AI in Real Estate 2026: How Artificial Intelligence and Virtual Staging Are Reshaping the Home-Selling Industry

4 ott 2026

Executive Summary

Artificial intelligence has moved from the margins of residential real estate to its daily workflow in less than three years. What began as agents experimenting with general-purpose chatbots to draft listing descriptions has become a layered technology stack touching nearly every stage of the home-selling process: pricing, prospecting, marketing, visual presentation, transaction management, compliance and financing.

The adoption data is now substantial enough to describe the shape of the market with some confidence. The consumerfinance gov board Website and also the 2025 Technology Survey, fielded in July 2025 among a random sample of active members, found that 41 percent of REALTORS® were currently using AI or generative AI, 20 percent used AI tools daily, and 32 percent had not yet used AI in their business at all. NAR’s 2026 survey, fielded in June 2026, showed the share of members using emerging technologies such as AI while still learning to use them effectively rising to 67 percent, up from 59 percent a year earlier. Time savings remains the dominant motivation, cited by 81 percent of respondents in 2026, up from 66 percent in 2025.

At the industry level, Morgan Stanley Research estimates that AI could automate 37 percent of tasks performed in real estate investment trusts and commercial real estate firms, representing about $34 billion in operating efficiencies by 2030. Its analysis singled out brokerage and real estate services as the segments with the largest potential uplift, with operating cash flow gains of up to 34 percent.

Within residential brokerage, one application stands out for the directness of its impact on how homes are presented and sold: AI virtual staging. Virtual staging addresses a long-standing problem, the difficulty buyers have imagining an empty or poorly furnished space, at a fraction of the cost and time of physical staging. Generative AI has collapsed the cost of a staged image from tens of dollars to, on high-volume subscription plans, well under a dollar, and reduced turnaround from days to seconds. At the same time, it has introduced new risks, most notably the unintended alteration of property features, which regulators have begun to address. California’s Assembly Bill 723, effective January 1, 2026, now requires real estate licensees to disclose digitally altered listing images and provide access to the originals.

This report examines the state of AI across the residential real estate value chain, with particular depth on virtual staging: its market economics, the evidence for its effectiveness, the technical architecture behind modern AI staging systems, and the regulatory and operational considerations that brokerages, technology vendors and listing platforms must navigate.

Key findings

Adoption is broad but shallow. A majority of agents now use some form of AI, but most use is concentrated in text generation for marketing and communications. In NAR’s 2025 survey, 46 percent of respondents reported no noticeable business impact from AI, while 17 percent reported a significantly positive impact.
Marketing is the leading use case. NAR’s technology reporting shows 56 percent of REALTORS® using AI to create social media posts and 52 percent using it to draft emails and follow-ups. Listing descriptions are among the most common applications.
Visual content is the next frontier. Photos remain the most valuable website content for buyers, cited by 41 percent of recent buyers in NAR’s 2024 Profile of Home Buyers and Sellers, ahead of detailed property information at 39 percent and floor plans at 31 percent. AI tools that improve or transform listing imagery, including virtual staging, address the content buyers value most.
Staging works, mostly through comprehension. NAR’s 2025 Profile of Home Staging found 83 percent of buyers’ agents said staging made it easier for buyers to visualize a property as their future home. Effects on price are more modest and more variable.
Regulation is catching up. Disclosure rules for altered images, fair housing obligations in AI-driven advertising, and emerging state AI laws are reshaping how AI can be deployed in listing marketing.
Infrastructure and governance determine outcomes. The firms seeing measurable returns treat AI as a governed operational capability, with quality control, disclosure standards and data policies, rather than as a collection of individual agent experiments.

Market Context: A Constrained Housing Market Raises the Stakes

AI adoption in residential real estate is unfolding against a difficult market backdrop. NAR’s 2025 Profile of Home Buyers and Sellers, based on a survey of buyers who purchased between July 2024 and June 2025, documented a series of record-setting extremes. First-time buyers fell to 21 percent of the market, the lowest share since NAR began tracking in 1981. The median age of first-time buyers rose to 40, and the median age of repeat buyers to 62, both all-time highs. Twenty-six percent of buyers paid all cash, matching the record. Sellers were older too, with a median age of 64, and had stayed in their homes a median of 11 years before selling, the longest tenure in the series.

