How to Use AI-Generated Visuals in a Proposal Without Overselling

What "Overselling" Actually Means With AI Visuals

Overselling happens when an AI rendering presents itself as design intent rather than concept exploration. The visual implies that the firm has committed to a form, a material, or a feature it has not actually engineered, costed, or validated. The architect's intent is usually fine. The problem lives in the gap between what the image suggests and what the design has earned.

AIA Trust frames the underlying phenomenon directly: hallucinations are "a phenomenon in which GenAI invents false information, images, or texts and confidently presents it in the output"⁴. Their published examples are the kind of failure modes every proposal lead should be able to point at: removal of existing immovable structures from a site rendering, and the suggestion of materials that either don't exist or lack the performance capabilities the project requires⁴.

Architect Magazine's "When Architecture Starts Hallucinating" describes a parallel category of failure: geometry hallucination, where asking the model for a building with five floors might produce seven, or load paths that never resolve to a foundation. Diffusion models generate images by learning statistical patterns in training data, not by reasoning about structure. They have no inherent model of thermal breaks, glazing ratios, or how a cantilever transfers load⁵.

Three categories of overselling show up in proposal work most often:

1. Hallucinated content: non-existent materials, miscounted floors, removed site features
2. Aesthetic implication: photoreal finish that reads as design commitment
3. Scope drift: visuals that imply scope items (rooflines, glazing systems, site work) the proposal did not actually price

Just because a rendering looks finished doesn't mean the design is. The AIA already has rules that govern this gap. They were written before Midjourney existed. They apply directly.

What the AIA Code of Ethics Already Says About AI Generated Architecture

The AIA Code of Ethics, last revised in April 2024, already governs AI generated architecture in client proposals. No new rule is needed. Four existing rules apply directly to AI-generated imagery, and AIA Trust commentary maps them explicitly onto generative AI use.

The four rules that matter for proposal work:

In January 2026, the AIA published its Position Statement on Artificial Intelligence and the Guidance for the Responsible Use of AI by Architecture and Design Firms⁶. Both reinforce the same principle that runs through the Code: the architect is the professional of record, and human oversight is the critical factor in responsible AI use.

AIA Trust translates the principle into a working analogy: "Treat GenAI outputs as you would work produced by a new associate: take the time to identify the underlying assumptions, verify their work, and put it through the appropriate quality control processes. You are ultimately responsible for that work."⁴

Read together, the rules require three things of a proposal team using AI generated architecture imagery:

• No misrepresentation of achievable results. The rendering must not imply a deliverable the firm hasn't committed to.
• Reasonable care in the production process. AI outputs need QC before they reach a client, the same way junior-associate output does.
• No use of AI imagery as design documents. Signed and sealed work requires responsible control that AI tools do not provide.

A firm-level AI governance strategy sits on top of these rules. The Code tells you what not to do. It doesn't tell you what to do on Tuesday. Here's a workflow that fits inside it.

A Three-Tier Workflow for AI Visuals in AEC Proposals

Use AI visuals at three escalating tiers: internal exploration, disclosed concept, and hybrid design-stage. The cleanest way to stay inside the AIA Code is to never let a tier-three visual reach a client.

The same logic that governs an AI decision framework for founders applies here: match the tool to the stage, and make the boundary visible.

44% of architects now use AI for concept images¹, and almost all of that volume lives in tier 2. The rule the tiers protect: never show AI to a client without context. Tier 2 needs language on the page. Here's what to use.

Disclosure Language That Doesn't Undermine the Pitch

Strong AI disclosure language tells the client three things in two sentences: what tool was used, what role it played, and that the design of record is developed by licensed professionals. Disclosure framed this way reads as professional rigor.

AIA Trust recommends being upfront about GenAI use in practice: "Be upfront about the use of GenAI tools in your practice. Ensure that all parties are comfortable with your use of these tools by taking the time to explain to relevant stakeholders how GenAI tools are responsibly incorporated."⁴ Proving Ground reaches the same conclusion: architects "may opt to implement an AI disclosure and disclaimer citing the use of AI in the course of their services."⁷

Two paste-able formats cover most proposal contexts.

