What does it say about the economics of content when the most visible site on the web loses significant traffic?
A status report by Wikipedia shows a significant decline in human page views over the last few months as a result of generative AI, “especially with search engines providing answers directly to searchers” [1].
Boost your skills with Growth Memo’s weekly expert insights. Subscribe for free!
- Evergreen content = Educational content covering established, timeless topics.
- Additive content = Content that provides net-new takes, insights, and conversations.
Wikipedia is an evergreen site. Even though it’s a user-generated content (UGC) platform like Reddit or YouTube, its primary purpose is to serve comprehensive definitions on established topics. Reddit, YouTube, and LinkedIn & Co. are about additive topics and insights.
AI destroys the value of one while raising it for the other.
Wikipedia’s human traffic has dipped -5% YoY, while scrapers grew by 10.5% and bots by 162.4% [2]. The fact that scrapers and bots together make up almost as much traffic as humans is symbolic of the eroding value of answering questions.
Even though Wikipedia’s direct traffic is up ~23% and Chat GPT referrals are up 3.5x YoY, Google referrals are down -35% because AI Overviews make it redundant for users to click through.
Over the same time that Wikipedia lost ~90 million visits, Google started showing a lot more AI Overviews that answer user questions directly – often based on Wikipedia’s content.
Almost 50% of Wikipedia’s queries display a large AIO at the top of the search results. That’s no outlier: Reddit is at 46% and YouTube at 38%.
Google and ChatGPT reward additive content.
YouTube’s citation rate jumped from 37% to 54% (up 17 percentage points) at the same time as Wikipedia dropped from 58% to 42% (down 16 percentage points). Video is replacing text as Google’s primary source for answers.
ChatGPT cites Wikipedia 3x more often than it mentions the site, while Reddit is at one-to-one and YouTube at ~250%! Since users don’t click citations, mentions are much more valuable. [3]
Pre-AI, the economics of evergreen content were net-positive because it attracted clicks from Google, some of which converted into customers. LLMs like ChatGPT, AI Overviews, or AI Mode are not incentivized to send out traffic but to give the best answer, which makes the experience more similar to TikTok than Search.
LLMs use web content like Wikipedia for training, but offer invisible citations instead of mentions. The net return is negative. Wikipedia has to convince donors that it’s still worth giving money, while its content is used as a utility for LLMs.
Over the last 12 months, sites offering additive UGC have gained LLM visibility [4]:
- Reddit.
- LinkedIn.
- Youtube.
- Quora.
- Yelp.
- Tripadvisor.
- Etc.
At the same time, content sites offering evergreen content lost significant amounts of organic traffic (and value):
- Stackoverflow.
- Chegg.
- Britannica.
- Wiktionary.
- History.com.
- eHow.
- Etc.
With fewer and eventually maybe zero clicks arriving [5], the value of creating evergreen content is questionable – not just for Wikipedia.
The fix is to shift focus from evergreen topics to net-new insights:
- Invest more in additive content: data stories, research, customer success stories, thought leadership, etc. Oura, Ramp, Okta, and others are already making the shift and hiring economists, journalists, and researchers. [6, 7, 8]
- Lower your investment in evergreen content in favor of additive content. We don’t know the right mix, but 50/50 or even 70/30 seems better than 80/20.
- When to keep evergreen content: For user experience (critical to understand a topic), Topical Authority, or when you can automate + enrich with unique data.
- When creating evergreen content, focus on hyperlong-tail topics aligned with your audience personas and positioning that no one else is visible for.
Evaluate additive content against influenced pipeline, LLM citations/mentions/Share of Voice, and publisher links/coverage.
Featured Image: Paulo Bobita/Search Engine Journal
