Scientific Foundation

AURAS-O currently represents a conceptual synthesis rather than a fully validated measurement model. It does not originate from a completed empirical validation study. It brings together mechanisms and findings from adjacent research traditions and translates them into management questions relevant to cultural organisations.

The framework provides a structure for examining conditions that may influence how museums, theatres, cultural destinations and other cultural organisations are represented and recommended by AI systems. It does not claim to establish the causes of a particular recommendation or to replace SEO or GEO.

Its six dimensions are working conceptual distinctions. Future empirical research may refine, challenge or restructure individual dimensions and their relationships. Making this status explicit allows readers to distinguish the framework’s proposed structure from the evidence on which it draws.

AURAS-O draws together perspectives on generative search, information retrieval, knowledge representation, reputation, recommendation, digital discourse and agentic AI. The following mapping indicates principal connections, rather than six independent literature streams.

Information Retrieval & Generative Search

→ Accessibility

Knowledge Representation & Entity Understanding

→ Understanding

Networks, Associations & Knowledge Ecosystems

→ Association

Reputation, Credibility & Signaling

→ Reputation

Digital Discourse & Dynamic Information Environments

→ Social Dynamics

AI Agents, Identification & Interaction

→ Observability

Research on generative search examines the retrieval and use of information within generated answers. The original GEO study evaluates interventions within a defined experimental setting; its results should not be read as a general guarantee of discoverability or recommendation. ^[1]

Knowledge-graph research provides a basis for considering entities, meaning and relationships together. This informs both Understanding and Association: an artist, collection and exhibition gain context through their connections. It does not establish that every generative system represents those connections in the same way. ^[2]

Research on the visibility of AI agents addresses identification, monitoring and records of agent activity. AURAS-O translates these questions into a cultural organisation’s readiness for interaction; that application remains a conceptual proposition, not a validated sector-specific measure. ^[3]

The connections to reputation, credibility, signaling and digital discourse remain broader areas of theoretical integration. They overlap with questions of source selection and recommendation, but should not be understood as independently validated explanations of AI behaviour. A critical survey of GEO likewise distinguishes multiple stages and uneven evidence across the field. ^[4]

1. Existing research
2. Mechanisms and concepts
3. Conceptual synthesis
4. Management questions
5. Six AURAS-O dimensions
6. Empirical validation

The six dimensions were developed as a management-oriented synthesis of recurring mechanisms identified across adjacent research fields. They are intended to make a fragmented body of evidence operationally intelligible for cultural organisations.

The progression describes the development logic, not a sequence of completed studies. Empirical validation remains work to be undertaken. Six dimensions provide a current working structure; neither their number nor their boundaries have been established as exhaustive or statistically independent.

The current framework should therefore be understood as a structured conceptual proposition whose boundaries and relationships remain open to empirical testing.

AURAS-O draws on sources with different levels of evidentiary weight. The website therefore distinguishes between:

Peer-reviewed research

Research assessed through a scholarly review process. The strength of a claim still depends on the study design, evidence and scope; peer review alone does not establish causality.

Scientific preprints and working papers

Research shared before, or outside, completed peer review. Findings may be useful and timely while remaining provisional.

Institutional research

Reports whose methods, data and institutional context need to be assessed individually. Institutional authorship is not itself a quality guarantee.

Industry and vendor research

Observations from operational systems that may offer useful insight, with attention to sampling, reproducibility and commercial interests.

These categories should not be treated as equivalent evidence, nor as a substitute for examining methods. Industry research can provide valuable observations from rapidly changing AI systems, but should not be interpreted as equivalent to peer-reviewed causal evidence.

Martinez’s survey highlights the importance of distinguishing experimental effects on already-retrieved content from broader claims about discoverability and downstream outcomes. AURAS-O therefore keeps conceptual propositions separate from tested relationships. ^[4]

An AI system may influence awareness, understanding or choice without generating a website visit.

Conversely, frequent representation within AI-generated responses does not necessarily produce additional referral traffic.

Traditional web analytics therefore capture only part of AI-mediated visibility. Representation, citation, referral and subsequent visitor decisions are different outcomes; observing one does not establish another. The distinction also follows the separation of visibility and downstream outcomes in Martinez’s survey. ^[4]

AI Visibility should therefore be examined independently from traffic, while downstream outcomes remain an important separate research question.

An observed answer is specific to the conditions under which it was generated. Outputs may vary according to:

• AI system and model
• Product or search mode
• Prompt formulation
• Language
• Location and context
• Personalisation
• Measurement date and time
• Stochastic variation

These factors need to be documented or considered in study design; their importance will not be identical across systems. Repeated observations can reveal variation, but cannot remove every source of uncertainty. Martinez discusses the need for repeated measurements and controlled comparisons. ^[4]

AI Visibility should therefore be treated as dynamic and probabilistic rather than as a permanently stable ranking.

A response cannot, on its own, reveal why a cultural organisation was included or omitted. Nor does a citation necessarily establish accurate interpretation, institutional credibility or influence on a visitor’s choice. AURAS-O currently offers no universally stable score; future empirical measures must specify what they measure and under which conditions.

The framework is intended to be empirically examined and refined. This includes testing whether the dimensions can be distinguished in practice, how they interact, and which observations meaningfully address the management questions they raise.

The current study of AI Visibility among cultural organisations in Austria and Germany is one research programme informing this process. It examines representation and recommendation across a structured set of prompts. It should not be understood as a promise of complete validation of all six dimensions.

Current research will inform the further development and operationalisation of the framework.

Explore Insights & Cases →