I introduced our exploration of artificial intelligence with three stories of potential futures, laying the groundwork for a broader exploration of how museums might navigate this swiftly changing landscape. Today Nik Honeysett, Chief Executive Officer of BPOC, takes us on the next leg of this journey: a tour of how AI is changing how people encounter information, and a framework for how museums might respond.

–Elizabeth Merritt, Vice President, Strategic Foresight and Founding Director, Center for the Future of Museums, American Alliance of Museums.

---

AI is reshaping the relationship between people and information. For museums, understanding the nature of that change matters because we cannot decide how to respond until we understand what is actually changing. The following four illusions: access, discoverability, certainty, and understanding, describe some of the ways AI is altering the relationship between cultural institutions, their knowledge, and the public.

The Illusion of Access

With Google AI Overview, when we ask for information about a museum, we are enticed into a conversation. “Do you want to know more?” It feels as though we are talking directly to the museum, but in reality, we are accessing an interpretation of it. Yesterday, the journey looked like this:

Question → Google → Museum Website → Visitor

Tomorrow it increasingly looks like:

Question → AI → Answer

The museum’s website hasn’t disappeared; it has become one source among potentially many, used to construct an answer and feed a conversation. This subtle change has profound implications. The conversation is increasingly no longer happening on our website, under our stewardship, or in the carefully crafted narratives we create. The conversation is happening somewhere else.

What can museums do?

First, we need to accept that bringing everyone back to the museum website may no longer be a realistic digital strategy. Instead, museums should think about how their knowledge travels beyond the website and how institutional authority accompanies it. That means publishing clear, well-structured, attributable information; making collections and interpretive content understandable to machines as well as humans; and paying closer attention to how third-party AI systems represent the institution. The objective is no longer simply to drive traffic to our websites. It is to ensure that when our knowledge travels elsewhere, its provenance, nuance, and institutional voice travel with it.

The Illusion of Discoverability

This leads to a second illusion. We assume that if information is published online, people will find it. That assumption was never entirely true, but AI is making it even less so. Discoverability is increasingly determined by algorithms, search rankings, recommendation engines, AI summaries, moderation systems, and “safety filters”. None of these systems necessarily intend to suppress information, yet they are all making choices. Every ranking system elevates some voices while reducing the visibility of others. Every recommendation system decides what is worth showing. Every AI summary decides what is worth mentioning. The consequence is subtle but important.

Censorship in the AI era does not necessarily require banning books, removing websites, or deleting content. Sometimes nothing is removed at all; it is simply undiscoverable. For museums, libraries, and archives, that raises uncomfortable questions. If your collection exists online but AI never mentions it… If your research is published but AI never cites it… If your interpretation is replaced by a simplified generative AI summary… Have you really been discovered?

To some extent, we have lived with this discoverability problem since the early days of the internet. We pursued SEO strategies and tactics trying to predict what terms the public would enter into Google’s search box to find us so that we would rank on the first page of results. We could “sponsor” words that we thought people would type in that would give us priority ranking. Now we have GEO (Generative Engine Optimization) and have the more difficult task of anticipating the kinds of questions and conversations people are likely to have about us, and publish authoritative content that can inform the answers AI systems construct. If we think we are the best cultural destination in the city, we need to talk about that. We also need structured, machine-readable content so bots can easily extract it.

If you think this means you should use generative AI text to create more content and help feed the beast, think again. There is compelling evidence to suggest that Google deprioritizes generative text, or rather the search algorithms “focus on rewarding people-first, high-quality content”. A recent study, updated for 2026 by @Rankability, found that 83% of top Google search results are not using AI-generated content. Bad news for leaders and managers who may be contemplating replacing human content writers with AI.

What can museums do?

The response should not be to produce more content faster. It should be to produce better content more deliberately. We have an advantage that generative systems cannot easily reproduce: original scholarship, collections knowledge, lived expertise, community relationships, provenance, primary sources, and distinctive institutional perspectives. Doubling down on high-quality human-generated content is therefore not nostalgia; it is a discoverability strategy. Museums should publish material that adds something genuinely new to the information ecosystem rather than contributing another statistically average version of what already exists. In an AI-saturated web, originality, authority, specificity, and provenance may become increasingly valuable signals. AI feeds on what already exists; museums should concentrate on creating what does not yet exist.

