AI Rendering Disclosure Rules for Architects in 2026

A hillside house that doesn't exist yet, rendered so convincingly it could pass for a finished photograph, sits close to the center of a European law now regulating "deepfakes." AI rendering disclosure has moved from an ethics-class talking point to a practical question with real answers: what that law requires, what your own professional body expects, and what the platforms you post on already enforce. This guide walks through each layer and gives you wording you can use today.

Key takeaways

• As of September 2026, the EU AI Act's Article 50 requires labeling realistic AI content, and architectural renderings are likely covered.
• US law doesn't require this, but the EU Act can reach non-EU firms whose AI output is used in the EU.
• Instagram, YouTube and TikTok require labels on realistic AI video; LinkedIn only shows a credential mark when a file carries one.
• The AIA's toolkit asks firms to disclose AI's role in client deliverables, not in every public post.
• Pair an explicit "AI-generated" label with a plain built-or-concept line to cut risk on any platform.

The short answer for US studios

If a rendering or video shows a project that isn't built, or a built project altered by AI in a way a viewer could mistake for reality, use two lines: an explicit "AI-generated" label placed where a viewer sees it right away, and a plain statement of whether the project is built or a concept. That combination reduces risk under the EU rule, matches the AIA's instinct toward candor, and heads off an advertising-law question. None of this is legal advice; a firm with real EU exposure should get a read from counsel.

The harder call is what counts as AI use that needs disclosure. AI-generated entourage or a sky replacement dropped into an otherwise conventional render, reviewed only inside your team, doesn't need a caption about it; a firm posting a video of a coastal retreat that doesn't exist, animated to look like a walk-through, does.

The EU AI Act: what Article 50 requires from August 2026

Article 50 of the EU AI Act took effect on August 2, 2026. Its fourth paragraph covers what the regulation calls a "deep fake": AI-generated or manipulated content that "resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful" (Article 3(60)). Whether a photoreal rendering of a site that doesn't exist yet counts as resembling something "existing" is the open question behind why lawyers say "likely" rather than "certainly."

Dezeen reported on August 10, 2026 that, according to Szymon Sieniewicz of law firm Addleshaw Goddard, realistic architectural visualizations are likely to fall inside that definition: "If it's realistic scenery with people walking around and it is AI generated, it will need to be labelled as AI generated" ("Realistic AI architectural renderings must be labelled under EU AI Act," Dezeen). A narrower regime applies to "evidently artistic, creative, satirical, fictional or analogous" work, but Dezeen reported that this is unlikely to cover content made partly for promotional purposes, such as renderings and ads: "for the standard businesses in the architecture and design space... it will be the case that they need to be labelled as AI-generated," Sieniewicz said.

Labels must be "clear and distinguishable" and shown "at the latest at the time of the first interaction or exposure" (Article 50); Dezeen reported that the Commission's Code of Practice on Transparency of AI-generated Content calls for an icon reading "AI generated" or "AI modified." For a Reel where a caption is truncated until someone taps "more," that timing rule is a reason to also put a short note on screen from the first frame.

The rule can reach beyond the EU: Article 2(1)(c) extends it to firms "located in a third country, where the output produced by the AI system is used in the Union" (Article 2). Whether a US firm's public post seen by EU viewers meets that bar isn't settled, so a studio with real EU exposure shouldn't assume exemption. Fines under Article 99 can reach €15 million or 3 percent of turnover, though small and mid-sized firms get the lower figure (Article 99); Sieniewicz doesn't expect regulators to move straight to heavy penalties.

AIA guidance and the Code of Ethics

US law doesn't require this, but the AIA has published its own position for members that points the same direction. The AIA AI Firm Toolkit recommends firms "disclose when AI played a substantive role in a deliverable that the client will rely on," naming an AI-generated rendering presented as a project visualization as one of its examples ("AIA AI Firm Toolkit," aifirmtoolkit.aia.org). That guidance is scoped to what a firm hands a client, not to public marketing posts — applying the same candor to a public caption is a reasonable extension of the idea, not something the toolkit itself requires.

The toolkit connects this to Rule 3.301 of the AIA Code of Ethics and Professional Conduct, which bars members from intentionally or recklessly misleading clients about the results their services can achieve. The rule binds AIA members specifically; non-members should check their state board's rules on misleading representations instead.

For a studio building AI video into its marketing, a sensible extension of that instinct — not itself AIA guidance — is to apply the same candor publicly when a rendering or video could lead a client, jury or buyer to believe something untrue about a project's status.

Advertising law: when an AI image becomes misleading

Setting AI aside, Section 5 of the FTC Act prohibits deceptive practices, and that doesn't change because an image was AI-generated rather than retouched by hand. The FTC's Policy Statement on Deception, issued October 14, 1983 and still the standard in use, treats a representation as deceptive if it's likely to mislead a consumer acting reasonably in the circumstances on a point material to that consumer's decision. For an architecture firm, the practical version is narrow: don't let a rendering or video imply a building is complete, occupied or code-compliant when it isn't. This is general information, not legal advice specific to your situation.

Platform labels on Instagram, YouTube, TikTok and LinkedIn

Independent of any law, the platforms you post to already require disclosure for realistic AI content, each its own way, as of September 2026.

