Non-commodity content: what it is and how to create it to avoid being replaced by AI Overview
AI Overview absorbs generic content without driving traffic to your site. The only antidote is non-commodity content: original, verifiable, and not replicable by AI. Here is how to recognize it and produce it.
If an AI could write it for you, the AI Overview will summarise it for you
GEO Displacement hits almost only one kind of content: commodity content. This guide explains what makes content non-commodity, how to spot it with a three-question test, and the four ways to produce it.
What is GEO Displacement, and why does it hit commodity content?
Google's AI Overview answers the user's query before they even click on a link. If your article contains generic information, the AI synthesizes it and displays it directly in the SERP: zero clicks, zero traffic. This is GEO Displacement — and it almost exclusively affects one type of content: commodity content.
Google's official guide of May 15, 2026 introduced a precise term to indicate the antidote: non-commodity content. Original, verifiable, not replicable by artificial intelligence. This guide explains what that means in concrete terms and how to produce it.
In short
- Content is commodity when anyone, including an AI, could write it the same way: the AI Overview absorbs it and your site is used as a source without being visited.
- Content is non-commodity when it holds information that exists only in your experience, your data or your expert point of view.
- The test is three questions: could an AI write it better than you? does it contain data or interpretations found nowhere else? would an expert find something they cannot find elsewhere?
- Four ways to produce it: proprietary data, verifiable direct experience, expert interpretation, documented case study.
- It does not guarantee a citation, but it is the necessary condition to become a source the AI Overview can select.
Commodity or non-commodity: what is the difference that decides your traffic?
The distinction is not about writing quality but about the replicability of the information: if anyone — including an AI — could write the same content the same way, the AI Overview synthesises it and your site is used as a source without being visited.
Commodity content
Anyone, including an AI, could write it the same way. Typical characteristics:
- Definitions taken from public sources (Wikipedia, Google's guide, official documentation)
- Lists of generic tips valid for any industry or company
- Summaries of other people's studies without original critical elaboration
- Step-by-step tutorials on standard operations without real variables
The AI Overview absorbs them because it can synthesise them with the same or higher quality.
Non-commodity content
It contains information that exists only in your direct experience, your data, or your expert point of view.
- AI cannot generate it: it wasn't there
- Google selects it as a citable source for precisely this reason
- It survives the summary: whoever wants the data has to come to you
How do you tell if content is commodity? The three-question test
Before publishing a piece of content, ask yourself these three questions in order.
Could an AI with access to all public sources write this article better than you?
Yes = commodity
If the answer is yes, the content is built on information that is already available and summarisable: the AI Overview will do it for you.
Does the content contain data, experiences or interpretations that exist nowhere else online?
No = commodity
This is the verifiable-originality test: a number you collected, a project you ran, a reading that requires your position in the industry.
Would an expert reader in the field find something they could not find on any other site?
No = commodity
The strictest test, and the one that matters for E-E-A-T: the expert reader is the best proxy for the source selector.
If you answered yes, commodity, no to all three: the content is at risk of GEO Displacement and is not a priority to publish as it is.
What are the four types of non-commodity content?
Four families, from the easiest to produce to the hardest to replicate. One per piece of content is enough.
Related: what AraLith COS is.
1. Proprietary and original data
Any number you have collected yourself, in your business, is not replicable. Surveys on real customers, project metrics, internal A/B tests, analysis of non-public datasets. You don't need a research team: even 20 responses to a questionnaire submitted to your customers on a specific industry topic are enough.
Practical example: an SEO agency publishes the results of a test conducted on 15 Italian websites optimised for AI Overview in the first quarter of 2026, with the citation rate before and after the intervention. No competitor has that dataset.
2. Verifiable direct experience
Saying "in my experience" is not enough: the experience needs to be made verifiable. Project names (where possible), dates, numerical results, screenshots, mistakes made. The key adjective is verifiable: Google evaluates E-E-A-T also on the ability to check the claims.
Practical example: the medichub.it case cited in Google's AI Overviews — a new domain, zero link building, cited as an authoritative source — is non-commodity content. Those who lived it can document it; anyone else can only cite it. You can explore this further in the AI Overview optimisation guide.
3. Non-replicable expert interpretation
An AI can synthesise what has already been written. It cannot produce an interpretation that requires positioning in the industry, relationships with specific stakeholders, or a verifiable professional history. Reasoned forecasts, critiques of widespread practices, comparative analyses with a vertical perspective: these are contents that an expert with real credentials can sign off on and an AI cannot fake.
4. Documented case study
The most effective format and the hardest to replicate. A real project with: initial objective, controlled variables, actions taken, measured results, errors and corrections, operational conclusions. Even a case study of a single anonymised client, if documented with rigour, is non-commodity by definition.
Frequently asked questions about non-commodity content
Non-commodity content is content that contains original, verifiable information that cannot be replicated by artificial intelligence: proprietary data, documented direct experience, expert interpretations, or real case studies. It is the type of content that Google selects as a source in AI Overviews because it cannot be synthesised without losing value.
The AI Overview synthesises answers directly in the SERP from indexed sources. If your content contains information also available elsewhere — definitions, lists of generic tips, summaries of public studies — the AI can answer the query without the user needing to visit your site. The result is GEO Displacement: stable impressions, zero clicks.
No. Real data from your business is enough: results from client projects (even anonymised), internal metrics, surveys with a few dozen responses, errors documented in a real case. The criterion is not the size of the sample but the originality and verifiability of the information.
It does not guarantee automatic citation, but it is the necessary condition to become a selectable source. The AI Overview selects sources that combine originality, thematic authority (E-E-A-T), AI-readable structure, and freshness. Non-commodity content satisfies the originality requirement; the other factors need to be addressed in parallel.
GEO (Generative Engine Optimization) is the set of techniques used to get cited by generative engines. Non-commodity content is its prerequisite: without verifiable originality, no structuring technique or schema markup will be sufficient. You can explore the full GEO strategy in the Google AI Overview guide and the AI Overview optimisation guide 2027.
Is your site losing traffic from AI Overview?
We analyze your content together and identify which ones are at risk of GEO Displacement and which have the potential to become sources cited by AI.
