What is information stratigraphy and the QISM model?

Research on AI Visibility mostly looks at the outcome: how often a company is mentioned, how sources are cited and how answers differ between models. The authors of the article ask a different question: what does the public information that AI can build an answer from actually look like? Below: what the model proposes, what it consists of and what it does not yet prove.

What is information stratigraphy?

Information stratigraphy is a way of describing the public information about a company, an organisation or a person as a system of distinct institutional levels. The name comes from geology, where stratigraphy describes how rock layers are arranged. The authors stress that the analogy does not mean the laws of geology apply to information; its purpose is to give the researcher a scheme for describing a complex object systematically.

The authors explain the need for such a scheme with examples. One organisation is documented in detail in its own materials but is almost absent from independent publications. Another has many press mentions, all of which go back to a single press release. A third has contradictory information in official registers and in derivative sources. According to the authors, a simple count of documents says very little about how such an environment is structured.

Where was the article on QISM published?

The article "Information Stratigraphy: A Five-Level Model of Sources Behind AI Visibility" was published on 20 October 2026 in the Newsletter on the Results of Scholarly Work in Sociology, Criminology, Philosophy and Political Science, volume 7, issue 2. The authors are Serhii Pontus and Leonid Kachuliak; the article gives QuorumLab, Finland as the organisation of both. The article is in English and openly available under the CC BY 4.0 licence.

What are the five levels of the QISM model?

The QISM model divides the public information about a company into five levels according to the procedure a document has gone through. Each level has sublayers; the table gives examples from the article.

The table shows that a level is not a rating of reliability. The authors say plainly that a company's website belongs to the open web but can be the primary source for its contact details and product specifications. The level shows where a document comes from, not whether it is true.

By what rule is a document assigned to a level?

The core rule of QISM is the institutional filter principle: a document is assigned to the level whose procedure of selection, registration, expert review or editorial approval it is subject to. What counts is not the status of the author or the reputation of the platform but the check this particular document has passed. For example, an independent journalist's article that has gone through editorial review belongs to the media level, while a paid press release on the same website without independent selection remains affiliated material.

The second key idea of the model is to keep three objects apart: the document, the claim and the producer of the information. A claim is a single statement about the company, such as its founding date or the name of its chief executive. One document can contain many claims, one claim can be repeated in many documents, and one author can publish documents at different levels. Besides the level, a document is marked with additional characteristics: affiliation, provenance (original or paraphrase), currency and mode of production, whether by a human, by AI or both.

What is the AI Access Pathway?

The AI Access Pathway is a separate dimension of the model that describes the technical and contractual channels through which an AI developer can obtain material from a platform. The authors name ordinary public access, search mechanisms, specialised APIs and data-supply agreements as such channels. It is not a sixth level but an additional characteristic: two documents at the same level can be accessible to AI in different ways.

The authors warn against conflating four things: the existence of an agreement, the technical possibility of obtaining data, the actual use of a document and its influence on an AI answer. One of them does not prove the others.

Why does QISM matter for AI Visibility?

QISM gives research an order in which the public information about a company is described first and only then compared with AI answers. The authors propose a six-step sequence:

1. describe the company's information environment level by level;
2. measure AI Visibility: how specific AI systems represent the company;
3. compare and identify correspondences and discrepancies;
4. formulate hypotheses about the possible role of levels, sources and access channels;
5. test possible causes with different methods;
6. repeat the measurement and refine the model.

This order helps to tell independent confirmation of a fact apart from repeated copying of the same report. The authors stress separately that if a new publication coincides in time with a change in an AI answer, that is not enough to claim the publication caused the change.

What does the QISM model not yet prove?

The article on QISM is conceptual and methodological: it contains no empirical data of its own, and the authors name its limitations directly. The five-level structure has not yet been tested for reproducibility or completeness. The model is limited to public information and does not examine closed training data, private data or the internal mechanisms of AI systems. Information about contractual access channels does not, on its own, show that specific documents have actually been used.

In the section on conflicts of interest, the authors state that QuorumLab provides commercial services in the field of AI Visibility and that QISM was developed as part of QuorumLab's research activity. The article does not evaluate the effectiveness of these services. The work received no external funding, and the authors disclose in the article that AI tools were used in preparing the manuscript.

So what is QISM?

QISM is a framework that divides the public information about a company or a person into five levels according to the check a document has passed, and separately accounts for how AI can gain access to it. It was proposed by QuorumLab researchers Serhii Pontus and Leonid Kachuliak in October 2026 as a basis for further research. The full text of the publication is openly available via DOI 10.61439/RSW39510. What QuorumLab is and how Soft Dev Oy is connected to it is explained in the article "What is QuorumLab and how is Soft Dev Oy connected to it?".