What problems could arise when AI cites AI-generated content?
When AI cites AI-generated content, it may create a loop of repeated summarizing, rewriting, and citation. This can make it harder for readers to find original sources and assess how much independent evidence supports the content. However, the articles cited here mainly show that this kind of citation occurs and suggest that sources may become harder to trace. They do not establish that citing AI content necessarily distorts information. Whether content uses AI, whether its sources can be traced, and whether it is accurate are three separate questions.
Citation loops have been observed, but that does not necessarily mean quality declines
In a July 2025 article, Ahrefs author Si Quan Ong reported that, among the pages cited by Google AI Overviews (AIO) that the study examined, 87.8% contained at least some AI-assisted content. He described Google’s AI summaries citing other AI-generated content as a feedback loop.
Ong also offered another explanation: this may partly reflect the makeup of content on the web. The article cited an earlier study of 900,000 pages, which found that 74% of new web pages contained AI-generated content. The fact that many cited pages use AI is therefore not enough, on its own, to prove that the system favors AI content. Nor does it directly demonstrate a decline in the quality of the cited content.
Repeated summarizing and republication may make the original evidence harder to trace
In an August 2026 roundup of discussions, CXL author Tarek Reslan noted a concern: a human writes content, AI summarizes it, a publisher generates an article based on that summary, and another AI then incorporates and summarizes it again. This information chain is a scenario illustrating a risk, not a process the article has established as widespread.
In this scenario, the article readers ultimately see may be several layers removed from the original material. Readers can assess it by asking:
- Does the citation point to the original content or someone else’s summary of it?
- Can the sources be traced back to the original author and full context?
- Are different articles merely restating the same material rather than providing distinct evidence?
These checks do not treat all AI content as untrustworthy. They help readers avoid mistaking repeated retellings for independent corroboration.
Faster spread remains a prediction, and detection figures do not measure accuracy
In May 2025, Ahrefs author Ryan Law predicted that continued citation of AI-generated content in new material would accelerate its spread. Ong also suggested that creators’ increasing use of AI could foster an ecosystem in which machines provide content to one another. These are the authors’ interpretations of trends, not inevitable future outcomes.
Likewise, “contains at least some AI-assisted content” does not mean “entirely generated automatically by AI,” and the detected proportion of AI content is not an accuracy rate. The articles cited here provide no measurements of how many errors this kind of citation causes. A more cautious conclusion is: AI citing AI makes it more important to check where content comes from; its trustworthiness still needs to be judged on specific sources and evidence.
AI citing AI-generated content: perspectives from across the web
If multiple articles mention the same information, can they count as independent corroboration?
Not necessarily. The discussions compiled by CXL describe a possible information chain: AI first summarizes an original text, a publisher generates an article from that summary, and other AI systems then incorporate and summarize it. If multiple articles follow the same chain, a higher article count does not mean more independent evidence. This scenario illustrates a risk; it does not establish that all similar articles come from the same source. Each still needs to be traced individually.
If many pages cited by AI summaries use AI, does that mean the system favors this content?
That cannot be determined from the percentage alone. A 2025 Ahrefs study reported that 87.8% of the examined pages cited by Google AI Overviews contained at least some AI-assisted content. The author also noted that this may partly reflect the composition of content on the web. That percentage describes the examined sample; it is not an accuracy rate for the cited pages, nor does it mean that the entire pages were AI-generated.
Will repeated citation of AI content necessarily accelerate its spread in the future?
Not necessarily. Faster spread is a prediction, not an established outcome. Ahrefs author Ryan Law argued that continued citation of AI-generated content in new material could accelerate its spread. Si Quan Ong suggested that creators' use of AI could give rise to an ecosystem in which machines provide content to one another. Both describe possible trends; neither establishes how quickly the content will spread in the future or that information will inevitably become distorted.
References
- 74% of New Webpages Include AI Content (Study of 900k Pages)ahrefs.com · 2025-05-19T11:21:50+00:00 — Discusses: Will repeated citation of AI content necessarily accelerate its spread in the future?
- AI Overviews Cite AI-Generated Content More Than Human Writingahrefs.com · 2025-07-14T09:00:12+00:00 — Discusses: If many pages cited by AI summaries use AI, does that mean the system favors this content?; Will repeated citation of AI content necessarily accelerate its spread in the future?
- ChatGPT Is citing Reddit less. What does It mean for marketers?cxl.com · 2026-08-27T16:10:28+00:00 — Discusses: If multiple articles mention the same information, can they count as independent corroboration?


Join the discussion 0 comments
What are you working through? Share a question or experience and let’s find the next step.
No comments yet. Start the conversation.