The Illusion of Certainty

A more dangerous illusion is certainty. Large language models are extraordinarily fluent – – we should expect this since they are constructing sentences based on statistical analysis of how we write. AI produces answers that are coherent, confident, and conversational, and we naturally associate confidence with competence. (If only they could add in a British accent.) But fluent does not mean factual. The problem is that AI is optimized to produce a confident answer. Whatever you ask it to do, no matter how bizarre or weird, it will have a go, and the result will be eminently plausible.

What can museums do?

We can respond by making uncertainty visible rather than treating it as a defect. Label what is contested. Explain where attribution is provisional. Show competing interpretations. Preserve gaps in the record rather than allowing technology to quietly fill them. Invite communities to explain where institutional knowledge is incomplete or disputed. In a world increasingly populated by fluent, confident answers, museums can distinguish themselves by being unusually honest about the limits of knowledge. “We do not know” is not institutional failure, but evidence of intellectual rigor.

The Illusion of Understanding

As information becomes abundant, judgment becomes scarce. Perhaps the greatest risk of AI is not misinformation; it is the illusion that understanding has occurred.

For decades, information science has described learning through the DIKW Pyramid:

Data → individual facts and observations

Information → organized facts with context

Knowledge → understanding developed through learning and experience

Wisdom → judgment about when and how to apply that knowledge

This model assumes that understanding is earned through progression. However, something very important is happening, and it’s not good. AI is compressing the journey from data to what appears to be knowledge. The learner receives a polished synthesis without necessarily experiencing the intellectual process that traditionally produced understanding. This is not simply faster learning; this is a different model of learning. Two people can arrive at the same answer. One has spent weeks reading, comparing evidence, questioning assumptions, and wrestling with ambiguity. The other has asked an AI assistant for a summary. Both may be able to explain the conclusion, but only one has experienced the journey, and that journey matters: it develops critical thinking, exposes uncertainty, reveals competing interpretations, builds intellectual humility, and, most importantly, develops judgment.

What can museums do?

This may be where we have the greatest opportunity. Museums have always operated at the upper levels of the DIKW Pyramid. They exist to cultivate judgment about knowledge: what is remembered, what is challenged, whose voices are included, and how society can learn from the past. They provide context, interpretation, multiple perspectives, and opportunities for reflection. They don’t merely answer questions; they help us ask better ones. Museums provide evidence that supports or challenges accepted narratives, and in an age when AI can generate explanations almost instantly, this role becomes even more valuable.

As AI makes information and synthesis abundant, museums can concentrate more deliberately on the things that remain scarce: context, judgment, accountability, lived experience, cultural responsibility, and wisdom. That means designing interpretation that exposes how conclusions were reached, encouraging visitors to encounter multiple perspectives, connecting knowledge to consequence, and creating opportunities for questioning and reflection rather than simply delivering answers. AI can help people know more things; museums can help people understand why those things matter.

A New Responsibility for Museums

These illusions describe the emerging reality of how people encounter information. But together they change the responsibility of museums. It is no longer enough to publish authoritative knowledge and assume that authority will survive intact wherever the information travels. Museums have a responsibility to ensure that society doesn’t simply gain faster access to knowledge but also continues to develop the wisdom needed to use it well. The institutions that have spent centuries helping society transform information into understanding may now have an equally important role in helping society transform AI-generated knowledge into human wisdom.

This is not a technology problem to solve. It is a question of institutional responsibility. My next essay for this series will be on how museums can live up to this responsibility through creating AI governance: the organizational policies, principles, decision-making processes, and procedures that guide how we select, use, evaluate, and oversee AI. This is the only mechanism that will give staff enough clarity to experiment confidently while establishing boundaries around privacy, intellectual property, transparency, cultural responsibility, sustainability, human oversight, and accountability.

The four illusions tell us what is changing. AI governance is how museums decide what they are going to do about it. That is where the next conversation begins.