Commodity vs non-commodity content: the distinction that decides if you exist in AI answers

What makes content commodity

Commodity content is not bad content. It is interchangeable content.

A 2,000-word article on "what is generative engine optimization" that synthesizes the same ten sources everyone else is synthesizing, presents the same framework everyone else is presenting, and reaches the same conclusions everyone else is reaching — regardless of how well it is written — is commodity content. If it disappeared from the internet tomorrow, anyone looking for that information would find it in five other places within thirty seconds.

The test is brutal in its simplicity: if your content disappeared, would anyone lose access to information they could not find elsewhere?

For most of the web, the answer is no. And the March 2026 core update, combined with the increasingly sophisticated content evaluation of AI systems, has made that "no" visible in rankings and citation rates in a way that was not apparent before.

Commodity content exists across three structural categories.

Aggregated knowledge without original synthesis. Articles that collect and restate what others have published. Guides that summarize existing research without adding original analysis. Comparison pieces that list features from vendor websites without independent testing. The content is accurate. It is useful. But it is not original — it is a redistribution of existing knowledge with no new information generated in the process.

Generic expertise without specific application. Content written by someone with genuine expertise in a field but at a level of abstraction that applies to no one's specific situation. "Here are five principles of effective leadership." "Here is how to build a content strategy." Theoretically correct. Practically applicable to nothing specific. Any competent professional in the field could have written it.

Templated format without distinctive insight. The GEO-optimized article that has answer-first paragraphs, clear H2s, FAQ schema, and structured data — but nothing to say that the structure is not already saying. Format without substance. The optimization signals are present; the reason to be cited is not.

Commodity content can rank. It can even be retrieved by AI systems performing web searches. What it cannot reliably do is be cited — because AI systems, like human readers, tend to cite sources that contain something worth citing.

What makes content non-commodity

Non-commodity content has one essential property: it contains information or perspective that does not exist in the same form anywhere else.

This is not about being first. A piece of content published in 2024 can be non-commodity if it contains original research that has not been replicated. A piece published this morning can be commodity if it synthesizes existing sources without adding anything new.

Non-commodity content takes several forms.

Original empirical data. Research conducted from primary sources — surveys, experiments, client datasets, proprietary benchmarks. Data that you collected, that reflects reality as you measured it, and that cannot be obtained by reading other websites. A GEO agency that publishes annual data on AI citation rates across platforms, drawn from its own client monitoring infrastructure, is publishing non-commodity content. A post that summarizes other agencies' published reports is not.

First-hand experience at specific scale. Documented outcomes from real implementations, at real companies, with real numbers. Not "clients typically see improvement in AI visibility after implementing structured data" — but "after restructuring 47 product pages with FAQ schema and answer-first paragraphs for a mid-market SaaS company, AI Overview citations increased from 3 to 19 over eight weeks." The specificity of first-hand experience at a defined scale is not replicable without having done the work.

Expert synthesis that requires expertise to produce. Analysis that connects concepts in ways that require genuine domain mastery. Not summarizing what others have said about the relationship between traditional SEO and GEO — but analyzing the structural mechanisms through which Google's March 2026 core update simultaneously changed ranking signals and AI citation patterns, drawing on first-hand observation of client data across both surfaces. The synthesis is non-commodity because producing it requires both the expertise and the access that most content producers do not have.

Documented failure modes and counterintuitive findings. The information that practitioners most need and most rarely find: what does not work, under what conditions, and why. The GEO playbook article that documents three structured data implementations that reduced AI citation rates — and explains the structural reason they backfired — is more valuable to a practitioner than ten articles documenting implementations that worked. Counterintuitive findings and honest failure documentation are inherently non-commodity because they require honesty and specificity that commodity content production systematically avoids.

Named, accountable human perspective. Opinions, assessments, and predictions attributed to a named expert with a documented track record in the field — and that represent a genuine, defensible position, not a consensus summary. Danny Sullivan's "commodity vs non-commodity" framing is itself an example: it represents a specific conceptual distinction, attributed to a named individual with institutional authority in the domain, that structures a conversation in a way that generic content does not.

Why AI systems apply the same distinction

The commodity/non-commodity framework is not unique to Google's ranking systems. It reflects a deeper principle about how information systems evaluate source quality.

AI language models — particularly the retrieval-augmented systems that power real-time search — face the same problem as Google's ranking algorithm: the web contains vast quantities of content that says the same thing in slightly different ways. Selecting among near-identical sources requires some principle of differentiation.

