Google’s March 2026 Core Update completed its rollout on April 8, 2026. Affiliate sites lost the most visibility of any content category — and the dividing line between losers and winners came down to first-hand experience and verifiable authorship. A few months and one further core update later, that reading has held up. This is a look back at what the data showed, and what it means for affiliate publishers going into the second half of 2026.

What happened

Google announced the March 2026 Core Update on X at around 6:59 ET on March 27, 2026, without an accompanying blog post. The rollout completed twelve days and four hours later, at 9:12 ET on April 8, per the Google Search Status Dashboard.

No new written guidance arrived during or after the rollout. Google pointed site operators back to the existing “Creating helpful, reliable, people-first content” page and the standard core updates FAQ.

Volatility was higher than the December 2025 update on every published metric. SE Ranking measured 79.5% top-3 URL turnover (versus 66.8% in December 2025) and 90.7% top-10 turnover (versus 83.1%). 24.1% of pages that had been ranking in the top 10 fell out of the top 100 entirely. The SEMrush Sensor peaked at 9.4–9.5 out of 10 during the rollout window.

For affiliate publishers, the practical question wasn’t how big the update was. It was whether their content sat on the side that gained or the side that lost. Industry tracking suggests the answer skewed sharply toward losses.

Google has since shipped a follow-up: the May 2026 core update rolled out from May 21 to June 2, and several trackers rated its volatility even higher than March. For this article, the direction matters more than the size. Post-rollout analyses describe the May update as rewarding the same signals of experience, authorship, and original data. Nothing in the May results points the other way, which makes March look less like a one-off correction and more like the start of a longer course.

Why affiliate content took the biggest hit

Affiliate sites lost more visibility than any other content category in this update. Third-party analyses report measurable ranking decline on roughly 71% of monitored affiliate domains. Average traffic losses ranged from 20–35% on affected sites, with over 50% on the strongest money pages.

Three patterns dominated the losers:

• Pure AI content farms. Sites built around scaled, unedited AI output saw 60–80% drops, according to multiple tracking vendors.
• Templated comparison pages without proprietary data. Pages that reproduced specs, prices, and Amazon copy without adding original measurement, photography, or testing lost rankings to pages that did.
• YMYL affiliates without verifiable author credentials. Finance affiliates publishing reviews under unnamed editors or generic bylines fell 40–55%, per Digital Applied’s tracking sample.

The pattern beneath those three: pages where Google could not identify a reason to trust the source of the recommendation. A list of “the best wireless headphones in 2026” carries weight if a named person tested them. The same list, written from public specs by an unattributed author, does not.

This is also where the smaller winning category showed up. Sites that publish original research (proprietary data, primary surveys, measured tests) gained visibility. One B2B SaaS sample tracked by Digital Applied measured a 22% visibility increase across that group through the rollout.

It wasn’t really about AI

Post-update analyses found near-zero correlation between AI assistance and ranking loss. Pages that used AI as part of a human-edited workflow were not targeted as a class. What lost was scaled content abuse: AI-generated drafts published without an editor, templated at volume, with no first-hand layer above the model output.

The distinction matters because it sets a different bar than “stop using AI.” A reviewer might use AI to draft a section. Then they test the product, rewrite the verdict in their own voice, add photos, and sign the page. That page operated inside an editorial workflow and was generally fine. A site that pushed model output to ten thousand URLs without that step was not.

Google’s Experience pillar, the first “E” in E-E-A-T, was the visible variable. Pages with specific first-hand use, original photographs, measured outcomes, and named authors with verifiable external profiles outranked the rest. They beat pages covering the same topic thoroughly but impersonally. Industry tracking found that around 73% of post-update YMYL winners had verifiable author credentials attached.

For affiliate publishers, the operational read is straightforward. AI is not the problem; absent authorship is. A page that says “we tested this for six weeks, here are the photos, here is what surprised us” reads to Google as a person. A page that says “this is one of the best options in 2026”, with nothing to back that up, increasingly does not.

What this means for ASA2 users

ASA2 has always been a manual-curation tool: you pick the products, write your own takes, and build your comparisons in the WordPress editor, with templates that present your content rather than generate it for you. That workflow happens to line up with what Google rewarded in this update.

The structural fit is direct. ASA2 sites that attach a named author and publish original photos alongside their product cards were already on the right side of the experience signal. Hand-curated collections fit the same pattern. A template doesn’t substitute for a take, but it also doesn’t get in the way of one.

That doesn’t mean every ASA2 site is automatically safe. Pages that lean entirely on the default product card without author commentary face the same exposure. So do pages without a clear byline, and pages that re-use Amazon’s marketing copy verbatim. The plugin makes the right pattern available; the site owner still has to use it.

ASA2 is already working on tooling that makes it easier to surface that author voice without leaving the WordPress editor. The tooling treats AI as an editor’s assistant for the writer, not a content generator that bypasses them. More on that in a later post.

For now, the practical move is simple. Look at the pages already performing on your site and ask one question. Would a reader believe a real person spent real time with this product? Where the answer is no, that is where this update points.

A short recovery checklist

Not every page needs the same treatment. Audit a representative slice first, then prioritise by traffic and revenue.

• Add a first-hand experience paragraph to your top-performing money pages. State the duration of use, the specific scenario, and one observation that wouldn’t be in the manual.
• Attach a named, verifiable author to every review and comparison page. Link to a LinkedIn or professional profile from the byline.
• Add proprietary data where you can: photos of the actual product, measurements, side-by-side tests, or direct quotes from buyers you’ve talked to.
• Consolidate thin pages. Merge weak duplicates into stronger consolidated pieces. Recovery comes from raising the floor, not editing every page in place.
• Audit AI-assisted copy for sections that read as generic. If a paragraph could appear unchanged on another site in your niche, rewrite it.
• Differentiate from Amazon’s product page. If your description and Amazon’s would convert to the same plain text, you have a duplication problem this update penalised.
• Review comparison tables for sources. Rows of public specs without context are part of the templated pattern.

What comes next

Recovery from a core update typically takes 4–12 weeks, and rankings usually move again at the next core update rather than drifting back in between. That next cycle has already come and gone: sites that reworked their pages after March had their first re-evaluation window during the May update. Sites that haven’t started yet haven’t missed anything permanent, since core updates now arrive every couple of months and the work done today sets the baseline for the next one. The brief is the same as Google’s quality guidance: write like a person who has used the thing.

For an in-depth read on what the update changed, see Search Engine Land’s post-rollout analysis and the rollout-complete coverage.