Ahrefs’ study of AI content in Google Search examines 331,000 pages using its own AI detector, ranking data, indexation signals and Search Console impressions. It finds AI-heavy pages throughout the top ten and no sudden traffic collapse over two six-month samples. From this, the headline concludes that Google does not punish AI content; it punishes bad content.
The research gets an important point right. AI-generated pages can be indexed and can rank first. That is useful evidence against the fear that Google detects every AI-written page and automatically excludes it. The article also openly acknowledges imperfect detection, selection bias and the many factors affecting organic performance.
But disproving a blanket ban is not the same as proving neutral treatment. A ranking system could allow some AI-heavy pages to succeed while still applying softer, contextual or site-level disadvantages elsewhere. The presence of AI content in top positions establishes possibility, not equal treatment. Finding no hard cutoff cannot support the headline’s much broader verdict.
The measurement problem makes that distinction more important. Ahrefs notes that its detector differs from anything Google might use. The detector estimates whether text resembles AI output; it cannot observe how a page was produced, how heavily a person edited it or whether original research was added. Its score may also correlate with repetitive language, predictable structure and other qualities that affect performance independently. The study is therefore comparing detector classifications, not controlled examples of the same content written with and without AI.
More importantly, “bad content” is never measured. The analysis does not assign a quality score or control for authority, links, topic, search intent, site age, publishing volume, editorial review or brand demand. Any of these could help explain why pages with higher detected AI levels receive fewer impressions and are indexed less often. The article’s quality explanation is plausible, but it remains a hypothesis layered onto observational data.
Those weaker results deserve more attention than the headline gives them. Low-AI pages had a higher indexation rate, while low and moderate groups received substantially more impressions. The six-month charts show no dramatic later collapse, but stability cannot reveal whether a disadvantage was already present when measurement began. Nor can a sample of surviving pages show what happened to comparable pages that never gained visibility.
Google’s own guidance supports a content-first approach: generative AI can be useful, while scaled pages that add little value may violate spam policy. That policy tells publishers what Google wants, not precisely how every ranking system behaves. Ahrefs’ evidence is consistent with it, but does not independently prove the mechanism.
A more accurate conclusion is narrower: there is no evidence here of a universal AI-content ban, and substantial AI use does not make ranking impossible. The data cannot establish that Google never penalizes AI-related patterns, and it cannot show that quality alone causes the observed gap. Publishers should judge AI by whether it improves accuracy, originality and usefulness—not by the comfort of an absolute headline. The trend is worth studying; it is not a final verdict.