Human-Written Content Continues to Outperform AI-Content

AI-generated content. We have all felt the allure of it, but is it effective?

In a word, no. Stated plainly, a recent study of 42,000 blog posts by Semrush confirmed a significant performance gap between human-written and AI-written content. 

In fact, human content was eight times more likely to rank in the #1 position on Google, dominating this highest ranking with an 80% share.

The Performance Gap, Breaking down the Semrush Data

The study analyzed 20,000 keywords and their top 10 results, uncovering that even “mixed” content, which combines human and AI efforts, only reaches the #1 rank 11% of the time, confirming that a human-led approach remains a requirement for top performance.

Looking at how various marketing teams are using AI, the focus is clearly on quantity, with 70% of teams saying they use AI to speed production, however, only 19% report an improvement in quality. This comes as no surprise to those of us in the trenches or even casual readers who have seen a massive influx of Digital Pollution, where brands sacrifice their “Source of Truth” status for the sake of output volume.

The Collapse of Low-Fidelity Content

A related 16-month experiment published by Search Engine Land provides another view of the long-term failure of unedited AI content. Researchers tracked 2,000 AI-generated articles across 20 new domains. In the first three months, the content saw promising growth, with 71% of pages indexed and a steady rise in impressions.

However, between months three and six, the system experienced a Ranking Collapse. The percentage of pages in the top 100 dropped from 28% to just 3%. Without the signals of authority, unique insight, or expert information, the search engines deprioritized the content. This supports what we have seen as well, AI content tends to be a flash in the pan, initially boosting impressions, then falling into obscurity.

The Importance of Information Gain

In order to get content to perform, as always, it needs to be valuable to your reader, but you also need to prioritize Information Gain. If your article provides the same facts as the top five results, its information gain score is zero.

Research from the Nielsen Norman Group suggests that users often find machine-written content less credible because it lacks nuanced examples. Google’s own Search Quality Rater Guidelines support this, stating that content created with little effort or originality should receive the “Lowest” quality rating. To secure a ranking, you must provide a unique perspective, proprietary data, or expert synthesis that a machine cannot simulate.

We have always stressed the importance of E-E-A-T signals to our clients, to varying success. Sometimes business goals or pet projects conflict with providing top quality content, so prioritizing search performance has always been an uphill battle. With the rise of AI-written content, this battle has become an all-out war as companies view AI as a panacea for the cost and time required for scaled content generation. 

However, as company after company begin to witness the failure of this approach, they are slowly coming around to a more sustainable integration of AI into the content production process. 

To be clear, we are not anti-AI, far from it. However, if you are treating Gemini, ChatGPT, and Perplexity as a substitute for a human staff writer, you are wasting resources and ultimately missing the value of these tools.