GEO strategy for AI search
Your customers no longer ask Google alone. They ask ChatGPT, Perplexity and other AI search tools. Instead of ten blue links they get a single, constantly shifting answer. If your brand is not in that answer, your competitor is.
Visibility in AI search is not a matter of luck. It is measurable and it can be optimised. That is GEO: generative engine optimisation. We measure where your brand stands today, show where you are losing reach, and make your content citable. With clear KPIs instead of gut feeling.
Like classic search engine optimisation, GEO optimises your own digital content. As a discipline it grew out of SEO and forms its newest branch.
Generative search engines are all search surfaces that can access a web index and whose answers are assembled by a generative AI:
- Google (AI Overviews / AI Mode)
- Bing (Copilot search widget)
- ChatGPT with web search enabled
- Perplexity
Simple chatbots lack the real-time connection to the live internet. They are not full search engines but pure answer machines. Disable web search in ChatGPT and this grounding disappears, at which point the tool no longer counts among AI search tools.
How do you optimise for AI answers?
Generative engine optimisation brings together two clearly distinguishable disciplines:
LLMO, large language model optimisation: optimising content for the training and brand understanding of language models.
RAG optimisation, retrieval-augmented generation optimisation: optimising for the retrieval systems of AI search.
An AI search tool either knows your brand directly or retrieves its knowledge about you from the internet. A complete GEO strategy works on both: the long-term brand understanding inside the models and the optimisation for short-term live retrieval.
LLMO strategy
The goal of an LLMO strategy is to anchor your brand, products and expertise deep in the data structures of language models, so that they recognise your brand as an authority in its field. That raises the chance of appearing prominently in relevant AI answers. LLMO is digital PR for language models.
Language models are a black box. Nobody outside the providers has direct access to the training data. The GEO community can only form hypotheses through experiments, much as SEO worked for a long time, before the Google leaks of May 2024.
What we do know: the semantic digital mention environment of a brand is the core of any LLM optimisation, because large parts of it feed into the training data. That environment can be represented as mathematical vectors. It is how AI search engines tell a medical product text from a homeopathic one, or how closely two brands sit together, purely from the proximity of numbers in vector space.
The biggest lever is therefore entity management: your brand should appear as a clearly identifiable entity across the entire web. This matters most in a house of brands, where the relationships between manufacturer, product family and product are tangled. Across systems such as your own website, Wikidata, Wikipedia and local listings, the representation often varies enormously. "Grown historically" is the usual explanation; cleaning that data up is the usual answer.
Further off-page factors support the LLMO strategy: brand mentions in trade media, blogs, guest articles, forums, reviews and social media. Even classic backlinks help your brand be perceived as an authority.
| AI search tool | Language model |
|---|---|
| Google AI Overviews / AI Mode | Gemini |
| Bing Copilot search widget | OpenAI GPT (customised) |
| ChatGPT | OpenAI GPT |
| Perplexity | model-agnostic (OpenAI GPT, Claude, Sonar) |
You should know which AI search tools your audience actually uses and concentrate on the surfaces with the greatest effect. Usually that is still Google search. In specific niches, particularly in B2B, the picture can look different.
RAG optimisation strategy
RAG, retrieval-augmented generation, connects AI search engines directly to external knowledge bases. Instead of relying solely on static training data, the system draws on additional sources. That raises the relevance and quality of answers considerably.
Optimising for RAG means everything that lets bots and crawlers read web content more easily. The RAG optimisation strategy therefore builds on classic technical SEO and on-page work. Anyone who has invested in those foundations is already well placed:
- crawling and indexing strategy
- no content hidden behind client-side JavaScript rendering
- structured data (schema markup) in the source code
- structured content: heading hierarchy, tables instead of walls of text, readable copy
- evidence in the text: statistics, quotes, sources
- maintained currency with a visible date
Different AI search tools sit on different web indexes. As with LLMO, the detail is worth checking: if your content never reached the relevant index, it cannot be cited in the corresponding tool.
| AI search tool | Web index |
|---|---|
| Google AI Overviews | Google web index |
| Bing Copilot search widget | Bing web index |
| ChatGPT | Labrador (own index), Bing web index (API partnership), Google web index (via third-party SERP scraping) |
| Perplexity | own web index, Bing web index (via API), further sources such as the Brave Search API |
GEO KPIs: tracking and monitoring
To measure the effect of GEO work, the most important method in our view is collecting synthetic data. We built our own tool for it: the netzstrategen AI Search Index.
