AI systems? Let me try to shed some light on this.

Are you still searching, or are you already generating?

For years, search engine optimization (SEO) has dominated companies' strategies to increase their visibility on the web - and thus traffic to their own websites. While it used to be possible to push websites to the top ranks of search results with tricks like keyword stuffing and automated link building, search engine algorithms have become increasingly sophisticated. For a while, it seemed that valuable content - those with positive user signals, clear added value for the recipient, and high-quality backlinks - had triumphed over the evil and irrelevant.

But then AI came along. It's flooding the net with millions and billions of content: texts, images, videos, comments, music pieces... Without labeling, control, without regard for consequences - and SEO-optimized by exploiting certain algorithms. Even though Google has been counteracting this since March with its "Helpful Content Update," the quality of search results remains mixed. AI-generated content is sometimes only recognizable at a second or third glance, and quality and plausibility must first be questioned and checked by the recipient. So if search results seem to consist only of advertising and AI junk, why not just ask ChatGPT or Perplexity directly? Well, here's the catch: Large Language Models and Generative Engines, in turn, absorb all this generated content and spin it together for the next queries, producing results of even worse quality. A downward spiral.

So, is the internet completely ruined?

Some people would say it's been dead for a long time. And they're certainly not entirely wrong. Now it's up to the big tech companies and AI firms to clean up, adjust their algorithms, and label fakes. Will it work? We'll see. Currently, marketing experts and SEOs are already dealing with what a world without Google traffic could look like - in other words, how to feed AI systems so that they display one's own brand content.

So how do we influence AI?

When you delve into this topic, you keep stumbling across new terms that represent this change. Let's start with a differentiation:

Large Language Model Optimization (LLMO): Strategies and techniques aimed at designing and optimizing content so that it can be better understood and processed by large language models (like GPT-4). Google also uses Large Language Models, so optimization can also be relevant for "classic" SEO.

Generative AI Optimization (GAIO) or AI Optimization (AIO): The optimization of content and data for use by generative AI systems, with the goal of obtaining high-quality and relevant generated content. In addition to the language models mentioned above, this can also include image, audio, and video generators.

Generative Engine Optimization (GEO): An evolution of search engine optimization (SEO) that focuses on optimizing content for generative models to ensure that this content is preferred and effectively used by AI systems.

Search Generative Experience (SGE): An advanced search technology that uses generative AI models to provide direct, context-related answers to search queries, instead of just displaying a list of links. Google, for example, rolled out initial SGE versions in the US last year, and recently gave insights into what these models mean for advertising during their Global Live Marketing Live Keynote. A rollout in Europe has not yet been scheduled.

Let's simplify a bit at this point and focus mainly on GEO. The core question: How does this approach differ from traditional SEO?telle ein bisschen vereinfachen, und uns hauptsächlich auf GEO konzentrieren. Die Kernfrage: Wie unterscheidet sich nun das Vorgehen von der traditionellen SEO?

SEO focuses on optimizing content for search engine algorithms, which traditionally are based on keywords and backlinks, among other things. GEO requires an even deeper engagement with understanding and generating content. GEO emphasizes user intent and context, meaning content must not only be comprehensive and well-structured but also relevant.

However, Google also tries to recognize context in search queries and varies the SERPs accordingly depending on the search query and intention. Structured content also has a positive effect on SEO (headline systems, classification according to Schema.org, etc.), and relevance is also considered through helpful content with added value and user signals. We assume that only the weighting of factors between SEO and GEO will differ. With GEO, some features that Google uses will certainly be left out, as they cannot be captured or can only be insufficiently captured. User signals, for example: Google can access click numbers in combination with dwell time and other influencing factors. An LLM like ChatGPT would have to rely on the built-in feedback function for similar findings. But let's be honest: who rates every answer?

What makes content relevant for AI?

If we knew that exactly, it would be far too easy. Experts assume that "press mentions," i.e., the mention of one's own brand, person, or company in established media, plays a major role. But the opinion-forming power of the masses is also taken into account: Wikipedia, Quora, IMDB, Yelp, and also the Reddit platform are used as training data.

But for SEO, these mentions and the associated links are also of increased relevance, which Reddit is currently feeling massively through so-called "parasite SEO."

It's also conceivable to influence the learning systems by having a critical mass of people repeatedly ask similar questions and then correct the answer so that the brand is mentioned positively. Negative influence is also possible here. There will certainly be many tests and insights gained in the coming months. Between us: Let's hope that  it’s not so easy to manipulate.

How can I optimize my website for AI? 10 tips

As you can see, it involves some effort to serve these "external relevance criteria." It doesn't usually happen that a reputable daily newspaper writes (unpaid) about the 10 best sneakers or power drills. So where should the press mentions come from? And establishing a presence on Reddit is at least as time-consuming as building up any other social media channel and involves a large investment in time and resources.

There are a number of ways you can implement some of this easily on your own website. Here are 10 tips:

1. It sounds banal, but to make your own content available to AI systems, you must first make sure that your robots.txt allows LLM User Agents.
2. Write your content so that AI systems can use it directly. For example, provide answers to questions and structure your content accordingly. This also helps optimize for voice search.
3. Use simple and easily understandable language. Explain technical terms.
4. Generally follow the EEAT criteria issued by Google for relevant content and keep your website always up to date.
5. Enrich your articles with expert opinions, quotes, and (real) user reviews. Additionally, use user-generated content and don't forget to make this indexable as well. It's important to label this as such with tags to avoid being penalized by search engines.
6. Choose user-friendly editorial features such as pros and cons lists or executive summaries to make it easier for AIs to cite you.
7. Write individual articles with their own URLs on subtopics and don't try to cover all aspects of a topic in one article. This point is somewhat in competition with classic SEO: Google now promotes content with different perspectives and a more holistic view. So it's about finding the right mix!
8. Pay attention to loading times.
9. Use structured data (schema.org)
10. Build your website formally correct, pay attention to headline structures, etc.

You can get even more tips from Malte Landwehr, Head of SEO at Idealo. According to some reports, he spent over 50 hours researching studies, blog articles, and papers to compile a list of the most important criteria. Really worth reading!

The best comes last: You might already be doing GEO without knowing it

Did you notice something? Most of the tips and relevance criteria have been valid for SEO for years. So old wine in new bottles: Good, human-made (!) and above all well-prepared content is still the be-all and end-all.

If you want to make your brand fit for what's coming or generally become more proficient in the topic of artificial intelligence, then take a look at our workshop offer or contact us directly.

PS: Why don't you ask ChatGPT which are the best healthcare agencies in Germany? If cyperfection isn't on the list, it's high time to correct the machine there ;)