Wikipedia is losing readers, Stack Overflow is losing contributors — is AI hollowing out these two sites in different ways?

This looks at whether Wikipedia and Stack Overflow face the same kind of pressure from AI. The short answer is that both are shrinking, but the studies are measuring different things. On Wikipedia the loss is in readers. On Stack Overflow it's in contributors. That gap matters more than the percentages do.

Wikipedia's loss happens at the point where someone reads. The Wikimedia Foundation reports that human pageviews fell about 8% in 2025, and it blames AI search answers and short social video. The foundation also admits the cause hasn't been independently checked Is generative AI actually driving Wikipedia traffic declines?. A cleaner test compared English Wikipedia, where Google's AI Overviews launched, with the German and French editions, where they didn't. Search referrals to English Wikipedia dropped about 5% relative to the other two Does AI search summaries divert traffic away from Wikipedia?. The mechanism is that Google's summary answers the question on the results page, so the click never reaches the source. Wikipedia still supplies the knowledge, but people see less of it under its own name. News publishers expect the same thing on a much larger scale, forecasting a 40–43% drop in search referrals over three years Will AI Overviews reduce search referral traffic to publishers?.

Stack Overflow's loss happens at the point where someone asks. The best study compares it with similar Q&A sites that ChatGPT affected less. New posts fell 16% after ChatGPT launched and reached about 25% within six months Did ChatGPT cause Stack Overflow posting to decline?. Here nothing sits between the user and the site. People simply take the question to a chatbot and never write it down in public. The drop was largest for programming languages that are best represented in ChatGPT's training data. So the topics where Stack Overflow's past answers most likely taught the model are the topics where the site is losing new questions fastest. It also isn't just low-value posts disappearing. Vote scores didn't change, which suggests well-received questions are going too, though votes are only a rough stand-in for quality Did ChatGPT displace only low-quality Stack Overflow posts?.

The difference matters because each kind of loss damages something different. When readers leave, as on Wikipedia, the articles still exist and can still be used. What weakens is the attention and goodwill that keep volunteers editing. When contributors leave, as on Stack Overflow, new knowledge stops being written down. The questions people now ask chatbots never become public pages that others can search, correct, or learn from. Other platforms show a different pattern again. On Medium and Quora, detector-flagged AI text rose from about 2% to 38% of posts, so those sites are filling up with AI writing rather than emptying out. Reddit barely changed Is AI-generated content rising faster on some platforms?. The pressure depends on what role a site plays: a reference people read, a forum people ask in, or a place people publish.

One idea in the corpus suggests a longer-term problem than lost clicks. If people increasingly hand searches off to AI agents, websites will be competing for the agent's choice rather than a human visitor's attention Will agents compete for attention just like users do?. If so, the 5–8% losses measured so far may be the early, human-scale part of a bigger change in how knowledge sites get found at all. The corpus doesn't yet have a study that compares the two sites directly on the same metric. A fair summary is that they are under related pressures, but the evidence doesn't show it is the same pressure.