SEO Experiment Shows Google’s Spam Update Penalizes AI Affiliate Sites

By Charles Ryder

If you’ve followed SEO over the last few years, you’ve seen a familiar formula: buy a partial‑match domain, launch a template, publish thousands of AI-generated pages targeting long‑tail “best + location” queries, then cash in with affiliate links. For a while, it worked—especially after mid‑2023, when tools like ChatGPT made scaled content production cheap and fast.

Then Google’s spam systems caught up, and it happened in public.

An SEO experiment published February 26, 2026 showed what happens when you stack a portfolio with every “low-trust” signal you can think of and run it into modern Google spam updates. The outcome was blunt: traffic dropped to zero, and it stayed there.

The AI affiliate boom: why programmatic SEO exploded

AI-driven programmatic SEO (pSEO) took off because it removed most of the friction from scaling. Marketers realized they could:

• scrape public datasets,
• feed them into templated pages,
• auto-generate “unique” copy at scale,
• and interlink aggressively,

Suddenly, building a 5,000–50,000-page affiliate site didn’t require a newsroom or a meaningful budget.

The targets were high-intent commercial queries like “best welding schools,” “best plumbing schools,” and “best electrical schools”—terms close to conversion. The bet was straightforward: Google rewards coverage and specificity, so create pages for every state/city/variant, wrap them in a decent-looking template, and let the long tail drive the clicks.

That bet paid off for a while. The “for a while” is doing a lot of work here.

The experiment: three PMDs, thousands of AI pages, fast indexing

In the experiment, SEO operator Tim Kraft bought three partial‑match domains built around “best [trade] schools” queries. Using a Python API plus ChatGPT, he generated thousands of pages programmatically.

Key characteristics of the sites (and why they matter):

• AI-templated content at scale: not a handful of assisted posts, but mass-produced pages.
• Public data aggregation: information compiled from available sources without meaningful original research.
• Minimal brand/authorship signals: little to no creator presence, expert oversight, or trust-building.
• Aggressive internal linking: classic pSEO architecture meant to spread authority and push rankings.
• Commercial intent baked in: the structure existed to rank and funnel clicks into affiliate outcomes.

The sites didn’t fail on launch. They indexed quickly. Impressions rose. Each site pulled roughly ~200 in-market clicks in the first couple of months. Anyone who has shipped pSEO projects knows that early traction can feel like confirmation.

In hindsight, it looks more like a probation period—long enough for Google to observe patterns before deciding whether the sites deserved to stick around.

The turning point: Google’s spam updates get sharper (2024–2025)

Google didn’t flip a switch one day and decide it “hates AI.” The public rollout shows a steady tightening around scaled content abuse:

• March 2024 Spam Update: explicitly targets scaled content abuse, including AI-generated spam.
• June 2024 Spam Update: continued spam detection improvements.
• December 2024 Spam Update: broader spam targeting, including site reputation abuse.
• August 2025 Spam Update: long rollout with steep drops for spammy sites—some hit within 24 hours.
• December 2025 Core Update: heavy volatility and broad quality reassessments, with many reports of demotions for AI-heavy, low-quality pages and lifts for stronger creator brands.

Google’s policy position has stayed consistent: AI isn’t “banned.” But automation used primarily to manipulate rankings violates spam policies. That distinction matters. “AI-assisted editorial workflows” and “AI-generated pages designed to rank” don’t get treated the same when trust signals are weak or missing.

The penalty: clicks drop to zero—and nothing recovers

In December 2025, the experiment’s sites flatlined. Clicks went to zero across all three.

No partial rebound. No quick refresh fix. No recovery from minor on-page tweaks.

Here’s the uncomfortable takeaway: when a site’s only defensible value is ranking, it has nothing left once ranking is removed.

Tim Kraft’s summary captured the risk of low-trust AI affiliate builds: combine thin value, weak credibility, and scaled automation, and the entire project depends on one fragile thing—Google’s willingness to rank it.

When that willingness disappears, the business disappears with it.

Why Google hit these sites (and why some pSEO brands still survive)

It’s easy to reduce this to “Google penalizes AI content.” That’s not accurate, and it doesn’t help anyone build better sites.

What Google appears to be targeting is the combination of signals associated with manipulation:

• scaled page generation with little editorial oversight,
• templated layouts where the keyword changes more than the substance,
• lack of real experience or expertise,
• no recognizable creators, reviewers, or brand entity behind recommendations,
• affiliate intent without unique value,
• and link structures built to push rankings rather than help users.

In other words: the footprint.

This also explains the programmatic SEO paradox. Some companies publish tens of thousands of pages and still do well. They earn that scale with real utility—original datasets, tools, community validation, product-led value, strong authorship, and brand trust. Scale isn’t the problem. Untrusted scale is.

What I’m taking from this as an operator

This experiment is a warning shot for anyone still treating AI as a content loophole.

1. Depth beats breadth in affiliate SEO
   Fewer pages with real differentiation (hands-on testing, expert input, original comparisons, updated data) will outlast thousands of liquid templates.
2. E-E-A-T isn’t a checklist—it’s the moat
   Authorship, editorial standards, reviewer bios, citations, firsthand experience, and a real brand presence aren’t optional anymore. They’re survival traits.
3. “SEO-only” sites are increasingly brittle
   If your site has no audience outside of Google, one update can wipe out your channel. Diversification isn’t a nice strategy—it’s risk management.
4. AI needs to be a workflow, not the product
   AI can speed up research, structure, drafting, and updates. But if the output reads like it was generated to fit a keyword pattern, Google is getting better at treating it that way.

Conclusion: the affiliate era isn’t dead—low-trust AI scaling is

Affiliate marketing isn’t going away, and programmatic SEO isn’t “over.” What’s fading fast is the idea that you can mass-produce affiliate pages with minimal oversight, minimal credibility, and minimal originality—and expect Google to bankroll it indefinitely.

If you’re building with AI in 2026, treat quality control, editorial process, and trust signals as infrastructure, not polish you add later. If you want a practical way to operationalize AI content without drifting into scaled-spam territory, consider a platform designed for responsible workflows and oversight—AIuthority is one option.