Where Museum Visitor Engagement Loses Depth

And why it matters most for international visitors

Museum visitor engagement loses depth, not breadth, at the points where a visitor wants to ask a question the panel does not answer, in a language the audio guide does not cover. International visitor numbers at flagship UK museums are at multi-year highs. The British Museum drew 6.5 million visits in 2024, its highest in a decade, with growth attributed to international visitors (ALVA 2024). Most institutions cover breadth competently through panels and audio guides. The gap is in depth. This post sets out the four moments where depth is lost, and how a multilingual AI character closes that gap alongside existing audio guides.

Museum visitor engagement is the depth and quality of interaction a visitor has with a museum’s collection. It is measured by dwell time, questions asked, content recalled, recommendation rate, and repeat-visit intent. For international visitors it is determined less by the collection itself than by the delivery layer: whether the museum can meet a visitor in their own language, at the moment they have a question, at the depth their curiosity deserves.

The four moments. Performit Live’s practitioner framework.

Note on this framework: the four moments below are Performit Live’s view from deploying AI characters in museum environments, most recently Private Forsythe at Waterloo Uncovered. It is a practitioner framework, not academic research. We share it because it has been useful in scoping every deployment we have run.

Museum visitor engagement tends to lose depth at four moments inside an international visitor journey:

1. The entrance. Orientation in a language the visitor reads only partially. Expectations are calibrated downward before the first exhibit.

2. The contextual exhibit. The artefact whose significance lives in a story the panel does not translate.

3. The unasked question. The moment of genuine curiosity that has nowhere to go.

4. The exit. Leaving with a shallower visit than intended, suppressing return visits and recommendations.

The collection is not the problem at any of these moments. The delivery layer is.

International visitors are arriving in record numbers, and they are the highest-value audience to engage well

The first thing to be clear about. International visitors are not declining.

ALVA’s most recent annual visitor figures report 157.2 million visits across 400 ALVA-member sites in 2024, with 193 of those sites reporting at least 38.7 million international visits in the year (ALVA 2024 visitor figures, alva.org.uk). Flagship institutions led the recovery. The British Museum drew 6.5 million visits, its highest figure since 2015, attributing the rise primarily to international visitors and major touring exhibitions (Art Newspaper, March 2025). The Natural History Museum took 6.3 million, up 11% year on year.

At national level, VisitBritain forecasts a record 43.4 million inbound visits to the UK in 2025, up 5% on 2024, with international visitor spending forecast at £33.7 billion (VisitBritain Inbound Tourism Forecast 2025).

So the question is not whether international visitors are coming. They are. The question is what experience they get when they arrive, and whether that experience reflects the value of the visit.

Two things make the international segment commercially important.

Per-visitor value. Inbound visitors to the UK spent an average of £806 per trip in 2024 (VisitBritain / ONS International Passenger Survey, 2024 IPS Total Inbound Update), materially above the domestic day-tripper average. They arrive with high intent and longer dwell expectations.

Advocacy reach. An international visitor who has a memorable visit recommends the institution across markets a museum’s paid marketing budget could never reach independently. A family from Osaka leaves with a story they tell colleagues in Japan. One who did not, simply does not tell anyone.

The commercial case for engaging international visitors well is straightforward. The question is whether the current delivery layer is built for it.

For the delivery-layer architecture, see Lumiere AI, our venue intelligence platform.

The four moments where depth is lost (in detail)

Moment 1: The entrance

The entrance frames the visit. Highlights are signposted, expectations are set, the visitor gets their first read of the site. For a domestic visitor who can read panels, ask a volunteer, and skim the literature, this works.

For an international visitor who does not share a language with staff, the entrance becomes a logistical exercise rather than an orientation experience. They collect a map they can partially read. They scroll an audio-guide menu. They make do, and their expectations are calibrated downward before they have seen a single exhibit.

Moment 2: The contextual exhibit

Not every exhibit is self-explanatory. The artefact on the plinth, the document on the wall. To unlock its significance the visitor needs the story behind the object: the period, the person, the event that makes it remarkable rather than simply old.

