If virtual reality invites us into another environment, augmented reality adds a digital layer to the one we already inhabit. In this 2024 research series, I consider how AI might make that layer responsive to places, objects and people. The central question is visual: who decides what appears between us and the world, and what deserves our attention?

Translation could become part of the view: captions for a conversation, an explanation beside an unfamiliar sign, or guidance in a museum. In education, a model could appear next to the thing it explains. Historical scenes might be reconstructed on site, provided visitors can distinguish evidence from interpretation. The value would lie in making context accessible without replacing the place with a screen full of information.

Architects could compare a proposed building with its surroundings; city residents could inspect plans for a street or park. Environmental scenarios might show possible effects of heat, pollution or changing land use. These images would need to identify their assumptions. A convincing overlay can make a proposal feel inevitable even when it represents only one possible future.

In the workplace, instructions might appear beside equipment, while retail experiences could help people judge the size of furniture or explore a product before buying it. Healthcare and rehabilitation offer other possible uses, from explaining anatomy to guiding an exercise. In these settings, accuracy, validation and professional judgement matter more than visual spectacle. An overlay is not a guarantee of a safe decision.

Artists could create works tied to a particular location yet changing with a visitor's actions. Performances, stories and shared digital objects might become part of an ordinary room or public square. Social filters and virtual figures could expand how people present themselves. The designer's challenge would be to make these encounters meaningful while respecting those who share the physical space.

Outdoor guides could explain a landscape; training aids could visualise movement; city services could present directions or warnings. Audio descriptions, captions and alternative interactions might also improve access. Each case would need testing in real conditions: information that covers a hazard, arrives late or demands too much attention can undermine the help it promises.

The same layer could become an advertising surface or a personalised filter on public life. Recognition of people and places raises questions about privacy, consent and whose version of the world is being shown. Users should be able to understand, adjust and switch off the layer. Those around them may also be affected without having chosen to participate.

For me, the most interesting prospect is not filling every surface with digital content. It is making visual knowledge available where it becomes useful: a diagram, a memory, a translation, a possible future. AI could make this language more responsive. Whether it deepens our relationship with the world or distracts us from it will depend on what we choose to reveal—and what we leave unobstructed.