Accessibility

Accessibility describes the conditions under which generative AI systems can reach, retrieve and process information about a cultural organisation. In the AURAS-O Framework, it concerns technical access and the availability of usable content: whether institutional information can enter the processes from which an answer is constructed.

The framework identifies crawlability, semantic HTML and structured data as practical indicators. These direct attention to how information is published, not simply whether a website exists. Content should be available as readable text, organised into meaningful sections and sufficiently self-contained to retain its meaning when retrieved separately from the surrounding page.

Accessibility is therefore an entry condition, not a guarantee of recommendation. A system may encounter an organisation through independent sources even when its own website is difficult to process. Reliable access to the organisation’s own account nevertheless creates the possibility of retrieving current, specific information. Whether that information is interpreted correctly belongs to the connected dimension of Understanding.

Current research suggests that access, content structure and source selection must be considered together. Information has to be available for retrieval, but inclusion in a generated answer also depends on how usable and relevant it is within the system’s wider information environment.

Aggarwal et al. examine changes to source content within a generative-engine benchmark. Their findings indicate that evidence-bearing additions, including citations, quotations and statistics, can affect visibility; effects vary by domain and method.^[1] Chen et al. extend the discussion to the sources selected by AI search systems, highlighting differences between owned, independent and social sources.^[2] Together, these studies suggest that accessible institutional pages participate in a broader source ecosystem rather than control the answer.

Serrano and Blum’s industry report, cited in the framework, associates structured, text-first pages with AI entry points.^[3] This is observational evidence, distinct from the benchmark experiments: it does not establish that a particular format will cause a cultural organisation to be recommended.

The shared implication is to make authoritative information retrievable and intelligible at the level of individual passages. These studies support examining accessibility as one condition among several, rather than reducing AI Visibility to keyword repetition or a universal technical formula.

For a museum, an exhibition page should make the exhibition title, dates, venue and curatorial context available as text. If those details appear only within a poster image, an AI system may not retrieve the same information that a visitor can see. Collection records can similarly expose an object’s name, creator and provenance in clearly identified fields.

A theatre or festival can publish each performance with its dates, location, participating artists and programme description on a stable page. This gives a retrieval system a coherent account without requiring it to reconstruct the programme from a visual calendar or a sequence of interactive filters.

For cultural heritage sites, accessible descriptions can distinguish the place, its historical significance and current visitor arrangements. Botanical gardens can connect readable collection information with named species, habitats and seasonal programmes. These are applications of the framework, not findings from the cited studies: the aim is to preserve cultural specificity when information is retrieved, while keeping the institution’s account available alongside external descriptions.