Does using AI-assisted content automatically trigger a Google ranking penalty? Research published by Ahrefs on July 27, 2026, covering roughly 331,000 pages across three datasets, does not support a simple yes-or-no answer. Pages estimated to contain heavy AI contribution can rank and can enter Google’s index. The same research also finds that those pages are associated with weaker organic visibility on average.
For brands, the useful conclusion is not about which tool opened the first draft. It is about how much original information, expertise, editorial control, and user value the page gains before publication. The study does not prove that Google applies an “AI percentage” threshold. It does provide measurable evidence for why scaled, low-value production remains a risky operating model.
What did the 331,000-page study measure?
Ahrefs used three distinct samples. Ranking analysis covered one million URLs from the top 10 positions across 100,000 search results in June 2026. Of those, 150,000 contained the minimum 350 words required for AI-content estimation. For indexation, Ahrefs sampled one million URLs, one page per domain, and found roughly 100,000 that were long enough to classify. The traffic analysis compared Google Search Console impression trends for 80,861 live pages.
Classification came from Ahrefs’ own AI content detector. Pages were grouped as low, below 20%; moderate, 20% to 50%; high, 50% to 80%; and very high, at least 80% estimated AI contribution. “Estimated” is the essential word. The detector does not observe the real production process. It infers probability from linguistic patterns. Ahrefs explicitly notes that its method is imperfect and different from whatever systems Google may use.
The detector should not be treated as a quality score for an individual URL. Across hundreds of thousands of pages, however, the dataset can reveal directional patterns in how production models and search performance move together.
The five findings that matter most
First, 5.3% of pages in positions one to three were classified as fully AI-generated, while 9% had at least 80% estimated AI contribution. Heavy AI use therefore does not automatically prevent a page from reaching the top of Google. At the same time, 82.2% of top-three pages had less than 50% estimated AI contribution. Pages with stronger human contribution remain the clear majority at the top.
Second, heavy AI contribution increases slightly as ranking position falls. Very-high-AI pages represented 8.4% of position one and 11.7% of position 10. The difference is not dramatic, but its direction is consistent. Average estimated AI contribution also rose from 27.1% in position one to 30.9% in position 10.
Third, the measured indexation rate was 49.28% for the low-AI group and 40.35% for the very-high group. That gap of almost nine percentage points deserves attention. Yet roughly 40% of the very-high group was still indexed, which is strong counter-evidence against a binary AI filter.
Fourth, low- and moderate-AI pages received two to three times more organic impressions than high- and very-high-AI pages. This was the largest performance gap in the study. Fifth, the heavily automated groups did not show a shared, sudden collapse after three to six months. Their impressions were lower but comparatively stable across the observed periods.
What the findings do not prove
The data does not prove that AI use directly causes weaker performance. The relationship may also run through other variables. Newer or lower-authority sites built mainly for search traffic may use AI more aggressively. Established brands may invest more in expert editors, original research, custom visuals, distribution, and brand demand, so they use less automation and perform better for several reasons at once.
The ranking sample also contains selection bias because every URL in the top 10 had already passed Google’s threshold for ranking. The indexation study used a blended definition as well. Ahrefs counted a page as indexed if it had at least one identifiable ranking keyword, one Search Console impression since January 2026, or an exact-URL result from a Google site search.
Reading the study as “AI content is safe” would therefore be as misleading as reading it as “Google penalizes AI writing.” A more defensible conclusion is that heavy AI use frequently appears inside the same production environments as low quality, and that the observed performance gap may be shaped by that combination.
Google’s official position has not changed
Google’s official documentation focuses on purpose and output rather than the production tool. Since 2023, Google Search Central has said that appropriate use of AI or automation is not against its guidelines. If content is original, useful, trustworthy, and created for people, the tool used to assist production is not automatically a negative signal. AI use does not create a special ranking advantage either.
The boundary appears in Google’s scaled content abuse policy. Producing many unoriginal pages with little or no user value, primarily to manipulate rankings, can violate the rules. Google does not limit that definition to AI text. Low-value content created through automation, human effort, or a combination of both can fall within the same policy.
Google’s current generative AI guidance emphasizes accuracy, quality, and relevance. That responsibility includes titles, descriptions, structured data, and image alternative text. When automation plays a substantial role, explaining who created the content, how it was produced, and why automation was useful can give readers important context.
The decision for brands operating in Turkey
Content scale is particularly tempting in ecommerce, healthcare, finance, tourism, and local services. Hundreds of product descriptions or location pages can be generated quickly. A language model can increase volume, but it cannot manufacture authority or trust. Incorrect prices, outdated regulations, invented features, and pages that miss search intent do not become durable assets merely because they were indexed.
That is why SEO services and content production should work inside one operating plan. Technical teams can monitor crawling and indexation while editors add original experience, data, examples, and visuals to each page. Digital marketing measurement should then determine whether the content produces qualified demand and conversions, not only visits.
What to do now: a six-step checklist
1. Segment the inventory by risk and value signals, not by an AI detector score. Large groups of pages built from the same template, unsupported claims, URLs with no organic impressions, weak engagement, and low conversion contribution should enter the first review batch.
2. Look for a genuine information gain on every page. If there are no customer questions, expert views, product-use evidence, field experience, original photography, first-party data, or clear comparisons, the page may only be repackaging what already exists online.
3. Position AI as a research and drafting assistant, not the final publisher. Source discovery, outlines, and variations can be accelerated. Facts, brand claims, legal language, and specialist recommendations still require expert verification.
4. Make authorship and process visible. Readers should understand who prepared the content, why that person is qualified to address the topic, and how automation was used when that context would reasonably matter. Editorial accountability is especially important in health and finance.
5. Move measurement beyond URL count. Track indexation, query diversity, qualified organic sessions, assisted conversions, revenue contribution, and maintenance cost together. More pages do not automatically create more business value.
6. Do not delete weak content blindly. Evaluate consolidation, rewriting, adding evidence, redirects, and noindex first. Apply changes to controlled page groups and monitor Search Console and analytics data for several weeks before expanding the decision.
Where to act now and where to wait
Act now when the production line is clearly optimized for volume. If teams create pages for tiny keyword variations, publish without sources, and limit editing to grammar checks, the workflow needs redesign. There is no reason to wait for Google to announce a new “AI update” before fixing a low-value system.
Wait before penalizing individual pages according to an AI detector score. Detectors can produce false positives and do not measure real usefulness. When a page loses performance, review search intent, technical issues, competitor value, site authority, freshness, and conversion contribution together.
Fark Studio perspective
The most valuable part of this research is that it replaces the question “Should we use AI?” with a better one: Does our production system add verifiable value to every page? If the answer is no, the issue is broader than tool choice. The brief, expert input, source verification, visual production, editorial standards, and measurement design are all incomplete.
If your brand plans to scale an AI-assisted content inventory, Fark Studio can audit the SEO and editorial risk before production expands. The goal is not to ban AI. It is to build the control layer that turns automation into original, useful, and measurable work.
Sources
Ahrefs, Google Doesn’t Punish AI Content; It Punishes Bad Content (331k Pages Studied), July 27, 2026.
Google Search Central, Google Search's guidance about AI-generated content, February 8, 2023.
Google Search Central, Google Search's guidance on using generative AI content on your website, updated December 10, 2025.
Google Search Central, Spam policies for Google web search, updated May 15, 2026.
