
What happened
Promptwatch tracking shows that the share of ChatGPT Search queries containing the site: operator jumped to 16–17% on August 8, up from a baseline of 0.3–0.5% in prior weeks, following OpenAI's August 6 announcement of updates to GPT-5.6 Sol to improve fact reliability and answer focus.
Why it matters
The shift reveals OpenAI is quietly changing how ChatGPT sources information—moving toward explicit domain filtering rather than relying on general web search. This represents a measurable design shift in a tool millions use daily, affecting which websites appear in AI-generated answers and suggesting OpenAI may be responding to concerns about misinformation or source quality.
What to watch
Promptwatch also reported on August 18 that ChatGPT appeared to have reduced Reddit's likelihood of being used in searches, though OpenAI has not publicly disclosed this change and system prompts remain largely obscured from outside view.
Summaries like this, in your inbox every morning.
Promptwatch occupies a growing niche called GEO (Generative Engine Optimization)—the chatbot equivalent of SEO, where companies track and consult on how to boost a website's visibility in AI tool responses. The firm's tracking data has become a window into otherwise opaque product changes at OpenAI, Anthropic, and Google, because those companies do not publish detailed logs of their system-level adjustments. The jump from 0.3–0.5% to 16–17% usage of the site: operator over a single day on August 8 is striking because it indicates a coordinated rollout, not a gradual shift. The brief dip to 0.15% on August 3–5 hints at a staged experiment or pre-launch test before the wider deployment. OpenAI's vague public announcement on August 6 about "more reliable facts" and "more focused answers" provides cover for the underlying change but does not explain the mechanism—that ChatGPT's search function may now use a structured shape like search(query, recency, domains) rather than simply encouraging users or the model to include site: syntax. The subsequent August 18 report that Reddit sourcing appeared to decline reinforces the picture of deliberate, incremental changes to ChatGPT's information diet, though without system prompt visibility, observers cannot confirm whether the discouragement is explicit or emergent from training shifts.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Circlia started a proof-of-concept for TASTE SIGNAL: visitors to Holland Village Cafe FUKUOKA taste cookies an…

In a webinar hosted by French Mathematical Society president Isabelle Gallagher, Fields Medal winner Hugo Dumi…

NVIDIA released Nemotron 3.5 Super VL, an open-weight model with 120 billion parameters that handles text, ima…

NVIDIA released Nemotron 3.5 Bolt 74B-A9B 24GB, an AI model for local agents that hits over 80 tokens per seco…

NVIDIA published a blog post titled "AI Security Is an Engineering Problem - How to Solve It at Every Layer of…

In the test, Haiku 5.5 on medium setting averaged scores close to Sonnet 5.5 on routine documents at about 1/1…
