The March 2026 Core Update finished rolling out two days ago. We’ve been watching it move across 200+ client sites in real time, and the picture that’s emerging is clearer than the noise on Twitter would suggest.

Here’s what the data actually shows.

Headline pattern: helpful-content keeps winning

Sites we’ve shipped engagement-led content into over the past 18 months are sitting flat or up. Sites where we only did link work and left content alone are seeing 3–8% organic traffic dips. Sites running heavy AI-drafted content with no human edit pass are seeing 15–40% drops.

This isn’t subtle. The classifier has gotten sharper, not softer.

What lost ground

Three buckets of pages are getting hit hardest:

  1. Thin product-comparison pages — "X vs Y" pages with 400–600 words, no original opinion, no editorial spine. Google has been clear about this for two years. The update is the enforcement.
  2. AI-only ingredient and glossary pages — supplement and food brands publishing 200+ ingredient stub pages with no real expertise are watching them de-index in waves. We’ve seen entire glossary trees lose 70%+ of their indexed pages.
  3. Listicles with no first-party data — "10 best X" content that just summarizes other people’s listicles. The classifier appears to detect derivative content even when the surface signal looks fine.

What gained

The surprise of the update is what gained ground:

  • Long product reviews with original photography and testing are up across our portfolio. Brands that invested in actual hands-on review content (with measurements, comparison shots, testing protocols) are sitting 5–15% higher.
  • Pages with strong author entity signals — bylines tied to LinkedIn, expertise statements, credentials — are holding through the update with negligible volatility.
  • Editorial roundups on tier-1 sites (the publications we place on for AEO) are up 8–12% on average, which has the second-order effect of lifting brand mentions inside them. AI engines reading those roundups are propagating those gains forward.

What this means for the rest of 2026

The classifier behavior says one thing clearly: Google is doubling down on classifying content by whether it shows original human work, not whether it answers the question. The two correlate, but they’re not the same.

If you’re running a content program right now, three asks:

  1. Audit your AI-drafted content. If a piece was published in 2024–2025 without a human revision pass, it’s a rank liability. Either revise or de-index.
  2. Invest in author entities. A real bylined author with credible LinkedIn presence is now a measurable ranking factor. We’re starting to recommend it as a baseline.
  3. Don’t panic on link-only sites. The 3–8% dip on link-only sites is recoverable. It’s a signal that link work alone has a ceiling — pair it with engagement and you’ll hold position.

We’ll publish a deeper data cut next week with the keyword-level breakdown. If you want us to run the analysis on your portfolio, reach out — we’re doing free post-update audits through May.