Get In Touch
The new reality – Position Zero – answers directly in the search interface or zero-click search. Modern SEO is no longer limited to keyword density and link mass. What is much more important is how artificial intelligence ‘reads’ your business – the content of your website and everything it finds on the web. And how it then interprets and transmits this data to the user.
Just yesterday, we were fighting for the top positions in search results, according to experts at Stelvel Ltd. Today, we are watching as the top ten results fade away at the bottom of the screen, negating the efforts of marketers. This is what Google’s expanded response and the AI assistant Overview look like. Traffic has been divided. Now, in addition to search robots, SEO must also take neural networks into account. The new optimisation strategy is called Generative Engine Optimisation (GEO).
How to adapt your website content for neural networks
Stelvel Ltd offers a simple and effective GEO algorithm: from collecting questions to the final page, which is easy for neural networks to cite. What do you need to consider and do?
- New type of semantics. Instead of high-frequency keywords, collect 50 relevant, frequently asked questions in your niche (using Keyword Planner, for example).
- Search results audit. Send this list of questions to ChatGPT, Gemini, and Perplexity, and request links to the resources where the information was taken from.
- Reverse engineering. Let the neural network analyse the sources of information, find regularities and structural patterns: lists, tables, micro-markup and the logic of presenting facts.
- Content adaptation. Create pages optimised for neural networks:
• Summary block: brief summary on the first screen (for snippet generation).
• Structure: clear headings, comparative tables, checklists.
• Glossary: block with definitions of terms at the end of the material. - Fast indexing. Post links to new materials in relevant Reddit threads, which can serve as a signal for neural networks.
- Consistency. Update pages monthly so that the algorithm reads the data as current.
Reverse engineering algorithms
Attempts to guess what an algorithm will like usually end up draining your budget. It makes more sense not to collect a semantic core blindly, but to find out right away which answers AI models consider high-quality. Simply upload 50 popular questions from your niche to ChatGPT, Gemini, or Claude, adding a prompt requesting links to primary sources.
Stelvel Bulgaria recommends selecting questions of varying levels of complexity, from simple ‘what is…’ to practical ‘how to…’ questions. This will provide a balance between introductory information and useful instructions.
Content for robots that people read
The more predictable the structure, the higher the chance of getting into Position Zero. Pages that appear in search results are almost always built according to a strict template that is easy for machines to read.
They have a clear hierarchy and elements that are easy to parse: from anchor headings to comparison tables and checklists – formats that AI reads instantly. The main trigger is the summary at the beginning of the article: two or three sentences summarising the essence, which the algorithm can use as a ready-made answer. And the glossary at the bottom of the page serves as a signal of expertise for information search, increasing the chances of getting into the search results for information queries.
This is reminiscent of ‘old SEO,’ but with one difference: the text is written not for keywords, but for meanings that the algorithm can verify. Use Schema Markup: structured data helps Google better understand your content.
Backdoor via Reddit
The main problem with the new search is indexing speed. For the model to include data in its response, it needs a strong external signal.
A link in a thematic thread on Reddit can act as a catalyst. Neural networks use this platform as a priority source: lively discussion guarantees instant indexing of the original source. And monthly page updates serve as a kind of signal to algorithms that the data is relevant and the source can be trusted.
How does AI traffic ‘sell’?
When a person visits your website from a regular search, they are only gathering information and looking for a solution. A user who clicks on a neural network recommendation has already received an answer and may have already made a decision. Stelvel company’s practice shows that GEO (neural network optimisation) has an unexpected effect – AI traffic consistently brings in more orders than classic search traffic. The neural network acts as a filter, and Position Zero immediately provides a basic answer, cutting out random transitions and ‘cold’ clicks.