SEO for Artificial Intelligence: Top Websites for LLMs and How the Influx of Marketers is Impacting Quality
Semrush analysts have identified the top ten most popular internet resources that large language models (LLMs) reference when generating answers. However, the algorithms’ high level of trust in these sites has spawned a new problem: platforms are being massively flooded with artificial content, which directly degrades the quality of the neural networks’ output.
According to recent data, artificial intelligence’s choice of information sources is neither random nor arbitrary; it is strictly segmented. Three major platforms have become the undisputed leaders in LLM citations. Wikipedia is traditionally used by algorithms as a foundational platform for extracting dry facts and encyclopedic references. Reddit has proven indispensable for analyzing live human discussions and informal opinions, while LinkedIn serves as the main provider of expert content that models turn to when generating answers in the B2B segment.
In addition to the leading trio, the top ten most frequently mentioned domains also include the video hosting site YouTube, the blogging platform Medium, the publication Forbes, as well as the official resources of Google and Microsoft.
The Dark Side of Popularity
Experts note that the predictability of language models has triggered the emergence of a new marketing trend – GEO (Generative Engine Optimization).
Understanding exactly which platforms AI relies on first and foremost, SEO specialists and brands have begun purposefully publishing hundreds of texts generated by artificial intelligence itself on these resources. The main goal of such actions is to force the language model to “swallow” the desired promotional material and start referencing the promoted product in its answers to real users.
Despite the high commercial effectiveness of this method, experts are sounding the alarm. A vicious cycle is emerging: new generations of neural networks are increasingly relying on marketing “spam” that was previously generated by algorithms. Because they are being trained on these artificial texts, the overall depth, objectivity, and quality of LLM responses are beginning to decline rapidly.
Other news from the world of artificial intelligence
