How AI is reforming the internet

Google once reformed the internet with its web search and built an empire on it. In 2024, its parent company Alphabet generated nearly $200 billion in revenue from internet search and advertising. Now the next reformation is upon us, which is likely to spell the end of the web search era. But what will replace it?

When Google launched its web search in 1998, it was impossible to predict what a huge impact this service would have on the usability of the internet. Before that, searching for information was much more difficult because search engines had not yet developed suitable means of organising the growing number of websites and content. They were more like cumbersome digital catalogues.

That changed abruptly with Google's PageRank algorithm. Suddenly, search results appeared sorted according to specific relevance criteria. Google finally achieved its commercial breakthrough two years later with the introduction of AdWords (now Google Ads). Now advertisers could rent a place at the top of the search results.

This business model remains one of the most successful in the digital economy to this day. And yet, Alphabet began a transformation years ago that will largely render this successful model obsolete. 

From search engine to answer engine

Google basically began its journey to becoming an answer engine back in 2012 with the Knowledge Graph. This semantic function answers simple questions directly with a short answer on the search results page, such as current sports results or stock market prices. Many more steps followed, such as Google Assistant, BERT, LaMDA and Bard, which later became Gemini.

These changes to web search alone led to a massive increase in zero-click searches, i.e. search queries where the searcher does not click on any of the links displayed.

Since 2024, Google has been systematically integrating AI-generated summaries into search results. In Germany, the ‘AI overview’ has been displayed since March 2025. Although the AI overviews contain links to the sources used, the answers are very comprehensive, so they are sufficient for many searchers. No more long clicks, just quick and easy-to-consume knowledge.

Once again, it is the radical simplification of a previously complex action that is driving Google forward. And once again, it will not just be a new feature, but will massively change the way the internet itself works.

The impact on the internet

Parallel to Google's rise, the internet developed from a niche for ‘digital nerds’ into an important business world for companies and organisations. It had long since evolved into an independent digital world that hardly anyone could ignore. Companies that were not active here did not exist.

Websites became digital presences and their content became essential infrastructure for brand visibility. All of this was based on the framework provided by Google's algorithms. Only content that followed the principles of search engine optimisation was displayed by Google in relevant search queries and directed searchers to the websites of the companies.

Over the years, this system became increasingly sophisticated. However, as competition grew, it became more difficult to secure one of the coveted spots at the top of the search results for relevant keywords. No wonder that search engine optimisation (SEO) also developed into a real business.

But that will soon be a thing of the past. Google will only respond to search queries with link lists as a side note. Instead, the desired information will be generated directly as an answer, as users of ChatGPT & Co. are already accustomed to. Since users will only click on the links contained in the answers in exceptional cases, operators of digital websites are faced with the question: How will potential customers find their way to their websites in the future? 

GEO instead of SEO

The AI-generated answers require information that may potentially come from the same sources that search engines previously used to compile the link lists. The crawler bots that collect information from the web are also not fundamentally different. Nevertheless, there are some significant differences between optimising content for search engines and for language models (LLM).

For large language models (LLMs), established SEO factors such as keywords, backlinks and domain trust are significantly less relevant than for traditional search engines. In large language model optimisation (LLMO), other criteria will play an important role. Here is a brief overview:

• High-quality content: Natural, structured content of high quality.
• Context: User-centric content that answers specific questions in a specific context.
• Conversational content: Content such as instructions, explanations and FAQ pages that answer specific questions in natural language, as LLMs do in their responses.

Conclusion:

The criteria for optimised visibility in language models read like a modern content marketing strategy: for years, user centricity, high-quality content and solutions to specific problems have been considered criteria for success in this area. So does this mean that nothing will change and we are only seeing a technological evolution rather than a revolution?

It's not quite that simple. While it will be possible for companies and organisations to build visibility in language models, the value of this presence has yet to be proven. Searchers will most likely quickly get used to the fact that the comprehensive answers provided by language models already fulfil their intentions. In many cases, it is doubtful whether they will then visit external websites that are mentioned or linked in the answers. Who devotes themselves with great interest to the references in a Wikipedia article?

It is not yet possible to reliably assess how quickly the reform of the internet will progress and what concrete effects this will ultimately have. What is clear, however, is that content-based corporate communications will have to strategically reposition itself.

Text: Falk Hedemann