Google’s spam update vs. AI affiliate sites: An SEO experiment

Remember when it was easy to rank partial-match domains and headings to commercially intended search queries?

When paired with the right methodologies and conversion-optimized widgets, you could silently earn tens of thousands of dollars in affiliate revenue per month with minimal maintenance.

It was possible to get by with just updating articles for relevancy and freshness signals, for example.

Pressure-testing Google’s spam update

Before the experiment, I had spent several months scaling an affiliate initiative in a much more above-board way for a longstanding website in a YMYL category.

We had success with hiring subject matter experts (SMEs) to write helpful, educational content that actually informed readers.

While the new content primarily targeted commercially intended keywords, that wasn’t the website’s sole purpose for existing. There were also thousands of pages of user-generated content (UGC) that inspired the new content, and visitors would navigate from them to convert, as well.

We had brand trust, original research, expert insights, and everything else you’d expect from a reputable publisher.

It was a perfect mix: verticalized legacy UGC with thousands of earned backlinks and a commercial lever that served a preexisting demand while adhering to industry best practices. It was a truly helpful experience.

The experiment: Scaling AI without trust

If the first model was built on trust and earned authority, this one would remove those signals entirely. 

During that time, influencers on LinkedIn were doing the same thing. Except they were using AI to generate thousands of pages by scraping and rewriting content, or by programmatically aggregating public data.

That’s when I searched in my couch pillows for a few dollars and bought three domains that partially matched the following queries: “best welding schools,” “best plumbing schools,” and “best electrical schools.”

The goal? Intentionally test a set of low-trust, high-scale tactics that are commonly promoted online and see how long they would persist.

I then used AI to make the websites pretty, fetched public data with a vibe-coded Python API call, and used ChatGPT to template all of the subheadings and paragraph text you would typically see ranking across the web. 

Within a few hours, with the help of liquid content, I published thousands of bottom-funnel pages across three websites. I was able to inject public data, target superlatives by program type and state, and include a directory with individual, templated pages per school.

I even leveraged aggressive internal linking practices that prioritized crawl coverage over user intent.

The setup violated almost every long-term trust signal — which made it a useful test of how the system would react.

All three sites shared the same traits:

• Zero brand signals.
• Programmatic AI-generated content.
• Public data aggregation.
• Aggressive internal linking.
• No original research or authorship signals.

Dig deeper: What 4 AI search experiments reveal about attribution and buying decisions

Confirmed: The data shows Google’s spam updates work

The websites worked briefly. Indexation was fast, pages surfaced for long-tail queries, and impressions climbed faster than expected.

Within their first couple of months, all three websites were generating about 200 in-market clicks each.

But as you can see, they flatlined hard during the first December spam update since their inception. In fact, clicks dropped to 0.

I tried turnkey data updates and adding a few performance-boosting plugins, but they never recovered.

In isolation, I’m not certain that any single one of these tactics caused the failure more than another. In combination, these tactics produced a site whose only defensible value was ranking in and of itself. Once that signal stopped being useful, nothing remained

The insight isn’t that the websites failed — it’s that Google tolerated them just long enough to learn from them.

Does affiliate content marketing still work?

Yes, affiliate content marketing still works as a monetization layer, but not as a growth engine.

There are plenty of websites that provide a helpful user experience while adhering to best practices and generating affiliate revenue.

As to how, refer to Google’s documentation on creating helpful, reliable, people-first content, where you can learn more about evaluating whether your website is publishing “content that’s created primarily for people, and not to manipulate search engine rankings.”

• “If the ‘why’ is that you’re primarily making content to attract search engine visits, that’s not aligned with what our systems seek to reward. If you use automation, including AI-generation, to produce content for the primary purpose of manipulating search rankings, that’s a violation of our spam policies.”

However, even when following best practices, the rise of AIO, the great decoupling, and dozens of other factors have made affiliate marketing less successful than it once was.

Fortunately, there’s an alternative. 

Dig deeper: Inside Google’s secret search systems: 1,200 experiments, AI agents, and entities

---

Where is content heading in 2026?

The real takeaway isn’t that Google cracked down on spam, or that affiliate content marketing stopped working. It’s that businesses built on a single, cheaply replicable distribution channel are exposed the moment that channel changes.

The next era of content will increasingly disadvantage businesses that treat search as their sole distribution channel.

Instead of focusing on easily replicable topics, many industry practitioners are shifting toward verticalized research and benchmarks that spark real conversations within communities. 

Content is no longer a series of pages intended to rank. Rather, it’s a combination of discovery, discourse, and thought leadership that spans many channels.

Discovery, discourse, and thought leadership

Hypothetical: You’re a SaaS business in the financial technology space that provides businesses with enhanced financial forecasting.

Instead of publishing landing pages that target “best financial forecasting software” or “most affordable financial forecasting software” (the SaaS equivalent of bottom-funnel ranking pages), consider doing deep dives with industry leaders who have something valuable to add to the conversation.

Rely on their insights to identify the largest gaps in financial forecasting in 2026 and validate: Does my product truly solve this? If it does, you may have found a perfect wedge into the community.

If not, there’s your roadmap.

Use these problem-and-solution insights to develop landing pages with interactive assessments paired with benchmarking reports informed by industry-leading organizations.

The “why” is that the content exists to help organizations contextualize both where they are and where they want to be.

While these assessments or studies may not rank in the first position in Google for high-volume search queries, you can instead leverage owned channels, partner distribution, paid media, and more to put them in front of your ideal clients.

These insights serve as a launching pad for communicating learnings from real conversations that aren’t easily replicated. By doing so across many different channels, you effectively enhance your ability to be everywhere.

If you execute well and provide true value, not only do you contribute to a community, but you may unlock the growth you’ve been after all along.

Companies like Stripe and its “Developer Coefficient” and HubSpot and its “State of Marketing” are doing this exact thing.

Dig deeper: 3 GEO experiments you should try this year

Content in 2026: Fewer pages, deeper moats

This model looks very different from scaling thousands of programmatic pages. It also comes with tradeoffs:

• Slower feedback loops.
• Less attributable ROI.
• Fewer “quick wins.”
• More dependence on distribution and partnerships.

In 2026, content is about fewer pages, deeper insight, a stronger point of view, and assets that are harder to replicate.

The spam update didn’t kill my niche websites for Christmas, but it exposed how thin the margin is for anything built without trust.

Search marketing isn’t about avoiding content penalties — it’s about building things that can’t be easily copied with AI.

View the full article