What Does the Machine Think of Us? Synthetic Perception and the Places It Misreads

 The session that stayed with me longest at the Destinations International Global Leaders Forum in Geneva was not about a new tool. It was a perception workshop led by José Filipe Torres of Bloom Consulting, and what made it memorable was a simple exercise. Participants rated a handful of cities on a set of attributes through a QR survey, then received a report comparing their own views with how AI systems describe the same places.

Destination organisations have spent decades managing how places are perceived by people. Torres argued that there is now a second audience, which he called synthetic perception: the way AI systems describe and evaluate a place based on the information available to them. Most of the AI conversation in our industry so far has been about discovery, meaning who gets found. This is a different question, and possibly a more uncomfortable one. It is about what gets believed.

Reputation has become partly a data problem

AI systems describe places from what has been written about them, and the session pointed to sources such as Wikipedia and Reddit as unusually influential. A destination with thin, outdated or stereotyped coverage will be described accordingly, and the description will sound fluent and confident. Dallas was the example used in the room, where images from a 1980s television series still sit comfortably alongside the contemporary city. Places change faster than their reputations, and a machine summary can freeze that lag in place.

Today's errors may become tomorrow's inheritance

The more unsettling idea was what Torres called hypersynthetic perception, where AI systems begin to learn from and influence one another. A single inaccurate description then stops being a mistake that can be corrected at the source. It becomes background knowledge, repeated across systems that never consult the place itself. The practical implication is timing. Monitoring how AI describes a destination is easier and cheaper before those narratives settle than after.

Peripheral places have the most to gain and the most to lose

For a destination far from the usual shortlists, this cuts both ways. Torres suggested that smaller destinations may be able to compete through content strategy in a way that was harder when visibility depended on promotional budgets. At the same time, a place with limited coverage is exactly the kind of place a model fills in with generalities. In Northern Finland, the question I took home is whether contemporary culture, technology, accessibility and year-round life are visible to these systems, or whether older images of darkness, cold and remoteness still do most of the describing. I do not yet know the answer, which is precisely why it seems worth measuring.

The most useful gap may be the one closest to home

One finding from the session deserves more attention than it received. Local professionals and residents tend to be more critical of their own destinations than outside audiences are, often because they see the problems up close. A perception audit that compares insiders, outsiders and machines therefore produces three views rather than two, and the differences between them are the information. Residents belong in that exercise, because they help deliver the experience that any perception is ultimately measured against.

What this means for behavioural targeting

This connects to something I keep returning to in my doctoral work on AI and peripheral tourism. Behavioural targeting rests on the assumption that behaviour is the signal: what travellers search for, click on and book. If an AI layer forms an impression before any behaviour becomes visible, then part of that signal is already a product of the machine's prior description of the place. Intent is shaped upstream of the point where we can measure it. Targeting a traveller whose shortlist was narrowed by an inaccurate summary means optimising against a distorted starting point. It also raises governance questions that are easy to postpone, such as who is accountable for how a place is represented, and how much of that representation we are comfortable leaving to platforms.

One line from the session, which I would treat as a synthesis rather than a quotation, framed branding as perception correction rather than invention. The task is to close the distance between what a place is and what is said about it. That distance is now also shaped by systems that will increasingly speak to travellers on our behalf. For destination organisations, reputation work is starting to look less like communication and more like stewardship of public information, and that is a role worth preparing for deliberately.