Skip to main content
Get started free

AI UGC Video for Real Estate: What You Can Legally Run

Three rules decide which AI UGC video formats a real estate agent can run: who appears on camera, how the property looks, and who may endorse you.

AI UGC Video for Real Estate: What You Can Legally Run

AI UGC video for real estate is creator-style vertical video for listings, produced with generated presenters, voice, or motion instead of a shoot. Real estate is the one vertical where three separate rules apply to it at once: who may appear in a housing ad, how the property may be depicted, and whether a synthetic speaker may act like a satisfied client. Those three decide which formats you can actually run.

Every guide on this subject hands you the same list. Talking-head listing intro, neighbourhood explainer, market update, day in the life. The list is fine. The problem is that it was written for ecommerce and then relabelled, and housing is not ecommerce.

Sell a candle with a generated presenter and you are in ordinary advertising law. Sell a house with one and you are in the Fair Housing Act, the REALTOR Code of Ethics, your MLS photo rules, a state disclosure statute, and a federal rule about who is allowed to give a testimonial. This guide runs the standard format list through all of it and tells you what survives.

Key Takeaways

  • Casting a face is a regulated decision in housing, not a creative one. Section 804(c) of the Fair Housing Act makes it unlawful to publish any advertisement for the sale or rental of a dwelling that indicates a preference based on race, colour, religion, sex, disability, familial status or national origin. When you generate the people in your ad, you are choosing that variable deliberately every time.

  • A generated presenter cannot play a happy client. The FTC’s rule at 16 CFR 465.2(a)(1) makes it an unfair or deceptive act to create a consumer testimonial that misrepresents that the testimonialist exists (ecfr.gov, accessed August 2026). Maximum civil penalty is $53,088 per violation.

  • The property footage is the part you must not generate. California AB 723 was chaptered on 10 October 2025 and requires a conspicuous statement plus a link to the original unaltered image whenever a listing advertisement uses a digitally altered one (leginfo.legislature.ca.gov, accessed August 2026).

  • Two disclosure clocks started this year. New York’s synthetic performer rule took effect 9 June 2026 at $1,000 for a first violation, and EU AI Act Article 50 deployer obligations have applied since 2 August 2026.

  • Meta already treats your ad differently. Housing is a special ad category, so age, gender and postcode targeting are removed to comply with fair housing law. Volume does not rescue a housing campaign the way it rescues an ecommerce one.

  • Four of the six standard formats survive. The two that do not are the ones every tool markets hardest.

What is AI UGC video for real estate?

AI UGC video for real estate is short vertical video made in the style of a personal social post rather than a produced commercial, where some part of it is generated: the presenter, the voice track, the captions, or the motion applied to a still. It runs on Reels, TikTok and Shorts. The subject is usually a listing, a neighbourhood, a market update, or the agent themselves.

The appeal is throughput. A listing needs content for the ten days it sits on the market, and hiring a creator per listing does not scale past a few properties a month. The constraint is that housing advertising is regulated in ways product advertising is not.

The three rule layers ecommerce UGC does not have

Here is the frame worth carrying. Every AI UGC decision in real estate falls into one of three layers, and a format has to clear all three to be usable.

LayerThe questionThe governing rule
CastingWho appears on camera?Fair Housing Act section 804(c); Meta special ad category
DepictionWhat does the property look like?NAR Code of Ethics Article 12; MLS photo rules; California AB 723
EndorsementIs the speaker acting like a client?FTC 16 CFR Part 465
DisclosureDoes the viewer know it is generated?NY GBL 396-b; EU AI Act Article 50; platform policy

Most format lists test against none of these. A few test against the fourth. The first three are where the expensive mistakes live.

Who can appear in a housing ad

Section 804(c) of the Fair Housing Act, 42 U.S.C. 3604(c), makes it unlawful to make, print or publish any advertisement about the sale or rental of a dwelling that indicates a preference, limitation or discrimination based on a protected class. Courts have read that to reach the use of human models as a way of expressing a preference.

HUD used to publish advertising guidance at 24 CFR Part 109 that dealt with this directly, including how models should be used so that advertising reads as open to everyone. Part 109 was removed from the CFR in 1996 (govinfo.gov, accessed August 2026). The statute it interpreted did not go anywhere, and fair housing organisations still teach the same standard.

