2.20Advanced10 min

AI Content and Google: Enforcement and EU AI Act

Lucas Blochberger·7 April 2026·Updated 11 June 2026

Definition

Google has been issuing aggressive manual actions for "scaled content abuse" since approximately June 3, 2025, completely deindexing affected sites. EU AI Act Article 50 requires transparency labeling for AI-generated public-interest text from August 2026, with penalties up to €35 million or 7% of global annual turnover.

Key Takeaways

✓Google Manual Actions for Scaled Content Abuse since approximately June 2025
✓Position-1 results are 8x more likely to be human-written
✓17.31% of top-20 results are AI-generated (Sept 2025, peak at 19.56%)
✓86.5% of pages contain some AI-content, but only 4.6% are fully AI
✓EU AI Act Art. 50: Transparency requirement from August 2, 2026
✓Exception: "Genuine human review" + editorial responsibility
✓Penalties: Up to €35 million or 7% of global annual revenue

The regulation of AI content is tightening simultaneously from two sides: Google is enforcing manually, the EU is regulating by law.

Google Enforcement

Since approximately June 3, 2025, Google has been sending waves of manual actions for "Scaled Content Abuse" and completely deindexing affected sites. The Quality Rater Guidelines (January 2025) defined generative AI for the first time and specify that pages that are "completely or nearly completely automatically generated" without "effort, originality and added value" deserve the lowest quality rating.

The Data

17.31 percent of the top-20 results are AI-generated (September 2025, Originality.AI). 86.5 percent of pages contain some AI content, but only 4.6 percent are completely AI-generated (Ahrefs, 600,000 pages). Position-1 results are 8 times more likely to be human-written (Semrush, 42,000 pages).

EU AI Act Article 50

The transparency obligations come into force on August 2, 2026. AI-generated text for public information requires labeling, except in cases of "genuine human review" and editorial responsibility of a person or organization. Penalties: up to 35 million euros or 7 percent of global annual revenue.

Data & Statistics

45 Prozent weniger minderwertige, unoriginale Inhalte in den Suchergebnissen nach dem March 2024 Update (erwartet waren 40 Prozent); Rollout abgeschlossen am 19. April 2024

Google (The Keyword / blog.google) - Google Search update for March 2024 [international] (2024)

Drei neue Spam-Richtlinien 2024: Scaled Content Abuse, Expired Domain Abuse, Site Reputation Abuse; die Scaled-Content-Richtlinie greift unabhängig davon, ob Inhalte durch Automatisierung, Menschen oder eine Kombination entstehen

Google Search Central Blog - What web creators should know about our March 2024 core update and new spam policies [international] (2024)

837 von 49.345 beobachteten Sites vollständig deindexiert (rund 1,7 Prozent); 100 Prozent der betroffenen Sites mit Anzeichen von KI-Content, 50 Prozent mit 90 bis 100 Prozent KI-Beiträgen

Search Engine Journal - Google's March 2024 Core Update Impact: Hundreds Of Websites Deindexed (Daten: Ian Nuttall / Originality.ai) [international] (2024)

74,2 Prozent von 900.000 neu erstellten englischsprachigen Webseiten enthielten KI-Content; nur 2,5 Prozent reiner KI-Text, 25,8 Prozent rein menschlich, 71,7 Prozent Mensch-KI-Mix

Ahrefs Blog - What Percentage of New Content Is AI-Generated? [international] (2025)

GPTBot ist der meistblockierte KI-Crawler mit 5,89 Prozent aller Websites (bis 7,3 Prozent über alle robots.txt); ClaudeBot-Blockrate wuchs um 32,67 Prozent, das stärkste Wachstum aller Crawler

Ahrefs Blog - The AI Bots That ~140 Million Websites Block the Most [international] (2025)

EU AI Act Artikel 50 gilt ab 2. August 2026: maschinenlesbare Markierung synthetischer Inhalte durch Provider, Offenlegung von Deepfakes und öffentlich informierendem KI-Text durch Deployer

EU Artificial Intelligence Act - Article 50: Transparency Obligations [EU] (2026)

Bußgeldrahmen: bis 35 Mio. Euro / 7 Prozent Jahresumsatz (verbotene Praktiken, Art. 5); bis 15 Mio. Euro / 3 Prozent (sonstige Pflichten inkl. Transparenz nach Art. 50); KMU jeweils der niedrigere Betrag

