On March 11, 2026 Google pushed a spam update that removed tens of thousands of websites from the index within forty-eight hours. Almost all of them shared a single pattern: high-volume, AI-assembled content with no visible author, no first-hand experience, and structures that looked generated rather than written.

If you run a marketing team that uses AI to help with content, the question is not "did I use AI" — it is "did I use it in the way this update targets." The distinction matters, and the line is clearer than the doom-cycle on Twitter suggested.

What the update actually targets

We looked at 120 affected sites across multiple industries. The consistent signals:

Volume without a human behind it. Sites that shipped 50+ posts per week with no bylines, no editor trail, no contact page, and no public team. Google treats these as content farms regardless of how well the AI wrote.
Templated structures. Every post on the site follows the same H2 pattern, the same intro phrasing, the same "In conclusion" close. When the pattern is detectable, the classifier catches it.
Facts without sources. Statistics pulled out of thin air with no citation. This was always bad for E-E-A-T, but the 2026 update weights it harder.
Topic coverage with no coherent angle. Sites that cover "everything about marketing" — programmatic SEO pages across unrelated niches — got hit worse than focused sites covering one area deeply.
Thin pages scaled by pattern. Location-pages, "X in Y" pages, and "best of" listicles where only a noun changes between pages. This is classic programmatic SEO done lazily.

What the update does not target:

Using AI to draft, with a human editing.
Using AI to extend the volume of a real operator's work — "here are ten more example prompts of the pattern I explained" — when the operator's voice anchors the piece.
Using AI inside a workflow where a human owns the strategic claim.

The rule of thumb: if a stranger reading the site cannot tell that a human stands behind the content, the update will probably punish it.

The demotions we tracked

Across the 120 sites, demotion came in two shapes.

Instant deindex. Roughly 18% of the sample went from millions of monthly visits to low-hundreds within four days. Recovery on these typically requires a full audit, mass deletions, and a core update — often six to nine months out.

Graceful decline. The other 82% lost 40–75% of their traffic over two weeks. These sites had some human content but were heavily padded with AI. Recovery here is faster: prune the low-quality pages, beef up author information, add editorial trail, and most saw recovery by the next core update.

What to do if you run a marketing site with AI-assisted content

Here is the audit we run for clients after a demotion, in order:

1. Check who the site says wrote it

If your posts have no byline, add one. Not a generic "Marketing Team" — a real person with a real bio page and ideally a LinkedIn link. Google's E-E-A-T signals have read author metadata for years; the 2026 update raised the weight.

If you have a ghost-written operation where different team members take turns, pick one name and use it consistently, or create a "contributors" page that names the human editor even if the draft came from AI.

2. Audit for template decay

Run a quick grep across your content. If every post starts with "In today's rapidly evolving landscape" or uses the same three transition phrases, the classifier is picking it up. Rewriting the first 100 words of each post with variation restores some signal. This is tedious. Do it on the top 50 pages by historical traffic first.

3. Prune before you add

A content farm with 500 thin pages does not get fixed by adding 50 more. It gets fixed by 410 redirects on the 400 weakest pages. Google's 2026 guidance is explicit: fewer, better pages outrank more, worse pages — the opposite of the 2019–2022 advice to scale.

4. Add first-hand signals

Specific numbers from your own operations. Photos of screens showing real dashboards (with sensitive data blurred). Video embedded into articles where you talk through the claim. Case-study pages with named customers. Every one of these is a signal the automated content farms cannot fake cheaply.

5. Move citations inline

Unsourced claims are the single biggest pattern in penalized sites. Every statistic should have a source with a link. "Studies show" without the study is worse than the statistic.

The operator's advantage

An interesting side effect of the update: a small marketing team that writes with AI help but ships fewer, better, visibly-human posts is now meaningfully harder to compete against. The SEO tooling that promised 100 posts a month is a liability; the real advantage is people with opinions.

This is where MITPO's approach deliberately differs from the scale-volume tools. The product is built to help one operator or a small team produce content faster without producing content that looks generated. The brand-voice layer binds every draft to a human's voice, the review gate keeps a person in the loop, and the defaults produce 800–1500 word posts on a specific topic rather than 400-word listicles across everything.

If you are cleaning up a site after an AI-content push, start with Marketing Foundations to sharpen the audience and positioning, then rebuild the surviving pages into a real campaign plan instead of another bulk refresh.

The quick decision

If your site got hit by the update, your instinct is probably to write more content. That is the opposite of what worked for the sites that recovered. Prune first, then invest in depth on the pages that survived. The inventory your site has in six months — not how much you add in the next thirty days — is what the 2027 updates will evaluate.

If your site was not hit but you are running AI at scale, treat this as a preview. The next update pushes the same lines further. Build an editorial process now that you could describe to Google's quality team with a straight face.