Semrush: AI Content Struggles Against Human Writing on Google

The Integration of AI in SEO Strategies

A recent study by Semrush, published on April 1, 2026, explored the impact of AI-generated content on Google rankings. By analyzing 42,000 blog articles and surveying 224 SEO professionals, the study revealed that 72% of SEO teams believe AI content performs as well as that produced by humans. However, the collected data nuances this perception.

The majority of SEO teams continue to favor a human approach in their content production. According to the study, 87% of respondents work in fully human or human-led workflows. The most common model, adopted by 64% of respondents, relies on human-led work assisted by AI. Additionally, 23% of teams still produce content without using AI.

The use of AI primarily focuses on text-related tasks, with 65% of respondents using it for research, editing, and on-page optimization. In contrast, its use is less frequent for creating visuals (28%), translation (15%), and video or audio production (9%), which require more specialized tools.

Perceived Performance of AI Content

SEO professionals recognize several advantages to using AI, including speed and assistance in ideation. Thus, 70% of respondents highlight faster and less costly content production, while 62% appreciate the help with brainstorming. However, only one in five respondents is convinced of the improvement in content quality.

Regarding SEO performance, 72% of professionals using AI content believe it ranks as well, if not better, than human content. This figure marks an 8-point increase from the 2024 study. Furthermore, 45% of respondents noted an improvement in the SEO performance of AI content over the past year, compared to only 6% who observed a decline. However, 25% of teams have not implemented tracking to distinguish the performance of AI content from that of human content.

Analysis of Blog Page Rankings

To confront the perceptions of SEO professionals with the reality of rankings, Semrush conducted an independent analysis. By collecting 20,000 keywords and extracting the top ten Google results for each, the company filtered 42,000 blog pages. These were classified using the GPTZero detector as human, AI-generated, or mixed, and then cross-referenced with their position in the SERPs.

The results show that content identified as fully human appears in the top position 80.5% of the time, compared to only 10% for AI-generated content. The gap narrows starting from the fifth position, where the difference between human and AI content becomes less pronounced.

Limitations of the Analysis

The results of this study urge caution before concluding that AI content harms SEO. Two methodological limitations should be noted. First, detectors like GPTZero have known flaws and cannot definitively determine the origin of a text. Second, the authority of a domain or article accumulates over time, thanks to backlinks and reputation signals sent to Google. Some well-ranked pages today may have gained their position years ago, regardless of their writing method.