A domestic visitor has multiple paths to that context: panels, volunteer guides, digital interactives, audio guide tracks. For an international visitor who does not read English fluently, the panels become visual texture rather than information. The audio guide, if it covers their language, typically delivers the headline and skips the depth.

Moment 3: The unasked question

Every visitor, at some point, wants to ask a specific question. Not the one the label answers, but the one prompted by genuine curiosity. What happened next? Why this object and not another? What does this connect to in the gallery upstairs? This is the highest-value moment of a museum visit. The moment real engagement happens.

For international visitors, this moment is routinely lost. The question, in their language, meets a volunteer who does not share that language. The question goes unanswered. The engagement passes.

Moment 4: The exit

What visitors leave with, the story they carry, the depth of understanding, the emotional resonance, determines whether they return, whether they recommend, and whether the visit was worth the investment to get there.

International visitors who navigated the gaps above frequently leave with a shallower experience than intended. Not because the collection failed them. Because the delivery layer did.

Why audio guides cover breadth well. Depth is a different job.

Museums have not ignored the international visitor. The standard delivery layer, the audio guide, has served the sector well for decades and remains the most cost-effective way to deliver baseline content across a whole site without screens, WiFi, or staff.

Audio guides are not the problem.

Audio guides do one job well: breadth. They deliver consistent narration across an entire site, in a handful of curated languages, without hardware at the point of display. Industry-standard platforms (Acoustiguide, Orpheo, Antenna) technically support 20-plus languages, but the cost of recording, translating, and maintaining each track means most museums deploy 5 to 10. The British Museum’s current Audio app, for example, ships in 10 languages. Arabic, present in its predecessor Multimedia Guide (2009-2014), has since been dropped, even as Gulf inbound visits to the UK have grown. That is a sensible cost-benefit decision for production-line content. It is also a structural ceiling on language coverage.

Audio guides were never designed to cover depth. Three structural limits surface when depth matters most.

1. The language ceiling. The international visitor mix at a flagship UK museum now includes Arabic, Mandarin, Hindi, Japanese, Russian, Korean, Portuguese and dozens of others. The UK as a whole receives inbound visits from 200-plus source markets (VisitBritain inbound tourism statistics). A 10-language library, by design, leaves the long tail uncovered.

2. The script problem. Audio guides deliver what curators decided to say, to every visitor, regardless of what the visitor actually wants to know. The military historian and the seven-year-old asking what soldiers ate for breakfast both receive the same track. This is not a flaw in audio guides. It is the nature of pre-recorded media. A 2012 Open University study of audio-guide formats found visitors using dialogic guides produced roughly 35 on-topic conversational turns per painting, versus 20 with scripted monologue (Elliston & FitzGerald, 2012, CSEDU 2012).

3. The volunteer dependency. Human guides remain the gold standard for deep multilingual engagement, when available. UK museum volunteer hours remain 16% below pre-pandemic levels, with multi-site organisations reporting the largest capacity gaps (Annual Museum Survey 2024, Arts Council England / Museum Development Network). The Museums Association’s 2024-25 Members Survey records overreliance on volunteers alongside difficulty retaining them as staff capacity contracts. Availability is unpredictable, language coverage is limited to who is on shift, and quality varies with the individual.

The honest framing. Audio guides cover the breadth of a site efficiently. AI characters cover the depth at the points that matter most. They are different jobs. They run side by side without overlap. (We argue this in the AI Character vs Audio Guide.)

A side-by-side comparison of how each delivery layer performs lives in AI Character vs Audio Guide.

What an AI character adds at the moments depth matters (named proof)

The AI character is not a better audio guide. It is a different kind of visitor engagement, deployed at the points where depth, language, and responsiveness matter most.

At Waterloo Uncovered, Private Forsythe is a Coldstream Guard from 1815. Visitors approach him, ask questions the platform supports, and he answers. Not from a pre-recorded script. From a governed knowledge base built with historians and curators to reflect the authentic experience of a soldier at Waterloo. What separates a character that convinces from one a curator can trust is covered in what makes a historical AI character believable.

A Japanese tourist asks about equipment. A French student asks about Wellington’s strategy. A child asks what the solider ate for breakfast. Each gets the answer they wanted, at the depth their question deserved.