These conditions shape the business case for AI in two ways. First, lower transaction volumes put pressure on brokerage margins and agent productivity, increasing the appeal of tools that reduce the cost of marketing a listing or servicing a client. Second, a buyer pool that is older, wealthier and more selective raises the bar for listing presentation. When fewer buyers are transacting, each listing competes harder for attention, and the quality of a listing’s first impression online matters more.

At the same time, agent representation remains near-universal. NAR’s 2025 profile found 88 percent of buyers purchased through an agent or broker, and 91 percent of sellers used a broker, an all-time high. AI in residential real estate is therefore overwhelmingly an agent-augmentation story rather than a disintermediation story. The relevant question for most market participants is not whether AI replaces agents, but which agents and brokerages use it to compete more effectively.

AI Adoption Among Agents and Brokerages

The adoption curve

NAR’s annual technology surveys provide the most consistent view of agent behavior. The 2025 survey, which received 1,241 usable responses from a random sample of 49,233 active members, found that 70 percent of REALTORS® used at least one form of emerging technology, with AI and generative AI the most common at 41 percent. Frequency data showed a market split almost evenly into thirds: 20 percent used AI daily, 22 percent weekly, 27 percent a few times a month, and 32 percent not at all.

The 2026 survey, based on 1,165 usable responses from 73,718 randomly selected members invited in June 2026, suggests the curve is still moving. The share of members actively using emerging technologies while still learning rose from 59 to 67 percent, and the share confident enough to teach others increased from 8 to 12 percent. NAR’s technology reporting for the same period indicates that 23 percent of members use AI daily and 55 percent report a positive effect on their business.

Smaller industry surveys point higher. A February 2026 adoption survey by Realtors Property Resource (RPR), with 225 respondents, reported 82 percent of participating members currently using AI, with writing listing descriptions (68 percent) and creating social media content (59 percent) the leading applications. The difference between these figures and NAR’s broader survey likely reflects sample composition: smaller, opt-in surveys tend to attract more technology-engaged respondents. For market sizing, the NAR random-sample figures are the more conservative and defensible baseline.

Which tools agents use

The tool landscape among individual agents remains dominated by general-purpose assistants. In NAR’s 2025 survey, the most cited AI tools were OpenAI’s ChatGPT (58 percent), Google’s Gemini (20 percent) and Microsoft’s Copilot (15 percent). Purpose-built real estate AI tools, including CRM-embedded assistants, AI-powered listing marketing suites, automated valuation tools and virtual staging platforms, are typically adopted at the brokerage or team level rather than by individual agents.

Broader technology use provides context. eSignature remains the most widely used tool at 79 percent, followed by social media at 75 percent, drone photography and video at 52 percent, and AI-generated content at 46 percent. The pattern is instructive: AI is being layered onto an already digital workflow, primarily in content production, rather than replacing core transaction infrastructure.

Impact: a widening gap

The most revealing statistic in NAR’s 2025 data is not adoption but impact. Seventeen percent of respondents reported a significantly positive impact from AI, 33 percent a moderately positive impact, and 46 percent no noticeable impact. In other words, roughly half of agents using or trying AI were not yet seeing measurable business results.

The commercial real estate sector shows a similar gap. A JLL global technology survey of more than 1,500 senior decision-makers across 16 markets, published in October 2025, reportedly found that 88 percent of owners and investors and 92 percent of occupiers were piloting AI, but only 5 percent said they had achieved all of their AI goals.

