Google March 2026 Spam Update & AI Patent: What Actually Changed

What the March 2026 Update Actually Targeted

Google's March 2026 Spam Update — rolling out across two weeks — generated significant ranking volatility, particularly for sites in the AI-generated content space, affiliate marketing, and thin local landing pages. Simultaneously, a newly published Google patent surfaced details about how their systems detect and discount AI-generated content at scale.

This post cuts through the speculation and focuses on what the evidence actually shows: which sites were affected, what signals the patent describes, and what the practical implications are for ethical SEO practitioners.

The Update's Apparent Targets

Based on analysis of ranking volatility data (SERPstat, Semrush Sensor, Mozcast all recorded significant volatility between March 8-22, 2026), the sites most affected shared common characteristics:

Mass-scaled AI content sites: Sites that had published thousands of AI-generated articles in 2024-2025 with minimal human editorial involvement saw dramatic traffic drops. This aligns with Google's Helpful Content system, which was incorporated into the core algorithm in 2023 — March 2026 appears to have been a recalibration of its sensitivity thresholds.

Thin affiliate aggregators: Affiliate sites with minimal original content (primarily product specifications pulled from manufacturer pages, thin reviews with no genuine product testing) lost significant rankings in product and "best X" queries.

Doorway page networks: Sites generating templated local landing pages at scale ("plumber in [city]" pages with only the city name differentiated) were filtered from local results.

What was largely unaffected: Sites with genuine editorial oversight, original research, demonstrated author expertise, and clear E-E-A-T signals.

The AI Detection Patent:What It Reveals

The patent in question (US Patent Application 20260089127, assigned to Google LLC, published March 2026) describes a system for "Assessing Linguistic Authenticity in Web Documents" — colloquially, AI content detection.

Key mechanisms described in the patent:

Perplexity scoring at the passage level: Large language models produce text with statistically predictable patterns — lower "perplexity" than human writing for the same topic. The patent describes scoring individual passages, not just whole documents, allowing detection of AI sections within otherwise human-written content.

Stylistic consistency analysis: Human writers have stylistic fingerprints — consistent syntactic patterns, characteristic vocabulary choices, idiosyncratic phrasing. AI-generated text lacks these fingerprints across document sets, even when individual documents appear coherent.

Temporal pattern analysis: The patent describes analyzing the rate of content publication relative to a site's historical baseline. A site that published 5 articles/month in 2023 and suddenly publishes 500 articles/month in 2025 is flagged for additional analysis — the volume pattern is itself a signal.

Entity attribution consistency: Does the claimed author (if named) have consistent attribution across the web? Does their writing style match their attributed content? Mismatches between claimed human authorship and detected AI patterns are weighted negatively.

What This Means for Ethical Content Strategy

The patent reveals that Google's detection operates at multiple levels simultaneously — not just "is this text AI-generated?" but "is the entity claiming authorship a real expert, and does the content match the depth of expertise that entity would have?"

This reinforces several principles:

AI as a tool, not a replacement: Using AI to outline, draft, or accelerate content that a genuine expert then reviews, enriches with original insight, and publishes under their verified identity — this is defensible. Pure AI-to-publish pipelines are now high-risk.

Original data and experience are the moat: The types of content AI fundamentally cannot generate without human input — original research data, first-hand experience, case studies from real client work — are exactly what Google's systems are calibrated to find and reward.

Author entity signals matter more than ever: Implement Person schema for all authors, maintain consistent publication history under real names, and build author entity signals (LinkedIn profiles, speaking appearances, published research) that create a defensible digital footprint.

For a deeper look at AI content strategy, see our AI-generated content and Information Gain guide.

At Verdant Mindset, our content strategy is built on genuine expertise and original research — the approach that algorithms increasingly demand. See our sustainable SEO services.