ChatGPT reduced how often it cites Reddit and Wikipedia in mid-September 2025, a shift captured through 230,000 prompts that reveals how quickly AI systems can reorder their preferred sources and what this means for publishers, marketers, and teams working in AI-driven search visibility.
A comprehensive study by Semrush spanning July 14 to October 12, 2025, has revealed a notable change in how major AI models reference online sources.
Analysts who monitored weekly citation patterns across ChatGPT, Google’s AI model, and Perplexity uncovered a dramatic decline in how often ChatGPT pointed users toward Reddit and Wikipedia.
The change became visible around September 11, when these two long-standing, high-volume sources began receiving far fewer citations in ChatGPT’s answers.
The dataset includes over 100 million citation events, offering a clear view of how quickly AI models can change their reference habits.
A Rapid Shift That Caught Researchers’ Attention
Throughout early August, Reddit and Wikipedia appeared frequently in ChatGPT responses.
Reddit citations peaked near 60%, while Wikipedia reached above 55%. Then, over the course of just a few days in mid-September, both experienced a steep decline.
Reddit dropped to roughly 10% of sampled answers, and Wikipedia dipped under 20 percent.
The pattern created a sharp contrast with other platforms. Google’s AI Mode continued favoring Google properties and professional networks like LinkedIn. Perplexity, meanwhile, showed only mild movements, with slight increases for sources such as Wikipedia and Forbes and a drop for Reddit.
A popular theory circulated soon after the change.
Google had discontinued support for the “num=100” parameter in its search results.
Some believed that if ChatGPT relied on deeper result pages to surface Reddit and Wikipedia links, the parameter’s removal might cause those domains to appear less often.
Several experts challenged this assumption. Semrush data showed that many Reddit and Wikipedia URLs already rank near the top of search results.
If that is true, access to deeper result pages would not have significantly affected citation patterns.
The timing aligned, but the magnitude of the drop suggested additional internal or systemic shifts.
The Domains That Rose as Reddit and Wikipedia Fell
As those two major sources lost visibility, other publishers began appearing more frequently within ChatGPT answers.
PR Newswire, Forbes, and Medium saw meaningful increases.
LinkedIn also gained traction across multiple platforms.
Google’s AI Mode showed stronger representation for YouTube and Google-owned sites.
Perplexity’s changes were subtle but consistent.
These variations highlight that each AI model relies on a distinct mix of sources, and those preferences can change independently rather than moving in unison across the industry.
Why This Matters for Content, Authority, and Marketing Teams
The way an AI model cites sources influences audience trust and shapes information pathways.
A cited domain often becomes the user’s next click, giving it a role similar to a high-ranking search result.
When citation patterns change abruptly, publishers and marketers can gain or lose visibility without warning. This makes the study’s findings directly relevant to businesses competing in search and AI-driven discovery.
These shifts show how quickly AI visibility can change. At Stan Ventures, we work with clients who need a steady presence across both search engines and emerging AI platforms, and this is exactly why we offer AI SEO services that look beyond keyword trends. Our approach strengthens authority signals and positions brands on the types of pages AI models frequently reference.
If your team is exploring how to stay visible as AI systems evolve, Stan Ventures can guide you with the right strategies.
Guidance for Organizations Preparing for Rapid Shifts in AI Behavior
Here are key considerations organizations should keep in mind as they navigate sudden and unpredictable shifts in AI behavior.
- Track citation patterns in your industry. They reveal which domains AI models currently consider useful.
- Strengthen your presence on authoritative publications that AI systems repeatedly cite.
- Diversify your publishing strategy instead of relying on one or two traffic sources.
- Add AI citation monitoring to your SEO workflow to detect early changes in content visibility.
- Produce content supported by credible insights, expert commentary, or unique data, which AI systems often favor.
Key Takeaways
- ChatGPT reduced Reddit and Wikipedia citations sharply around mid September 2025.
- Google’s “num=100” change does not fully account for the scale of the shift.
- Different AI models reacted in different ways, showing platform-specific behavior.
- Several mainstream publishers gained visibility after the drop.
Monitoring AI citations is becoming essential for SEO teams and publishers adapting to new discovery methods.
Zulekha
AuthorZulekha is an emerging leader in the content marketing industry from India. She began her career in 2019 as a freelancer and, with over five years of experience, has made a significant impact in content writing. Recognized for her innovative approaches, deep knowledge of SEO, and exceptional storytelling skills, she continues to set new standards in the field. Her keen interest in news and current events, which started during an internship with The New Indian Express, further enriches her content. As an author and continuous learner, she has transformed numerous websites and digital marketing companies with customized content writing and marketing strategies.