Several factors explain the gap in residential brokerage:

Task selection. Most agent AI use targets low-value writing tasks. Saving 15 minutes on a listing description is useful but rarely changes business outcomes.
Lack of integration. General-purpose chatbots sit outside CRM, MLS and transaction systems, so outputs must be copied and pasted, limiting automation.
Quality control burden. AI outputs require review for accuracy, fair housing compliance and brand consistency. Without standard processes, the review time can offset the time saved.
Skills. Only 12 percent of respondents in NAR’s 2026 survey felt confident enough to teach others, indicating that most users remain early in their learning curve.

Motivations and intentions

Agent motivations have shifted toward productivity and client experience. In 2026, 81 percent of REALTORS® cited saving time as a primary reason for adopting new technology, 71 percent cited improving the client experience, and 57 percent cited closing more deals, each up from the prior year. Looking ahead, 54 percent expressed interest in using generative AI for content creation in the next year, 51 percent in client relationship management tools, and 48 percent in AI-powered lead generation and follow-up.

Brokerage-level adoption

Brokerages occupy a different position from individual agents. They control technology budgets, brand standards, compliance obligations and data. Larger brokerages and franchise systems increasingly provide AI tools centrally, including:

AI writing assistants embedded in listing and marketing platforms
Automated lead routing and response
Virtual staging and photo enhancement for listing media
Transaction document review and checklist automation
Agent coaching and performance analytics

Central provision changes the economics. A brokerage can negotiate volume pricing with AI vendors, standardize disclosure and quality control, and capture data on which tools actually improve listing performance. For virtual staging in particular, brokerage-level procurement allows consistent labeling, approved styles and a single audit trail of original and altered images, which matters increasingly for regulatory compliance.

AI Across the Home-Selling Value Chain

AI applications in residential real estate can be mapped against the stages of a typical listing: pricing, prospecting, preparing and marketing the property, managing showings and offers, and closing. Each stage has a different maturity level, data dependency and risk profile.

1. Valuation and pricing

Automated valuation models (AVMs) are the most mature AI application in residential real estate, predating the generative AI wave by more than a decade. AVMs estimate a property’s market value using statistical and machine-learning models trained on public records, sales history, listing data and, increasingly, imagery and text from listings.

Consumer-facing AVMs illustrate both their power and their limits. Zillow publishes accuracy figures for its Zestimate: according to its accuracy data refreshed in August 2026, the nationwide median error rate was 1.78 percent for homes on the market and 7.20 percent for off-market homes. The gap is significant. On-market estimates can incorporate the list price, listing description, photos and days on market, effectively anchoring on information the seller and agent have already provided. Off-market estimates rely on public records and comparable sales, with substantially more uncertainty. On a $500,000 home, a 7.2 percent median error corresponds to roughly $36,000 in either direction, and by definition half of estimates are off by more than the median.

For agents and brokerages, AI-assisted pricing tools increasingly blend AVM outputs with comparative market analysis workflows, adjusting for condition, renovations and micro-location features that public data captures poorly. Computer vision models that assess condition and finish quality from listing photos are an active area of development, linking valuation directly to the visual content of a listing.

2. Lead generation and client acquisition

AI is reshaping how agents identify and convert prospects. Common applications include:

Predictive seller models that score homeowners on their likelihood to sell, using signals such as tenure, equity, life events and market activity
Automated lead response, with AI assistants answering portal inquiries, qualifying leads and booking appointments around the clock
Personalized nurture campaigns, where AI drafts tailored emails and texts based on client behavior and preferences
Conversation intelligence that summarizes calls and flags follow-up actions inside the CRM

These tools are where NAR members express growing interest: 48 percent in the 2026 survey said they were interested in AI-powered lead generation and follow-up over the next year. They also carry compliance risks, including consent requirements for automated texting and calling, and fair housing obligations when targeting advertising.

