ChatGPT Search Underlying Mechanism Changes, Begins Large-Scale Targeted Scraping of Specific Websites
ChatGPT Search's underlying mechanism has changed, beginning large-scale targeted scraping of specific websites.
Promptwatch monitoring shows that after August 8, the proportion of site: queries rose from an average of about 0.37% to 16.8%, an increase of approximately 45.5 times.
The average number of background searches per response has nearly doubled.
The new process may first discover candidate sources across the web, then lock onto specific domains to find documents, rules, and updates.
The GEO strategy has thus become a two-layer competition: domains are selected first, and pages then compete for citations. Websites should prioritize building a first-hand content system that is crawlable and verifiable.
Search traffic is concentrating on official and authoritative domains, with changes in event-driven citation structures putting pressure on sites relying on UGC or aggregated content, while first-hand authoritative content sites benefit.
Source: Public information
ABAB AI Insight
OpenAI adjusted its search behavior after the GPT-5.6 update, with monitoring platforms like Promptwatch capturing a surge in the use of the site: operator, continuing the evolution of AI search from scraping open web pages to prioritizing trusted sources.
Targeted queries prioritize official domains, motivated by the aim to enhance factual accuracy and reduce low-quality content interference, strategically replacing global ranking with domain-level trust filtering.
This is similar to Google's early weighting of authoritative sites and the current preference of AI Overviews for official sources, marking a transition in generative search from "broad citation" to "precise verification."
Essentially, this is a reconstruction of information filtering in technological substitution. Search is shifting from keyword matching to entity and domain trust, with mechanisms locking in through site: to reduce hallucinations and noise, forcing content strategies to shift from page SEO to domain-level authority building.
ABAB News · Cognitive Law
Targeted scraping allows domains to enter the scene before pages.
First-hand verifiable content is the new citation ticket.
AI search's trust filtering is harsher than traditional ranking.