This is what changes the exit experience. Not the collection, that was always there. The delivery layer, at the points where depth matters most, finally matches the collection.

What it generates beyond the visit

Every question asked is logged. Every gap in the knowledge base, every question the character could not fully answer, is surfaced in a curator dashboard. Over time the knowledge base improves. The character gets smarter. The next visitor who asks a similar question gets a better answer.

The experience does not just serve visitors. It learns from them. That is what we mean by a venue intelligence platform. Museum visitor engagement that compounds with every visit.

How to know where an AI character fits at your site

The question for most museum directors is not whether the technology works. Private Forsythe is proof that it does. The question is where it fits in an existing delivery layer.

In our experience, three factors determine the answer. Footfall. Language mix. Existing infrastructure.

Footfall at high-engagement points: entrance, flagship exhibits, transition galleries.

Language mix of international visitors: how far the audio guide library falls short of the actual visitor profile.

Existing infrastructure: where audio guides, volunteers, and signage already do the job, and where they do not.

For some sites the right answer is AI characters at the entrance and the flagship exhibit, with audio guides continuing to cover site-wide breadth. For others it begins with a single character at the highest-traffic gallery. For some, audio guides remain the foundation and the AI character is a phased upgrade at specific points.

The two-minute Readiness Scorecard gives a personalised view of where a heritage site stands across the dimensions that matter before any deployment.

Want to see what this looks like at your site?

We built Private Forsythe for Waterloo Uncovered. In twenty minutes we can show you Private Forsythe running live, and walk through what a character would look like at your museum. The history it would tell, the languages it would cover, and what deployment actually involves.

Book a Discovery Call 

Explore the Heritage approach

Take the 2-minute AI Character Readiness Scorecard

FAQ

Q1. What is museum visitor engagement?

Museum visitor engagement is the depth and quality of interaction visitors have with a museum’s collection. It is measured by dwell time, questions asked, content recalled, and recommendation rate. For international visitors it is determined less by the collection itself than by the delivery layer: whether the museum can meet a visitor in their own language, at the moment they have a question, at the depth their curiosity deserves.

Q2. Are museums losing international visitors?

No. At most UK and European flagship sites international visitor numbers are at or near record highs. The British Museum drew 6.5 million visits in 2024, its highest figure since 2015 (ALVA 2024). The challenge is not loss of visitors but loss of depth of engagement once they arrive. Audio guides cover breadth efficiently. Depth at high-engagement points is where the gap is widest.

Q3. How does AI improve museum visitor engagement?

AI characters improve museum visitor engagement by responding to a visitor’s actual question in 40-plus languages in real time, allowing visitors to ask anything rather than receive a fixed script. Every question is logged, surfacing gaps in the knowledge base that curators close over time.

Q4. What is the difference between an AI character and an audio guide?

An audio guide delivers a fixed pre-recorded track in a limited set of languages and covers site-wide breadth efficiently. An AI character responds to a visitor’s specific question in 40-plus languages, at the depth the visitor wants, and logs every interaction. The two roles are complementary, not competing. Most museums run both. See AI Character vs Audio Guide for the full comparison.

Q5. Does an AI character replace audio guides or volunteers?

No. Audio guides remain the most cost-effective way to deliver breadth across an entire site. Volunteers remain the gold standard for in-person depth. AI characters extend reach into the languages, hours, and high-footfall points current delivery cannot cover at scale.

Q6. Who built Private Forsythe at Waterloo Uncovered?

Private Forsythe was built by Performit Live on the Lumiere AI platform, in partnership with Waterloo Uncovered. He is an AI Coldstream Guard from 1815 with a governed knowledge base developed with historians and curators.

Q7. How many languages can a Lumiere AI museum character speak?

A Lumiere AI museum character supports 40-plus languages in real-time conversation, including Arabic, Mandarin, Hindi, Japanese, Russian, Korean, Portuguese, and most European languages, with mid-conversation language switching.

Q8. What does it take to deploy an AI character at a museum?

A typical deployment is a 12-week process covering knowledge-base build with curators, character design, governance configuration, on-site install, and post-launch knowledge tuning. The Heritage page explains the steps in detail.