Short caption (~15 words), used adjacent to an image:

Concept exploration generated with [tool name]; design of record developed by licensed architects at [firm].

Long paragraph (~45 words), used on a proposal page that contains AI imagery:

Selected concept visuals in this pursuit package were produced using generative AI image tools as part of our early-stage option exploration. All design decisions, technical specifications, and signed and sealed work are produced and reviewed by the licensed architects of [firm] under our standard quality control process.

Three things to avoid:

• Don't bury it. Disclosure that lives in 6-point footer text reads as a hedge. Put it next to the visual.
• Don't hedge. "May include AI-assisted elements" is weaker than "concept exploration generated with [tool name]." Be specific.
• Don't apologize. Disclosure is a quality-control statement. Frame it as part of how the firm produces work, and it strengthens the proposal.

Disclosure handles the visible risk. There's a less obvious one underneath: aesthetic anchoring.

The Hidden Trap: Aesthetic Anchoring

Aesthetic anchoring is what happens when a developer sees a photoreal AI rendering early in a project and the form, materials, and feel of that image become the standard the project is held to— regardless of whether the design is buildable at the proposed budget. It's the most expensive trap in AI visuals because it doesn't show up until contract or construction.

Architect Magazine documented the mechanism: when developers are shown gorgeous AI renderings early in projects, it sets an aesthetic anchor that can lead projects to be anchored to forms that are fiscally irresponsible before construction begins⁵. The client points at the rendering during a design review and says, "Build this." The design didn't exist yet. The image did.

The legal mechanism is well-described even though no case law exists yet. Construction-law commentary from Fabyanske, Westra, Hart & Thomson notes that "companies that overpromise and underdeliver on AI capabilities can face claims for fraudulent inducement, which involves allegations that a company made false or misleading representations to secure a contract."⁸ The same logic translates to AI-rendered building forms presented as design intent. Whether the exposure resolves into litigation is open; the mechanism is documented. The hidden costs of AI projects include the downstream contract dispute exposure that aesthetic anchoring creates.

Three signals that aesthetic anchoring is happening with a current client:

• The developer references the proposal rendering by visual detail during design review meetings
• Budget discussions assume material specifications the design team has not actually validated
• Pushback on schematic design feels like "downgrade from the proposal" rather than "design development"

The countermove comes from the rendering studios themselves: intentional imperfection. Over-polished imagery erodes client trust because it looks impossible⁹. Slightly looser, more sketch-like AI visuals at the concept stage signal authenticity and reduce the anchoring effect. Match the finish level of the visual to the certainty level of the design.

Pull it together into a checklist your proposal team can use on Monday.

What to Do on Monday— A Proposal-Team Checklist

The shortest path to AI generated architecture visuals that win work without raising risk is a five-item checklist for every proposal: tier the visual, disclose its role, prevent anchoring, document the human work, and assign a sign-off owner.

1. Tier the visual. Before any AI image goes into a pursuit package, confirm which tier it belongs to (internal, disclosed concept, or design-stage). If it's tier 3, replace it with a traditional rendering or hybrid output.
2. Disclose its role. Use the standard short caption or long paragraph from the disclosure templates above. Place it adjacent to the visual, not buried in a footer.
3. Prevent anchoring. For any photoreal AI image, pair it with a written statement that form, materials, and scale are subject to design development. Where appropriate, choose a less-polished finish level to match the certainty of the design stage.
4. Document the human work. Note which licensed professionals are responsible for the design of record, echoing the AIA Position Statement and Rule 4.102^{3 6}. This is the single sentence on the page that protects the firm.
5. Assign a sign-off owner. One named person at the firm signs off on every AI visual before it leaves the building (principal, BD lead, or technology officer; pick one and keep it consistent).

Five items. Same checklist every proposal. Same person signs off. That is what responsible AI use looks like for an AEC firm: a workflow that any team member can run. Embedding the checklist into the firm's broader practice of building AI culture makes adoption sticky rather than dependent on individual discipline.

A handful of questions come up enough that they deserve direct answers.

FAQ

Is it ethical to use AI generated architecture renderings in an architectural proposal?