The principle that both Google's systems and LLMs appear to apply is information value: the degree to which a source adds information that is not already in the synthesis. A source that says what five other retrieved sources already say adds marginal information value. A source that adds a unique data point, a distinctive perspective, or a specific documented outcome adds high information value — and is more likely to be included in the synthesis.

This is the mechanism behind the "Information Gain" ranking signal that Google reinforced in the March 2026 core update. And it is the mechanism behind the citation patterns that AI visibility researchers have documented: sources cited in AI answers tend to be the ones that contain specific, verifiable, original information — not the ones that best summarize what is already known.

The practical conclusion is uncomfortable but clear: generic, well-optimized content performs worse in AI-mediated discovery than specific, expert-grounded content — even when the generic content is better structured, fresher, and more thoroughly formatted.

Format signals matter. They matter less than substance.

The content audit framework: applying the distinction to your existing corpus

Most brands looking at this framework face the same challenge: a large existing content corpus, most of which was produced under a different optimization model, with varying levels of commodity and non-commodity content distributed unpredictably across it.

The audit is the first step. Here is how to approach it.

Step 1 — Apply the disappearance test to every high-priority page. For each of your top 50 organic pages, ask: if this page disappeared, would anyone lose access to information they cannot find elsewhere? Document the answer honestly. Pages that fail the test are commodity pages — candidates for either upgrade or consolidation.

Step 2 — Identify what you know that others do not. What data do you have that is not publicly available? What client outcomes have you documented with specific numbers? What implementations have you run that produced counterintuitive results? What expert perspectives do your named authors hold that are not consensus positions? This inventory is your non-commodity asset base.

Step 3 — Map asset base to content opportunities. For each piece of proprietary knowledge, identify which commodity pages in your corpus it could upgrade. Original client data can upgrade a generic case study. A documented implementation outcome can upgrade a generic best practices guide. A named expert's distinctive position can upgrade a consensus summary. The non-commodity upgrade does not require starting from scratch — it requires injecting original information into existing structure.

Step 4 — Build production systems for ongoing non-commodity content. The most common failure mode: a brand produces one piece of genuinely non-commodity content (an original research report, a detailed case study), generates strong AI citation performance, and then returns to producing commodity content because non-commodity production requires more effort and organizational infrastructure.

Non-commodity content at scale requires systems: regular data collection from client implementations, a structured process for documenting and publishing first-hand outcomes, named expert contributors with defined publication cadences, and editorial standards that explicitly reject commodity content regardless of its GEO formatting quality.

The counterintuitive implication: less content, more citations

The commodity/non-commodity framework has a counterintuitive operational implication that most content teams resist.

Producing fewer, better pieces of genuinely non-commodity content generates more AI citations than producing more, well-optimized pieces of commodity content.

This runs against the volume logic that has governed content marketing for fifteen years. Publish more, cover more queries, build more topical authority through breadth. The GEO era inverts this logic for citation performance: depth and originality in a focused area outperforms breadth and genericness across a wide area.

A brand that publishes twelve genuinely non-commodity pieces per year — each containing original data, documented outcomes, or distinctive expert perspective — will accumulate more AI citations than a brand that publishes fifty commodity pieces per year, regardless of how well those fifty pieces are structured, formatted, and optimized.

The operational implication is a reallocation: fewer total pieces, higher investment per piece, more rigorous editorial standards, and explicit organizational commitment to the data collection, expert contribution, and honest documentation that non-commodity production requires.

That is a harder organizational change to make than adding FAQ schema to existing pages. It is also the change that makes the difference between a content strategy that generates AI citations and one that generates AI retrieval without attribution.

Conclusion

Danny Sullivan's commodity/non-commodity distinction is not a new content quality framework. It is a precise articulation of what the best content practitioners have always known: content that contains nothing original is content that will eventually be outcompeted by content that does.

What is new is the mechanism of outcompetition. In the traditional search era, commodity content could survive through technical optimization, link building, and publishing velocity. These tactics generated rankings that persisted long enough to justify the investment.

In the AI-mediated discovery era, the survival window for commodity content is shorter. AI systems make source selection decisions in milliseconds, applying information value assessments that have no tolerance for interchangeable content. The commodity article that ranked for three years on a well-optimized keyword will not be cited in AI-generated answers — because the AI has no reason to cite it when five other sources say the same thing.

The brands that understand this early enough to reorient their content production around genuine originality — original data, first-hand experience, named expertise, documented failure modes — are building citation assets that compound. The brands that respond by better formatting their existing commodity content are optimizing the wrong thing.