Generated people change the risk profile here in a way that is easy to miss. When you hire a creator, casting is constrained by who answered the brief. When you generate a presenter, you set age, apparent ethnicity, family composition and everything else with a prompt, on every clip, forever. A brokerage running one generated presenter across two hundred listings has made a single casting decision at scale, and it is documented in the prompt.

The practical answer is not to avoid generated people. It is to treat the roster the way a compliant agency treats a shoot: vary it deliberately, keep the ads open on their face, and never let the model default decide. Holding one creator persona steady across a campaign is a real capability, and in housing it is also a decision you should be able to explain.

Meta has already built its side of this in. Housing is a special ad category, which strips age, gender and postcode targeting from the campaign. Meta introduced these restrictions to comply with fair housing law (about.fb.com, accessed August 2026). The volume-and-variants playbook that works for a skincare brand does not transfer cleanly, because half the levers are gone.

Can an AI presenter give a testimonial about your service?

No. This is the cleanest rule in the set and the one most often broken.

The FTC’s Rule on the Use of Consumer Reviews and Testimonials took effect on 21 October 2024. Section 465.2(a) makes it an unfair or deceptive act and a violation of the rule for a business to write, create or sell a consumer testimonial that materially misrepresents, expressly or by implication, that the testimonialist exists, that they had experience with the service, or what that experience was (ecfr.gov, accessed August 2026).

A generated person on camera saying your brokerage sold their house in nine days fails clause (a)(1) on its own terms. The testimonialist does not exist. It does not matter that the sale was real, and it does not matter that you added a small caption saying the video uses AI, because the misrepresentation is about a person rather than about the technology.

Civil penalties under section 5(m)(1)(A) of the FTC Act are capped at $53,088 per violation (ecfr.gov, accessed August 2026).

The line that keeps you clear is simple. A generated presenter may narrate, explain, host and present. It may not have bought a house from you.

What the property may look like on camera

The property footage is the half of the frame you should generate least.

NAR’s Code of Ethics Article 12 requires REALTORS to be honest and truthful in their real estate communications and to present a true picture in advertising, marketing and other representations. Standard of Practice 12-10 extends that duty to internet content and images specifically, and its fifth clause prohibits “otherwise misleading consumers, including use of misleading images” (nar.realtor, accessed August 2026). A generated shot of a room that does not look like that room is a true-picture problem before it is anything else.

California went further and made it statutory. AB 723 was chaptered on 10 October 2025 as Chapter 497. It requires a licensee using a digitally altered image in an advertisement or promotional material for a property to include a reasonably conspicuous statement on or adjacent to the image saying it has been altered, plus a link, URL or QR code to a publicly accessible page carrying the original unaltered image (leginfo.legislature.ca.gov, accessed August 2026).

MLS rules add a third layer that varies by board. Common requirements are an on-image label reading “Virtually Staged” or “Digitally Altered” that stays readable at thumbnail size, disclosure in the listing remarks, and the unaltered original included in the photo set. Check your own MLS handbook rather than a summary, because these genuinely differ.

None of this bans virtual staging, and virtual staging remains one of the highest-value uses of generation in this industry. It bans undisclosed alteration presented as the property. If you want the prompt-level detail on doing it well, we keep a set of virtual staging prompts for exactly that job, and AI real estate photography covers the stills side.

The workable split: generate the presenter, the captions, the b-roll of nothing in particular, and the motion on a still you actually own. Shoot the property.

The disclosure layer moved twice this year

Two rules changed inside the last three months, which is why guidance written in 2025 is already wrong.

New York. General Business Law section 396-b took effect 9 June 2026. Anyone dealing in property or services who produces an advertisement must conspicuously disclose that a synthetic performer appears in it, where they have actual knowledge. A synthetic performer is a digitally created asset made with generative AI that is intended to give the impression of a human performer not recognisable as any identifiable person. Penalties run $1,000 for a first violation and $5,000 for subsequent ones (nysenate.gov, accessed August 2026).

The EU. AI Act Article 50 deployer obligations have applied since 2 August 2026. If you deploy a system that generates or manipulates video constituting a deepfake, you disclose that the content is artificially generated, clearly and at first exposure (artificialintelligenceact.eu, accessed August 2026). The watermarking duty sits with the model provider. The human-facing disclosure sits with you.

