Here is an awkward truth that architects rarely say out loud: most clients cannot read drawings. Not properly, anyway. A plan is an abstraction. A section is a stranger one. Even elevations, which look superficially like pictures of buildings, flatten away the very things — depth, light, material, atmosphere — that make a building feel like anything at all.

For centuries the profession worked around this. Architects built physical models, sometimes at enormous expense. Brunelleschi is said to have built a model of his dome for Florence Cathedral partly to convince a sceptical committee, and partly, the story goes, to hide his method from rivals. The model was a persuasion device as much as a design tool. That dual purpose — design instrument on one hand, sales instrument on the other — has followed visualisation ever since, and it explains a lot of the tension in how the tools get used today.

The client's inability to see what they were buying had real consequences. Buildings got signed off on trust and delivered as surprises. Sometimes pleasant ones. Often not. Anyone who has worked in practice long enough has a story about the client who approved every drawing, then walked into the finished space and asked why the ceiling felt so low.

The story of architectural visualisation isn't a smooth curve. It moves in jumps, and each jump changed who got to participate in the conversation.

1. Perspective drawing (15th century). Alberti and Brunelleschi's codification of linear perspective gave architects a way to show depth on a flat sheet. For the first time, a drawing could approximate what an eye would see. It remained a skilled craft — good perspectivists were expensive and rare well into the 20th century.
2. Photorealistic hand rendering (19th–20th century). Watercolour perspectives, then airbrush and gouache renderings, became the currency of competitions and client presentations. Firms employed dedicated renderers whose entire job was making the unbuilt look inevitable.
3. CAD and early 3D (1980s–1990s). Computers took over drafting first, then modelling. Early renders were crude — flat lighting, plastic-looking materials — but they were cheap to revise, and that changed everything. A view that once took a week to repaint could be re-rendered overnight.
4. Real-time and immersive visualisation (2010s onward). Game engines, VR headsets, and GPU rendering collapsed the feedback loop from days to seconds. A client can now stand inside a building that doesn't exist, turn their head, and say "the window should be bigger" — and watch it get bigger.

Each leap did the same thing in a different way: it moved the moment of understanding earlier in the process, and it widened the circle of people who could genuinely participate in design decisions.

It's tempting to file visualisation under marketing — pretty pictures that help win work. And yes, that's part of it. But the deeper effect is on decision-making, and decisions are where buildings succeed or fail.

Consider what changes when a client can actually see:

• Changes happen when they're cheap. Moving a wall in a model costs minutes. Moving it on site costs weeks and a variation order. The earlier a problem is spotted, the less it costs to fix — this is the single most repeated maxim in construction, and visualisation is the most direct tool we have for acting on it.
• Approval means something. When a planning committee or a board signs off on a photorealistic view or a VR walkthrough, they've approved something they genuinely understood. Disputes later — "that's not what we agreed" — become rarer and easier to resolve.
• Non-experts get a real vote. Community consultation used to mean pinning plans to a church hall wall and hoping. Now residents can be walked down a proposed street at eye level. Whether that makes consultation more honest or just more persuasive is a fair question (more on that below), but it undeniably makes it more accessible.
• The design team catches its own mistakes. This one gets less attention. Architects fool themselves too. A stair that works in plan can feel oppressive in a walkthrough. Rendering a space forces decisions — about materials, junctions, light fittings — that plans let you defer indefinitely.

There's also a quieter cultural effect inside practices. Younger architects who grew up with game engines think in three dimensions from the first sketch. The old workflow — design in plan, check in section, render at the end — is inverting. In some studios the model is the design, and the drawings are extracted from it almost as an afterthought, a legal formality for the contractor.

This brings us to the uncomfortable part. Visualisation has always been part persuasion, and modern tools are extraordinarily persuasive.

Every experienced planner has seen it: the render where the street trees are twenty years more mature than anything that will be planted, where the sky is permanently golden, where the neighbouring building that would cast the whole plaza into shadow has been quietly de-emphasised. None of these choices is a lie, exactly. Together they add up to one.

Some jurisdictions have started pushing back. Verified view methodologies — where the render must be built on surveyed camera positions and validated photography — are now standard for tall-building applications in cities like London. The idea is simple: if an image is going to be treated as evidence, it should meet evidentiary standards.

There's a parallel worry heading the other way. As AI image generation gets folded into the pipeline, it's becoming trivially easy to produce a gorgeous "render" of a building that no one has actually designed — no structure, no fire strategy, no buildable detail behind the pixels. The image precedes the architecture entirely. Clients see these images, fall in love, and then a design team has to reverse-engineer a real building out of a hallucination. Sometimes that's a productive provocation. Sometimes it's a slow-motion disappointment with a signed contract attached.

The discipline's old rule still holds, and it's worth stating plainly: a visualisation is a claim about the future, and claims can be honest or dishonest regardless of how they were made.

What good looks like: a working checklist

For clients and developers commissioning visualisation — or architects deciding how much of the fee to spend on it — a few practical tests separate useful imagery from expensive decoration:

• Does the image show a decision that still needs to be made, or one already locked in? Money spent visualising settled questions is marketing budget, not design budget. Both are legitimate; just know which one you're spending.
• Can the viewpoint be checked? If a render claims to show the view from a neighbour's window or a station platform, someone should be able to stand there and verify it.
• Are the conditions typical or exceptional? Ask to see the scheme in flat grey light, in winter, at dusk. A building that only works at golden hour doesn't work.
• Is the model shared or siloed? A visualisation built from the live design model stays true as the design changes. One built as a separate artefact starts drifting from reality the day it's finished.
• Who is the audience, really? An image made to win a committee vote and an image made to help a design team choose between two staircases should not look the same, and shouldn't cost the same either.

None of this is complicated. Most of it is just the willingness to treat images as working documents rather than trophies.