Yes, under the AIA Code of Ethics. Rules 3.301 and 4.201 govern client representations and apply directly to AI imagery, provided the visuals do not mislead about achievable results, professional qualifications, or scope³. The architect remains the professional of record⁴.

Do I have to tell my client I used AI to create the rendering?

The AIA Code does not explicitly require disclosure, but AIA Trust strongly recommends transparency about GenAI use in client work⁴. Framed disclosure naming the tool, its role, and the human oversight in place reads as professional rigor and protects against misrepresentation claims under Rule 3.301 and Rule 4.201³.

Can AI-generated renderings be used as construction documents or signed and sealed work?

No. AI outputs are concept artifacts, not professional documents. AIA Rule 4.102 requires responsible control over signed or sealed work, which AI tools cannot provide³. AI may inform exploration behind the scenes; the deliverable is produced and reviewed by licensed professionals.

What's the actual win-rate lift from AI in AEC proposals?

Inconclusive. Only 35% of AEC firms link AI adoption to higher proposal win rates or revenue, per the 2026 QorusDocs 10th Annual Proposal Management Survey². The lift depends on how AI is used. Adoption alone does not produce it.

What's the single biggest risk of using AI renderings in a proposal?

Aesthetic anchoring. The client fixes on a form, material set, or feature before the design is technically and economically validated, leading to budget and feasibility disputes downstream⁵. Construction-law commentary frames the legal exposure as analogous to fraudulent inducement when AI-rendered representations secure a contract the firm cannot deliver against⁸.

AI Speeds the Concept; the Architect Closes the Loop

AI generated architecture has already changed how the first 48 hours of a project conversation can look. The question for an AEC firm in 2026 is whether your firm has a workflow that turns those visuals into won work without turning them into a misrepresentation problem.

AI speeds the concept exploration. The architect closes the loop on what gets built. The firms that know which of those two jobs AI is doing for them are the ones landing in the 35% of AEC firms that can show AI lifts proposal win rates^{1 2}. Human oversight and professional accountability remain the critical factors⁶.

If mapping AI to your firm's proposal workflow feels like a heavier lift than your team can absorb alongside live pursuits, an outside implementation partner can shortcut the design of that workflow. Dan Cumberland Labs helps AEC firms work through these decisions through AI strategy services, using the same operating-manual approach that produced this article.

References

1. Chaos Group / Architizer, "The 2026 archviz pulse: 10 stats defining the new era of design" (2026) — https://blog.chaos.com/statistics-ai-architectural-design-and-visualization
2. Stargazy (citing QorusDocs 10th Annual Proposal Management Survey 2026), "4 numbers every AEC proposal leader should know in 2026" (2026) — https://stargazy.io/resources/4-numbers-every-aec-proposal-leader-should-know-in-2026
3. American Institute of Architects, "AIA Code of Ethics & Professional Conduct" (2024) — https://www.aia.org/code-ethics-professional-conduct
4. AIA Trust, "Ethical Challenges of Generative AI in Architectural Practice" (2024) — https://theaiatrust.com/ethical-challenges-of-generative-ai-in-architectural-practice/
5. Architect Magazine, "When Architecture Starts Hallucinating" (2024) — https://www.architectmagazine.com/technology/when-architecture-starts-hallucinating/
6. American Institute of Architects, "Architects and AI: Practical guidance for a changing profession" (2026) — https://www.aia.org/aia-architect/article/architects-and-ai-practical-guidance-changing-profession
7. Proving Ground, "Code and Conduct: Five areas where AI confronts the Architect's Ethics" (2025) — https://provingground.io/2025/10/22/code-and-conduct-five-areas-where-ai-confronts-the-architects-ethics/
8. Fabyanske, Westra, Hart & Thomson, "Legal Risks of the Use of AI in the Design-Build Process" (2025) — https://www.fwhtlaw.com/blog/2025/05/16/legal-risks-of-the-use-of-ai-in-the-design-build-process/
9. The AEC Associates, "Why Architectural Renderings Lose Clients Before The Proposal Is Reviewed" (2025) — https://theaecassociates.com/blog/why-architectural-renderings-lose-clients-before-proposals-are-even-reviewed/