Platforms. Meta’s mandatory advertiser self-disclosure covers social issue, election and political ads. Separately, from 1 June 2026 Meta applies automated detection to identify ad media created or edited with third-party generative tools and attaches an “AI info” label through About this ad, with no advertiser action needed (about.fb.com, accessed August 2026). TikTok requires AI disclosure across organic posts, branded content and ads. Assume the label appears whether or not you apply it, and write copy that reads fine with it attached.

Which formats survive all three layers

Here is the standard six-format list, scored against the framework.

FormatCastingDepictionEndorsementVerdict
Agent talking-head listing introReal agent, no issueReal listing photosAgent is not a clientRun it
Neighbourhood explainerGenerated presenter is fine if the roster variesStock or shot b-rollInformationalRun it
Objection answerSameNo property depiction neededInformationalRun it
Market updateSameCharts and text, not propertyInformationalRun it
Before and after revealNo people neededThe whole risk sits hereNoneOnly with AB 723-style disclosure and the original alongside
Day in the life of a buyerHighest fair housing exposureReal footageReads as a client storyDo not generate the buyer

The two weakest formats are the two the category markets hardest, because a synthetic buyer walking into their new home is the most emotionally effective thing a generator can make and the single riskiest thing to publish. If you want that story, film an actual client with written consent.

The four that survive share a property. The generated person is a narrator, never a customer, and the property on screen is real. For the production steps once you have picked a format, how to make a real estate video from listing photos covers the build.

The 403% number, and what to budget from instead

You will meet a statistic in every article on this topic: listings with video receive 403% more inquiries, attributed to the National Association of REALTORS. We went looking for it in NAR’s own research and could not find it in any NAR publication. Several secondary write-ups trace it instead to a vendor page. It has been repeated since roughly 2015 without an accessible primary source, and AI real estate video marketing traces it and a second zombie figure in full.

We are not saying video does not work. We are saying do not build a budget on a number nobody can produce the study for, in a year where your own analytics will tell you the truth for free. Post ten pieces, measure saved and shared rather than views, and use that.

That habit matters more here than in ecommerce, because housing campaigns run without the targeting levers that normally let you buy your way to a signal.

What it costs to produce

Generated video is priced by the second, which makes it the most expensive operation in any AI creative stack. That is the number to plan against.

In DesignerBox, video cost is credits per second multiplied by duration. Veo 3.1 generates native audio alongside the picture, in 4, 6 or 8 second clips at 720p, 1080p or 4K (deepmind.google, accessed August 2026), and an 8 second Veo 3 clip with audio runs 6,400 credits. A Seedance Pro Fast clip at 720p and 5 seconds runs 150. Images are 5 credits each.

Plans run Free at 112 credits, Basic $15 for 500, Pro $35 for 1,000, Premium $75 for 2,500, and Ultra $200 for 8,000. AI video needs Premium or higher. Before committing, count the clips a listing actually needs and multiply, because the gap between the cheapest and dearest video model is more than fortyfold. The full plan ladder has the rest.

For a listing at three clips a week over two weeks, the cheaper models make the arithmetic work and the flagship ones do not. That is the honest answer, and it is the same answer we give ecommerce brands in how to create TikTok video ads with AI.

Where DesignerBox fits, and where it does not

DesignerBox is a campaign pipeline built on a photo you already own. Upload a listing photo and it produces the stills, the vertical video and the ad frames from that source, with 13 image and video models across six providers behind one subscription. Video Studio handles the motion, Ad Studio sizes the frames for Reels, TikTok and Facebook, and the AI talking avatar tool runs the presenter step. The real estate pages cover the vertical specifically.

The fit is good because of the direction of travel. Everything starts from your real photograph rather than a text prompt, which is precisely what the depiction layer above demands. Nothing comes out looking generic AI, because nothing was invented from nothing.

What DesignerBox does not do: it does not check your MLS handbook, it does not apply your state’s alteration disclosure for you, and it does not lip sync footage you already shot. On that last one, when a real person appears in the ad sets out the tooling split. For the voice track specifically, AI voiceover for UGC ads compares the three production paths.

Compliance is yours. The tooling only decides how fast you can produce something compliant.

Tool choice narrows this further than the rules do. We compared which AI avatar tool will let you use your own face.

FAQ

Yes, with conditions. The generated element must not be a client giving a testimonial, the property must be depicted truthfully with any alteration disclosed, the casting must not signal a preference under fair housing law, and the video must carry whatever disclosure your state and platform require.

