The GEO era: ChatGPT and the new search paradigm

The end of traditional search: Welcome to GEO

The search ecosystem is undergoing its most radical transformation since the consolidation of Google's PageRank algorithm in the late 1990s. For decades, SEO (Search Engine Optimization) has revolved around link architecture, keywords, and domain authority. However, the emergence of large language models (LLMs) with web browsing capabilities has given way to Generative Engine Optimization (GEO). This new paradigm does not seek to rank a link in a list, but to influence the informative synthesis of an autonomous agent. As analyst Simon Willison points out, GEO is no longer an academic theory, but an operational reality where tools like Promptwatch monitor how companies are cited—or ignored—in the responses of ChatGPT, Claude, or Gemini.

The leap in 'site:' operator usage: A revealing technical signal

Data analysis from Promptwatch following the deployment of the GPT-5.6 'Sol' model in August of this year has exposed a deliberate strategy by OpenAI. Historically, the use of the site: operator in ChatGPT's internal queries remained in a marginal range of between 0.3% and 0.5%. Following the staggered rollout between August 3rd and 5th, this figure experienced exponential growth, reaching 17% on August 8th.

This statistically significant jump coincides chronologically with OpenAI's official update, which promised greater "factual reliability" and "more focused" responses. The technical hypothesis, supported by experts like Willison, is that OpenAI has integrated an internal domain filter, possibly with a logical structure like search(query, recency, domains). This implies that, instead of performing an open search across the entire web, the model's inference engine now pre-filters sources it considers authoritative, prioritizing specific domains based on the query. This move is reminiscent of the early days of web directories, where human curation (or in this case, algorithmic curation) prevailed over massive indexing.

Why does this change the rules of the game for companies?

Traditional SEO was based on the premise that a blue link on a search engine results page (SERP) would lead the user to a website. In the GEO environment, the user often receives the definitive answer within the chat interface. This poses an existential challenge for referral traffic.

• From acquisition to citation: The value of a brand is no longer measured solely by clicks, but by the frequency with which the model uses it as a primary source of truth. If a website is not indexed by the agent as a "reliable" source, it becomes invisible, regardless of its ranking on Google.
• The erosion of forums: A critical finding from Promptwatch is the drastic reduction in citations coming from Reddit. This suggests that OpenAI is applying a quality filter that penalizes user-generated content in favor of institutional, technical, or journalistic sources. Companies that relied on organic visibility in forums are losing a historical source of traffic.
• System opacity: Unlike Google, which has historically offered tools like Search Console to diagnose indexing issues, OpenAI keeps its system prompts and source selection algorithms under absolute secrecy. This turns GEO into a "black box" discipline where professionals must constantly perform reverse engineering to understand why one brand appears in the chat and another does not.

Impact and historical comparisons

We are witnessing a phenomenon comparable to the transition from directory-based navigation (like Yahoo!) to algorithmic search engines. Back then, companies had to learn to optimize their metadata; today, they must optimize their brand semantics to be "readable" and "citable" by AI. The impact on marketing budgets is direct: investment is shifting from link building toward the creation of structured, technical content that facilitates the work of LLMs. The "death" of referral traffic is not an exaggeration, but a reconfiguration of the attention economy where the website becomes a knowledge base for AI, rather than a final destination for the user.

Speculation and future: The challenge of veracity

It is essential to maintain skepticism: although Promptwatch data is solid, there is no official confirmation from OpenAI regarding a change in its search algorithm. However, the correlation between the GPT-5.6 update and the change in search behavior is too consistent to be random. It is speculated that this trend toward the use of domain operators will continue, which could lead to a more fragmented web, where only domains that manage to pass the AI's "quality filter" will have relevance. For companies, the future of digital work will consist of managing brand identity in an ecosystem where the intermediary is not a search engine, but a knowledge agent that decides, in milliseconds, what information is worthy of being presented to the end user.