In a previous post, I explained how almost all SEO is becoming programmatic. We established that SEO is entering a meta-stable equilibrium with two key characteristics.

1. All marketers have access to the same foundational models. This means everyone will be able to tap into the same level of “expertise” (i.e., reasoning capabilities within their models).

2. All marketers have access to the same tools (like daydream). This means everyone will also have the same ability to distribute the content they create with that expertise.

If you’re investing in SEO using the same foundational models and tools as everyone else with no unique perspective or information, you’ll be performing Baseline SEO.

Baseline SEO: Content generated using publicly available AI models, resulting in undifferentiated output lacking unique human insights or proprietary data.

The proliferation of AI-generated content means that undifferentiated Baseline SEO articles offer no unique value to platforms like Google, ChatGPT, or Perplexity, leaving them with no reason to prioritize such content in their results.

A notable example of Baseline SEO

Twitter user Jake Ward provided a striking example of Baseline SEO last year, when he exported a competitor's sitemap, converted their URLs into article titles, and then used a tool called Byword to generate 1,800 articles. This approach perfectly encapsulates Baseline SEO: creating content without unique data points or differentiation, relying purely on a foundational model's pre-trained capabilities. The initial results seemed promising — the domain they performed this experiment on generated 490K monthly visitors. A few months later, however, the domain lost nearly all of its organic traffic as Google delivered a manual penalty that was later confirmed by Jake himself. This indicates that while Google’s algorithm still has some glaring gaps in catching bad behavior, it has a strong bias against Baseline SEO.

Add unique data or insights to go beyond Baseline SEO

To go beyond Baseline SEO, focus on contributing unique insights and information that foundational models do not have access to. While this has always been true — Google was granted a patent for an “information gain score” in June 2022, which appears to reward publishers for creating content with new and differentiated insights — it’s becoming even more important in the AI era, where “good” but not great content can be produced at virtually no cost, instantaneously, by almost anyone.

There are three ways you can enrich your content with unique insights:

1. Leverage unique, first-party data

Zapier’s integrations pages are a good example of using first-party data to enrich a page.

Zapier and Canva exemplify a data-driven approach to content creation.As Zapier onboards more app developers, the number of combinations for "How to integrate [SaaS A] and [SaaS B]" content grows exponentially. Similarly, Canva's content expansion mirrors its user activity: as users create more templates and engage more intensively with the platform, Canva rapidly generates corresponding "templates for [x]" pages.

These companies built a powerful flywheel by leveraging user-generated data to enrich their content. As their user bases grew, so did their dataset, enabling them to create more pages and programmatic page combinations. This strategy created a self-reinforcing cycle: more users led to more data, which in turn produced more diverse and valuable content, attracting even more users.

2. Distill meaning from complicated data sources with AI

Power’s programmatic page for a breast cancer clinical trial.

Vast amounts of poorly formatted data reside on government databases and websites. AI can serve as a valuable tool to distill complex, hard-to-parse information into more digestible content for the average reader. A perfect example of this is WithPower.com, which uses AI to simplify descriptions of drug trials from clinicaltrials.gov, then targets keywords in the format of "clinical trials for [specific condition]."

3. Add human insights to programmatic content

An example of one of LinkedIn’s collaborative articles.

While AI can be useful for fleshing out the skeleton of an article, it can benefit greatly from the addition of human perspectives and user-generated content. LinkedIn’s expert-recommended collaboration articles (see above) provides the best current example of this strategy at scale, but there’s still plenty of room for improvement. When you read one of these collaborative articles, you’ll find both the AI-generated and human-contributed content bland, generic, and shallow.

AI will raise the quality bar for content

As foundational models continue to improve, the level of quality required to go beyond Baseline SEO will be raised drastically by indexing layers like Google, Perplexity and OpenAI — forcing companies to differentiate by adding unique and proprietary data. As that happens, there will be less value in publishing merely "ok" or "good" content. The only content that will survive is content that adds something beyond the vast knowledge base that LLMs will accumulate over time. This in turn, should lead to the internet having more useful content, not less!

Thanks for reading! If you liked what you read here, please feel free to email [email protected] with your thoughts.