Museum and gallery websites often treat the exhibition image as the page.
A full-screen installation shot loads first. Motion follows. The title appears over the image, while dates, artists, curators, ticket details, and supporting context sit behind tabs, sliders, or JavaScript.
The result can look exceptional in a design review and still perform poorly as a publishing system.
The risk has grown with AI Overviews. Google can now answer questions about an exhibition before the searcher opens an institution’s site. If the museum page is slow, thin on crawlable text, or hard to index, the summary may rely on press coverage, listings, or other secondary sources instead.
The visual-first web is not invisible because AI cannot process images. It becomes invisible when the facts surrounding those images are difficult to retrieve.
The Image Payload Became the Page
Images are already the heaviest part of the typical web page. The 2025 Web Almanac found that the median mobile homepage weighed 2.56 MB, with images accounting for 1,059 KB and JavaScript adding another 697 KB.
Museum sites can exceed that pattern quickly. They need high-resolution artwork, installation photography, collection thumbnails, video, interactive timelines, and custom transitions. The problem is not using those assets. It is loading too many before the visitor or search crawler reaches the exhibition information.
A 6 MB hero image does not communicate more authority than a well-sized responsive version. A carousel containing 20 installation views does not need to download every slide on first load. Text should not wait for a client-side animation to finish before appearing.
Performance also affects the people museums are trying to reach. Visitors may search on mobile while traveling, using public Wi-Fi, or standing outside the venue. The page should deliver the exhibition name, location, dates, price, and booking route before the full visual layer finishes loading.
The design question is not whether the site should feel like the institution. It is which assets have earned a place in the first screen.
AI Overviews Need Evidence They Can Retrieve
Google says pages must be indexed and eligible for a search snippet to appear in its generative AI features. Its current guidance also tells site owners to keep content crawlable, use clear technical structure, and follow JavaScript SEO practices.
That creates a weak point for exhibition pages built as visual shells.
An image may contain the artist’s name, exhibition title, and opening dates as part of the artwork. A human can read it. Search systems still work better when those facts also appear as text, headings, metadata, and structured page elements.
Google’s image guidance makes the relationship explicit. It uses nearby text, captions, titles, filenames, and alt text alongside computer vision to understand an image.
Traditional rankings do not guarantee an AI citation either. Ahrefs analyzed 863,000 search results and four million AI Overview URLs in 2026. Only 37.9% of cited URLs also appeared within the first 10 result blocks for the same query.
That means institutions cannot rely on a strong domain name or an old first-page position. Their exhibition pages need to supply direct answers to the related questions AI systems generate: who is included, what period the work covers, why the exhibition matters, where it is held, and when it closes.
Without those facts, the museum may provide the subject while another publisher receives the citation.
Museums Already Own the Best Source Material
The irony is that museums have exactly the material AI search is supposed to value.
They hold curator essays, provenance records, object metadata, installation photography, conservation research, artist biographies, lecture transcripts, and archival video. The weakness is often how that material is connected and published.
Culture24’s “Let’s Get Real” review said the cultural sector was still struggling to evaluate online success more than a decade after its first research program. Its earlier report had been downloaded nearly 24,000 times, showing how persistent the question remained.
Some institutions are moving in a stronger direction.
ArtDaily reported that the Museo Nacional del Prado launched Canal Prado in 2026 to gather and organize its growing library of art videos and lectures. It also covered the ALBERTINA’s first online catalogue raisonné of Gustav Klimt drawings, which turned collection material into a searchable research resource rather than a simple image gallery.
Those projects give each object, lecture, or topic a stable destination with its own context.
The same principle should apply to an exhibition. One page should contain the essential facts, curatorial argument, named works, credits, accessible image descriptions, related events, press material, and booking information. Video transcripts and object records should link back to it.
That structure serves visitors, researchers, journalists, and AI systems without writing a separate version for each one.
Keep the Visual Experience, Expose the Facts
The fix is not turning every museum site into a plain database.
It is separating the visual experience from the information architecture, then making both work together.
A release checklist should cover:
- Server-rendered exhibition titles, dates, venue details, and summaries
- Responsive AVIF or WebP images with width and height attributes
- Descriptive filenames, captions, and useful alt text
- Lazy loading below the first screen
- Indexable URLs for exhibitions, artists, objects, videos, and events
- Article, Event, VideoObject, and organization markup where accurate
- XML sitemaps that include current exhibition and image URLs
- Transcripts for lectures and curator videos
- Internal links between collection records and exhibition pages
- Search Console checks for indexing, Core Web Vitals, and AI-feature visibility
The build brief should ask studios that handle design-heavy, visual-first builds to protect the institution’s art direction without hiding its publishing layer. That means designers, developers, curators, accessibility specialists, and content teams agreeing on what must appear in the page source before motion and interaction are added.
Google does not require a special AI file or a rewritten “GEO” version of every page. It asks for useful content, clear structure, crawlable access, and strong image and video practices.
Museums do not lack distinctive content. They risk packaging it in a way that makes secondary coverage easier to retrieve than the institution itself.
The next generation of cultural websites should still make people want to see the work in person. They should also make the institution the easiest source to find, understand, and cite.
