AITourPilot Try Henry

What we have built, and how we checked it.

Guides for more than thirty museums and heritage sites – art museums, an archaeological city, a seventeenth-century warship, a codebreaking estate, a working farmstead – and the discipline behind every sentence Henry speaks.

On the platform

The scale it already runs at.

30+

Museums and heritage sites

10,000+

Works, each with its own record

32,000+

Sources behind them

2,400+

People with a biography

How it is checked

Every claim is weighed before a word is written.

Facts are gathered, then arbitrated against their sources one at a time, and an editorial pass goes looking for what slipped through. For Forum Romanum all 760 historical claims were arbitrated individually, and then a red team was set on the result to attack it.

It catches exactly what an unchecked model gets wrong: an invented person, a date pulled from the wrong foreign source, a name that drifted. None of it reaches a visitor. The four layers, in order.

  • It corrects the internet

    On one site the check overturned an attribution that is widely repeated online: the excavation had been led by someone other than the famous name usually credited with it.

  • It says when a thing is a copy

    At another site the first thing the guide says, before anything else, is that the object in front of you is a full-size copy and the original is in the national museum.

  • It gives dates as ranges

    Where the evidence spans a range, he says the range. Ancient casualty figures are attributed as ancient accounts, not repeated as modern fact.

What real visitors changed

Visitors interrupt a guide who talks too long.

In field testing, a long opening answer was cut short by the visitor in 56% of cases. So Henry’s first answer is now 60 to 80 words, by rule, in every guide we build. It is the kind of thing that only shows up when real people are standing in a real room, and it is now enforced by a test rather than left to a writer’s judgement.

The standard

MEUX is the standard for a meaningful museum visit. We built against it.

Researchers at Warwick and Oxford published it in 2023: a model for museum exhibition user experience. Not a checklist of features, but a description of what actually makes a visit meaningful to the person having it – and a hard standard to clear.

We built against MEUX, and we meet all of it but one — the shared family visit, which is on the roadmap.

We name the one because it is what makes the rest worth believing. Anyone can claim to meet a standard nobody can check. This one is open access: the paper is linked below, and a curator can read it and hold us to it line by line.

Why it is worth reading anyway

The model was validated across 28 interviews, 272 staff surveys and 341 visitor surveys – and one of its findings stands on its own: what museum staff believed made a visit meaningful and what visitors themselves said diverged significantly.

King, E., Smith, M. P., Wilson, P. F., Stott, J., & Williams, M. A. (2023). Creating Meaningful Museums: A Model for Museum Exhibition User Experience. Visitor Studies, 26:1, 59–81. Open access, CC BY 4.0, funded by the AHRC.
Read the paper

See what AITourPilot could do for your museum.

Five minutes, and you will know.