The marketing team gets the mandate: double the number of blog posts this quarter, same budget. Someone suggests AI. Six months later, organic traffic has flatlined and some of the content has dropped out of the index. It’s a pattern that repeated often enough in 2025 and 2026 that Google started describing it explicitly in its own guidelines. So the question isn’t “should you use AI for SEO content,” it’s “how do you do it so it pays off instead of costing you.”
What Google Actually Says About AI-Generated Content
Google Search Central’s official position has been consistent for years: what matters is the quality and usefulness of content for the user, not the method used to produce it. That’s not an unlimited license, though. The 2025 and 2026 updates to the Quality Rater Guidelines explicitly state that pages made “wholly or almost wholly” of AI content, without original input, experience, or verification, should receive the lowest possible quality ratings. Google draws a line between AI as a tool that supports a well-built editorial process and AI as a content factory that replaces that process entirely — and only the second one is treated as a problem.
The Data Cools the Enthusiasm: Perception vs. Reality
Semrush analyzed 42,000 blog posts tied to 20,000 keywords and cross-referenced that data with a survey of 224 SEO specialists. The result is telling: 72% of respondents believe AI content ranks as well as or better than human-written content, and 45% say AI content performance has improved over the past year. The hard SERP data says something different — fully human-written content holds the #1 position with over 80% probability, while content flagged as fully AI-generated reaches the top spot only about 9% of the time. The human-content advantage holds across the entire first page of results, though it clearly narrows from position five downward.
Interestingly, another large-scale study from Ahrefs — based on 600,000 pages ranking for 100,000 phrases — shows a seemingly contradictory picture. As much as 86.5% of top-20 pages contain at least some AI-generated content, and the correlation between the percentage share of AI in the text and ranking position was just 0.011 — practically zero. But the two studies measure different things: Semrush evaluates content classified as “pure AI” with no human intervention, while Ahrefs counts any AI involvement, even minimal, in a human-assisted process. There’s really one conclusion here: it isn’t the mere fact of using AI that determines ranking, it’s the degree of human control over the final text.
When Automation Ends in a Penalty
The risk is real and measurable. According to Originality.ai data, content flagged as AI-generated accounted for 17.31% of top-20 search result pages in September 2025, down from a peak of 19.56% in July 2025. It’s worth remembering what happened before that: after Google’s March 2024 core algorithm update, the share of AI content in top results dropped sharply to 7.43% — the clearest evidence yet that Google can selectively penalize mass-produced, low-quality AI content, even without officially penalizing the technology itself.
SEO consultant Glenn Gabe’s (GSQi) analysis of Google’s August 2026 spam update shows this in concrete cases: a site with more than 1.5 million indexed URLs, 85% of them programmatically generated with AI, lost visibility site-wide; an affiliate site with content scored at 97-100% probability of being AI-generated saw a drop across nearly 14,000 keywords. Independent site HouseFresh learned this the hard way too — after the March 2024 update it lost 91% of its organic traffic, from roughly 4,000 to 200 daily visits, as mass-produced reviews from large publishers overtook it in the results. The common denominator in these stories isn’t AI itself — it’s scale without quality control.
Where AI Actually Pays Off
The same Semrush report also shows the upside. 64% of teams use a “human-led, AI-assisted” model — the one that both Google and the hard ranking data treat most favorably. More than 65% of specialists use AI for research, editing, and text optimization rather than generating finished articles from scratch. The biggest, most consistently confirmed benefit is speed, cited by 70% of respondents, while only 19% believe AI genuinely improves content quality without additional editorial work. AI is excellent at cutting production time for briefs, headline variants, first drafts, or competitor analysis — but it doesn’t replace the expert verification and real experience that users and the E-E-A-T guidelines expect.
How to Approach This Sensibly
The practical takeaway from all this data is more measured than the marketing promises of “AI content in a minute” tools. SEO content automation works when it’s one stage of a controlled process — with a research brief, fact-checking, review by a subject-matter expert, and a genuine point of view — and it stops working the moment it turns into an unsupervised content production line. Companies that treat AI as an accelerator for their team’s work, not a replacement for it, see real time savings without losing visibility. Those that bet on scale alone sooner or later run into the same algorithm update that hit HouseFresh or the sites analyzed by GSQi.
At Unomage, we’ve designed B2B marketing automation in this spirit from day one — as support for teams, not a replacement for them, with quality control built into every stage of the process instead of bolted on at the end. If you’re wondering how to safely and measurably bring AI into your own content strategy, check out platform.unomage.com or get in touch — our team in Warsaw is happy to review your content process and show you where automation delivers a real edge, and where it’s better to leave more room for a human.
This article was created with the help of the Unomage AI platform.