Meta requires people to use its disclosure tool for organic posts with photorealistic video or realistic-sounding audio that was digitally created or altered, and says it may apply penalties for skipping it ("Labeling AI-Generated Images on Facebook, Instagram and Threads," Meta, February 2024). Since September 2024, the "AI info" label stays on the post for AI-generated content, and moves into the post's menu for content only edited with AI ("Meta's Approach to Labeling AI-Generated Content and Manipulated Media," Meta, April 2024).

YouTube requires creators to disclose AI content that looks realistic; those who don't may have a label applied for them, or face removal or suspension from the Partner Program ("Disclosing use of GenAI content," YouTube Help). TikTok requires a label — your own caption, sticker or watermark, or its AIGC label — on AI or significantly edited content showing realistic people or scenes; unlabeled content "may be removed, restricted, or labeled" by TikTok's team ("Integrity and authenticity," TikTok Community Guidelines). It also auto-labels uploads carrying a C2PA Content Credential ("Partnering with our industry to advance AI transparency and literacy," TikTok Newsroom, May 2024).

LinkedIn doesn't ask creators to self-disclose. Its Content Credentials feature reads a signed C2PA manifest embedded by the tool that made a file and shows an icon when present; LinkedIn says "it's not yet possible to identify and label all AI-generated and modified content," so an absent icon isn't proof content is human-made ("Content Credentials on LinkedIn," LinkedIn Help).

Two more changes landed since, as of September 2026. Instagram's August 31, 2026 update renamed its "AI creator" label to "AI-generated profile," cutting reach for an account whose main subject is an AI person; it doesn't cover a studio account that just posts project films with Arch2Video's small fictional cast ("Instagram puts new limits on undisclosed AI profiles," TechCrunch). TikTok's guidelines were revised effective September 24, 2026, sharpening the labeling duty above with automatic Content Credentials detection. Both add checks on top of the captions below — see Meta's AI info label on ads and Content Credentials in AI video files.

Disclosure wording you can copy

A caption alone isn't automatically enough, and neither is a status line without naming AI: pair an explicit "AI-generated" label with a specific line about the project, then turn on the platform's own tool where one exists.

For unbuilt or concept work: "AI-generated concept study for a proposed clinic interior. Design not yet built."

For an AI-animated video of a real, completed project: "This video is AI-generated from photographs of the finished project. The people and some details shown are generated and not part of the actual building."

For a rendering shared as a design study: "AI-generated early-stage concept for a courtyard garden addition, exploring massing and material options."

Keep "AI-generated" in the line itself — TikTok's own audience research found the term is widely understood — and add the project's specific status alongside it, not in place of it. A caption naming only the built status, without saying AI was involved, tells a viewer something true but incomplete.

Building a disclosure step into your review process

The easiest way to handle AI rendering disclosure well is to fold it into a review step you already do, not a separate compliance task.

When you review a rendering or video before it goes out — the same pass where you'd catch a warped railing or a wrong material call — add one question: does this show something that isn't built, or use AI in a way a viewer could mistake for an unaltered photograph? If yes, write both lines before the post goes live.

If you use Arch2Video to turn renderings and photographs into video, review every frame: AI-generated motion interprets your images and can change details, and every film also includes a small original fictional cast, so people in your source photos are replaced rather than animated. State the built status regardless of the tool, and if you don't own the source images outright, confirm the license covers video and AI-assisted use before the film goes out (see photo licensing terms).

Build the habit into wherever your team reviews a project, including an award submission, where some programs bar altered images entirely. Our own example films are labeled as concept studies for the same reason.

Frequently asked questions

Does the EU AI Act apply to a US architecture firm?

It can. Article 2(1)(c) extends the Act to firms outside the EU when the output their AI system produces is "used in the Union." Whether a US firm's public post, seen by EU viewers, counts as use in the Union isn't settled yet. A firm with real EU clients or audience should treat that exposure as genuine risk rather than assume it's exempt, and get legal advice.

Do I need to disclose AI use on every rendering my firm produces?

No. The AIA's toolkit ties disclosure to deliverables a client will rely on, not every internal draft, naming an AI-generated rendering presented as a project visualization as one example. A working sketch used only inside your team doesn't need a caption. Applying the same candor to a rendering or video you post publicly is a sensible habit, not something the toolkit requires.

Is there a US law that requires labeling AI images?

Not specifically for architecture firms. The FTC's deception standard bars any ad, AI-made or not, from creating a false impression material to a viewer's decision. California's AI Transparency Act, delayed to align with the EU Act under AB 853, mostly binds large AI providers, not the firms using their tools. The platform rules above govern a US studio's actual day-to-day posting.

Do I need to label renderings made with V-Ray, Enscape or similar software?

Not for AI reasons. A conventional 3D render made without generative AI doesn't trigger any of the disclosure rules in this guide. It still needs an honest built-versus-concept caption if it shows a project that isn't finished. If you add AI-generated entourage, a sky replacement or similar AI edits into that render, treat that layer the same as any other AI-generated content and label it.

Does labeling a video as AI-generated hurt its performance on social media?

Not according to YouTube, which states that disclosing AI content doesn't limit a video's reach or its ability to earn money. Meta and TikTok haven't published a similar statement about reach, though both already require disclosure regardless. A short, honest caption carries far less risk than a misleading video getting flagged or reported after the fact.