What to Ask a Rendering Studio About Its Use of AI

Quick answer: Ask whether generated imagery appears in your final deliverables, which drawing set the model is built from, what happens to the set when a material changes in month three, who is accountable if an image shows something unbuildable, whether any post-processing is generative, whether the deliverables are yours to publish, and whether the studio will label which images are generated. Then put the answers in the scope.

Two years ago nobody asked a visualization studio about its software. The question now comes up in almost every scoping call, and it deserves a better answer than a slogan in either direction. A studio that says it uses no AI at all is telling you something about its marketing, not necessarily about its pipeline. A studio that leads with AI may be describing a genuine efficiency or may be describing your deliverables.

The useful move is not to pick a side. It is to ask questions specific enough that the answers mean something. Here are the ones that separate a real answer from a comfortable one, and what each answer tells you.

The test worth applying to any vendor

Start from what you are actually buying. On a project where imagery goes in front of buyers, lenders or a review board, you are buying four things: architectural accuracy, consistency across every deliverable, control over changes, and accountability for what the image claims. Any tool that gets a studio to those four faster is fine. Any tool that costs one of them is not, regardless of what it is called.

That framing is more useful than asking whether AI is involved, because it applies to decisions that have nothing to do with AI. A studio that outsources views to three freelancers who never see each other's work also fails the consistency test. The tool is not the issue, the property is.

The questions, and what the answers tell you

1. Does generated imagery appear anywhere in our final deliverables?

The single most important question, and it should get a direct yes or no. A good answer distinguishes clearly between the front of the process, where generated imagery is genuinely useful for exploring direction, and the deliverable set you publish. An answer that stays abstract about "AI-enhanced workflows" without saying what reaches your files is the answer to be careful with.

2. Is the model built from our drawings, and which set?

This establishes the source of truth. You want to hear your drawing set and revision named, because that is what makes an image traceable back to something you can check. If the answer is vague about where the geometry comes from, everything downstream inherits that vagueness, for the reasons explained in why AI renderings lose architectural accuracy.

3. If we change a material in month three, what happens to the set?

This is the question that reveals whether there is a pipeline or a folder of files. In a model-based process the change propagates: the same model feeds the stills, the animation and the walkthrough, so they cannot disagree. If the answer involves regenerating images, ask what else changes when they do, since regeneration alters things nobody requested. This is the substance of consistency across views.

4. Who is accountable if an image shows something that cannot be built?

A studio working from your documentation can trace a discrepancy: either the model is wrong or the drawing is, and either is fixable at the source. That chain of accountability is a real part of what you are paying for. Where no such chain exists, the exposure lands on whoever published the image, which is you.

5. What post-processing is applied after rendering, and is any of it generative?

A reasonable question rather than a hostile one. Post-processing is normal and always has been. What you are separating is processing that refines an image derived from your model, from generation that introduces content that was never in the model. Ask specifically whether anything is added rather than adjusted.

6. Can you confirm the deliverables are ours to publish?

Provenance and usage rights matter more on a marketing asset carrying your project's name than on an internal deck, and generated imagery raises questions here that a modeled render does not. This is covered from the exposure side in the risks of AI generated images in real estate marketing.

7. Will you label which images are generated and which are modeled?

A studio comfortable with its process will not mind telling you which is which. This matters practically rather than philosophically: it tells your team which images are safe to put in front of a buyer and which are internal, and it prevents a concept image from drifting into a brochure eight weeks later because nobody remembered where it came from.

What to put in the scope

Conversations get forgotten and scopes do not. A few lines carry the answers into the agreement:

• Name the drawing set and revision the model is built from.
• State whether generated imagery may appear in published deliverables, and if so where.
• State that all deliverables derive from a single model set, including animation, VR and plan visuals.
• Require that design changes propagate to every affected deliverable rather than only the one being revised.
• Confirm usage rights over everything delivered.

Why "we do not use AI at all" is not automatically the best answer

It is worth saying plainly, because the opposite bias is now common. Refusing a tool that improves speed without touching accuracy is not rigor, it is just slower. Generated imagery is genuinely useful early: exploring a palette, testing a mood, producing options quickly when nothing is committed. A studio using it there and modeling everything that ships is doing the right thing, and a studio that treats the whole subject as a threat may simply not have thought about it.

What matters is the boundary and whether the studio can articulate it. A vendor who can tell you exactly where the line sits in their process, and why it sits there, has thought about your exposure. That is the signal.

Our own answer, for the record

We use AI where it improves speed and efficiency without giving up accuracy, consistency, control or accountability, and we do not use it where it would cost one of those. The imagery we deliver for leasing, pre-sales, investor presentations and approval submissions is modeled from your documentation. We think that answer should be easy to get from any studio you are evaluating, including this one, which is why the questions above are written so they can be asked of us too.

For how the work is scoped, see real estate rendering services and what 3D rendering costs, or send your drawing set for scope against a real project.

Frequently asked questions

Should I avoid studios that use AI?

Not as a rule. Using generated imagery early to explore direction, while modeling everything that ships, is a reasonable process. What matters is whether the studio can tell you exactly where the boundary sits in their workflow and why. A vendor who has thought about that has thought about your exposure.

What is the single most important question to ask?

Whether generated imagery appears anywhere in your final deliverables. It should get a direct yes or no, with a clear distinction between exploration at the front of the process and the set you publish. Abstract answers about AI-enhanced workflows that never say what reaches your files are the ones to probe further.

How do I know if a studio has a real pipeline?

Ask what happens if you change a material in month three. A model-based process propagates the change across stills, animation and walkthrough because they are views of one model. If the answer involves regenerating images, ask what else changes when they do.

What should go in the scope rather than staying in conversation?

The drawing set and revision the model is built from, whether generated imagery may appear in published deliverables, that all deliverables derive from a single model set, that design changes propagate to every affected deliverable, and confirmation of usage rights over everything delivered.

Why do usage rights come up with generated imagery?

Generated imagery can echo the work it was trained on, which is a different provenance question than a render modeled from your own drawings. On a public marketing asset carrying your project name, that is worth settling before publication rather than after.