Do I have to disclose AI in a real estate video?

In New York, yes, if a synthetic performer appears, effective 9 June 2026. In the EU, yes, under AI Act Article 50 since 2 August 2026. On TikTok, yes, across organic and paid. On Meta, mandatory advertiser disclosure covers political and social issue ads, but automated detection has applied an “AI info” label to commercial ads since 1 June 2026 regardless.

Can I use an AI avatar as a fake happy client?

No. The FTC rule at 16 CFR 465.2(a)(1) makes it an unfair or deceptive act to create a testimonial that misrepresents that the testimonialist exists. Maximum civil penalty is $53,088 per violation. A generated presenter may narrate and explain, but it may not have been your customer.

Does virtual staging need to be disclosed?

Yes, in practice everywhere. California AB 723 requires a conspicuous statement plus a link to the original unaltered image. Most MLS boards require an on-image label and the unaltered original in the photo set. NAR Article 12’s true-picture duty applies regardless of state.

Which AI UGC formats are safest for real estate?

The agent’s own talking head, neighbourhood explainers, objection answers and market updates. All four use a narrator rather than a customer, and none requires generating the property. Before-and-after reveals need alteration disclosure, and a generated buyer story is the format to avoid.

How long should a real estate UGC video be?

Fifteen to thirty seconds for social feeds. Generated clips arrive shorter than that. Veo 3.1 produces 4, 6 or 8 second clips, so a 30 second piece is assembled from several, which is also why the credit cost adds up faster than people expect.

Can I generate the property interior instead of shooting it?

Not as a depiction of that property. Generating an interior that does not match the real room is a true-picture problem under NAR Article 12 and a disclosure obligation under California AB 723 and most MLS rules. Shoot the property, then stage or relight the real photograph and disclose the alteration.

Does DesignerBox work for real estate?

Yes. It starts from a photo you already own and produces stills, vertical video and ad frames from it, which matches what the depiction rules require. It does not handle your MLS or state disclosure obligations, and it does not lip sync existing footage.

Where to go next

Score your current content calendar against the four-layer table before you produce anything else. Most agents find that two of their planned formats need a change of speaker or a disclosure line, and that the fix is a rewrite rather than a reshoot.

Then pick the production path. If a generated presenter is doing the talking, AI UGC ad generators covers what the category costs. If you want one face held steady across a whole campaign, reusing one AI presenter across a campaign explains the mechanics and the consent question underneath it.

For the budget question underneath all of this, real estate videographer cost vs AI video runs the per-listing comparison against what a shoot actually costs.

Sources

All accessed August 2026.

  • FTC Rule on the Use of Consumer Reviews and Testimonials, 16 CFR 465.2, including the prohibition on misrepresenting that a testimonialist exists: ecfr.gov
  • FTC maximum civil penalty amounts, 16 CFR 1.98: ecfr.gov
  • Removal of HUD’s fair housing advertising regulations at 24 CFR Part 109: govinfo.gov
  • NAR Code of Ethics Article 12 and Standard of Practice 12-10, 2026 edition: nar.realtor
  • California AB 723, chaptered 10 October 2025, digitally altered real estate images: leginfo.legislature.ca.gov
  • New York synthetic performer disclosure, GBL 396-b, effective 9 June 2026: nysenate.gov
  • EU AI Act Article 50 transparency obligations, applicable from 2 August 2026: artificialintelligenceact.eu
  • Meta housing special ad category and ads fairness restrictions: about.fb.com
  • Meta generative AI transparency for ads, including automated detection from 1 June 2026: about.fb.com
  • Veo 3.1 native audio, clip durations and resolutions: deepmind.google

Regulatory and platform requirements verified from the primary sources listed above as of August 2026. This is not legal advice, and MLS rules vary by board. Individual results vary.

Vytas

Founder at DesignerBox

Vytas is a founder at DesignerBox, from the team behind LoadFocus, FocusBox and PostNext. He writes about turning one product photo into a full campaign, and the pipelines that keep every asset on brand.

Follow along on Instagram at @designerboxai for campaign breakdowns.

Every top video model, one bill

Veo 3.1, Sora 2 Pro, Kling 2.6 Pro, Seedance 2.0 and Runway Gen-4.5 are built in. Switch models per shot without a second subscription or a second login.

Start free

Upload one product photo. Ship the whole campaign, without a photoshoot.