EU Artificial Intelligence Act - Article 99: Penalties [EU] (2024)

In Österreich nutzten 2025 30 Prozent der Unternehmen ab 10 Beschäftigten KI, gegenüber 20 Prozent 2024

Statistik Austria - IKT-Einsatz in Unternehmen [Österreich] (2025)

Mobil 2024: 59 Prozent gute LCP-Werte, 74 Prozent gute INP-Werte, 79 Prozent gute CLS-Werte; LCP ist der am schwersten zu bestehende Core-Web-Vitals-Wert

HTTP Archive - Web Almanac 2024, Performance chapter [international] (2024)

Bei AI Overview klickten US-Nutzer nur in 8 Prozent der Fälle auf ein klassisches Ergebnis (vs. 15 Prozent ohne); nur 1 Prozent klickte auf einen Quell-Link im Overview

Pew Research Center - Google users are less likely to click on links when an AI summary appears [USA] (2025)

AI Overviews korrelieren mit einer um 58 Prozent niedrigeren durchschnittlichen Klickrate auf Position 1 (Update Dezember 2025; April-2025-Studie: minus 34,5 Prozent)

Ahrefs Blog - Update: AI Overviews Reduce Clicks by 58% [international] (2026)

FAQ

Does Google automatically penalize AI-generated content?

No. Google does not categorically classify AI content as spam. What matters is the primary purpose: automation including generative AI is only considered spam if the main purpose is to manipulate rankings. The Scaled Content Abuse policy from the March 2024 Core Update applies regardless of whether content is created through automation, human effort, or a combination. An AI-assisted article with genuine added value, evidence, and experience is not a violation; mass-produced, unoriginal pages without utility are.

What is Scaled Content Abuse?

Scaled Content Abuse is one of three new spam policies Google introduced with the March 2024 Core Update. Google defines it as creating many pages with the primary purpose of manipulating search rankings rather than helping users. Typical is the creation of large volumes of unoriginal content with little or no value to users. The policy applies regardless of creation method, including purely human or hybrid-generated mass content. Violations can lead to complete deindexing.

When does the AI content labeling requirement under the EU AI Act take effect?

Article 50 of the EU AI Act takes effect on August 2, 2026. Providers of AI systems must mark synthetic audio, image, video, or text content as artificially generated in a machine-readable format. Deployers must disclose deepfakes and disclose AI text published with the purpose of informing the public about matters of public interest. The text requirement does not apply to every AI text, but to content with a public information purpose.

What are the penalties for violations of the AI Act transparency obligation?

For violations of the transparency obligations under Article 50, a framework of up to 15 million euros or 3 percent of worldwide annual turnover applies, whichever amount is higher. The frequently cited 35 million euros or 7 percent applies exclusively to prohibited practices under Article 5, not the labeling requirement. For SMEs and start-ups, the lower of the specified amounts or percentages applies.

Should I block AI crawlers like GPTBot and ClaudeBot in robots.txt?

This is a strategic consideration. A block via robots.txt protects content from training and response use, but reduces the chance of being cited in LLM responses, which harms GEO visibility. According to international surveys, GPTBot at 5.89 percent is the most blocked AI crawler, with ClaudeBot's block rate growing fastest. It makes sense to differentiate between training crawlers and retrieval crawlers rather than blanket blocking. Blocking Google-Extended does not affect normal indexing by Googlebot.

How does a site recover after a Manual Action for Scaled Content Abuse?

In case of a manual action, the cause is stated in Google Search Console. The process: remove or substantially improve affected thin and mass content, then submit a reconsideration request. After successful review, the action can be lifted within weeks. If the site was algorithmically devalued by a Core Update, there is no notification and no reconsideration request; the only solution is to improve content quality site-wide and wait for re-evaluation in a later update, which can take months.

How do a manual action and an algorithmic devaluation differ?

A manual action is triggered by a Google reviewer, reported in Search Console, and requires a reconsideration request after remediation. An algorithmic devaluation by a Core Update does not appear as a notification, but only shows in traffic decline. It cannot be resolved through a reconsideration request, but only recovers with a later update when Google re-evaluates the improved content quality.

How does your website perform?

Get a free, AI-powered SEO report of your website by email: technical SEO, on-page, keywords & competitors. No obligation.

Get a free SEO audit →

Previous← Edge SEO: CDN-Level Optimization with Cloudflare Workers NextConsent Mode v2 and Privacy Sandbox: What Remains After 2025 →