The first step is to develop a golden prompt set with you, based on real user data. We query those prompts at regular intervals across the common AI search tools. The simulated data shows where your website and your brand have weak spots.
- AI search citations
- How often your content is linked as a direct source in AI answers.
- AI search brand mentions
- How often your brand is named in the generated text.
- AI search share of voice
- Your share of mentions compared with the whole competitive field inside AI search.
- AI search brand sentiment
- The tone and perception of your brand in the answers language models give.
- AI search brand position
- Where your brand is placed within an AI-generated list of answers.
We complement this by analysing server log files and impressions from Google Search Console and Bing Webmaster Tools, comparing the synthetic data against real AI crawler visits. That shows which crawlers reach which pages and which pages are cited most. Especially when dealing with artificial intelligence, every claim belongs to a valid data source.
How a GEO project runs with us
Measure the status quo: GEO audit
We record how often and in what context your brand appears in AI answers. And who is named instead. That covers your website, market and competition, plus a baseline measurement of your GEO KPIs. Technical hurdles are part of it: is the site accessible to AI crawlers, and is the content indexable?
Develop the strategy
From the audit we derive your priorities as a threefold plan: quick wins on the citability of existing pages, mid-term coverage of the golden prompts, and long-term work on entities and reputation.
Implement and adjust
The measures fit into your existing content strategy. What stands at the end is authority in your core topics rather than a one-off campaign: evidenced, repeated and verifiable.
Our GEO products
The sequence is a clear ladder: measure first, then plan, then keep at it.
GEO workshop
Building on the audit, we develop your golden prompt set so GEO KPIs become measurable. It usually also reveals where additional content is worth creating.
GEO audit
The audit reviews your information architecture and whether AI search engines can read it, with a focus on RAG optimisation. The result is a concrete action plan.
GEO tracking & monitoring
We measure your visibility over time with the AI Search Index, so you can see whether the measures work and how you develop against direct competitors.
Frequently asked questions about GEO
- What is a GEO strategy?
- A GEO strategy makes sure a brand is named and correctly represented in the answers of AI systems such as ChatGPT, Perplexity or Google AI Overviews. It combines two disciplines: LLMO for the knowledge inside the model, RAG optimisation for the retrieval system. It complements the SEO strategy.
- What is the difference between GEO and LLMO?
- GEO is the umbrella term, LLMO one of the two pillars beneath it. LLMO optimises what a model knows about a brand from training. RAG optimisation, the second pillar, optimises what a system looks up at runtime.
- Does GEO replace classic search engine optimisation?
- No, GEO complements SEO. With RAG systems good SEO is even a precondition, because many AI systems fall back on the organic search index.
- How long does GEO take to work?
- That depends on the pillar. RAG optimisation can show an effect within days or weeks. LLMO can take considerably longer, because a new model generation may be needed before a language model has learned a piece of information.
- How do you measure visibility in AI answers?
- Through golden prompt sets rather than keyword sets. Citations and mentions side by side already give a good picture, as long as the prompt set was chosen for real search demand and relevance to the brand.
- What is a grounding page?
- A grounding page is a central, structured information page on your website that serves AI search engines as a reliable source. It bundles the essential brand and product information and so improves the citability and accuracy of AI answers.
- Do we need an llms.txt?
- The llms.txt is not yet an established standard, though querying it is a growing trend among SEO tool vendors such as Semrush. In the audit we check whether it makes sense for your site, or whether a grounding page is the better investment.
GEO does not work on its own
SEO improves visibility in classic search results; GEO extends it to generative answers. Technical accessibility, indexing, information architecture, strong content, entities and authority are the shared foundation. That is why we develop both as parts of one search strategy rather than separately.
Your contact
SEO Strategist
As an SEO strategist, Sven Eric Maier bridges the gap between user-centric content and technical performance. His expertise is rooted in years of e-commerce experience and building his own platforms across the web.
Where does your brand stand in AI search today?
In the GEO workshop we develop your golden prompt set and measure the starting point.