3. Listing preparation and marketing

Marketing is the most widely adopted category of AI use and the area with the most direct impact on how a home is presented. It includes:

Listing descriptions generated from property data, agent notes and photos
Social media content, including captions, short videos and carousel posts
Photo enhancement, such as exposure correction, sky replacement, lawn enhancement and twilight conversions
Virtual staging and decluttering, which add or remove furniture in listing photos
Virtual renovation, which visualizes changes to finishes, paint or flooring
Floor plan generation from photos, scans or video walkthroughs
Video and 3D tours, including AI-assisted editing and automated tour creation

Because photos are the most valued website content among buyers, AI applications that improve listing imagery have outsized influence on buyer engagement. Virtual staging, examined in detail in the next sections, is the most consequential of these applications because it changes not just the quality of an image but what the image depicts.

4. Showings, communication and negotiation

AI assistants increasingly handle scheduling, showing feedback collection and routine client communication. In NAR’s technology reporting, 52 percent of REALTORS® use AI to draft emails and follow-up communications. Some platforms offer AI summaries of offers, comparison of offer terms and drafting of counteroffer communications, always under agent supervision.

Negotiation itself remains firmly human. AI can summarize terms and model scenarios, but the judgment, relationship management and fiduciary responsibility involved in negotiation are not tasks brokerages are delegating to automated systems.

5. Transaction management and compliance

Transaction coordination is document-heavy and deadline-driven, making it well suited to AI-assisted automation. Applications include extracting key dates and terms from contracts, flagging missing signatures or disclosures, generating task checklists, and answering agent questions about brokerage policies. AI document review can reduce errors, though brokerages remain responsible for the accuracy of transaction files.

6. Mortgage and closing

Mortgage lenders use AI for document classification, income verification, fraud detection and underwriting support. Title and escrow companies use AI to detect wire fraud patterns and verify identities. These applications are largely invisible to buyers and sellers but affect closing timelines and risk.

Where value concentrates

Across the value chain, the applications with the clearest measurable returns share three characteristics: they replace a cost the brokerage or agent already pays, they integrate into existing systems, and their outputs can be verified. Virtual staging fits all three. It directly substitutes for physical staging or human-made virtual staging, it plugs into the listing photo workflow, and its outputs can be checked against original images. This combination explains why virtual staging has become one of the most commercially significant AI applications in residential listing marketing.

Virtual Staging: Market, Evidence and Economics

Defining the category

Virtual staging is the digital addition of furniture, decor and sometimes lighting to photographs of empty or sparsely furnished spaces, producing images that show how a property could look when occupied.

It’s a very powerful AI tool if you use the Best Virtual Staging Software: AI can create incredibly realistic and stylish results with DIY real estate home staging apps for Realtors and AI Interior Design 3D Apps.

It sits within a broader family of listing image services that vendors often bundle together:

Virtual staging: adding furniture and decor to empty rooms
Virtual decluttering or item removal: removing existing furniture or clutter, often as a precursor to restaging an occupied home
Virtual renovation: visualizing changes to fixed features such as flooring, cabinetry, paint or countertops
Image enhancement: exposure correction, white balance, straightening, sky replacement, lawn greening and twilight conversion

The distinction matters commercially and legally. Adding movable furnishings is broadly accepted in listing marketing when disclosed. Changes that alter how the property itself is represented, such as removing defects, changing views or replacing finishes, carry significantly greater legal and ethical risk.

Three generations of production

The category has evolved through three production models, which coexist in today’s market.

Manual compositing. Early virtual staging relied on photo editors cutting furniture from product images and pasting it into listing photos, painting in shadows by hand. Quality depended entirely on individual skill, and perspective errors were common.

3D rendering. The second generation borrowed from architectural visualization. Artists reconstruct room geometry from the photograph, place licensed 3D furniture models in a virtual scene matched to the camera, render them with physically based lighting, and composite them into the photo. This approach produces accurate perspective and convincing shadows but is labor-intensive, which is why human-made staging is typically priced per image with one- to two-day turnaround.

Generative AI. The current generation uses diffusion-based image generation combined with depth estimation, segmentation and layout models. Users upload a photo, select a room type and style, and receive staged images in seconds. Many vendors now operate hybrid models in which AI produces a first pass and human editors correct errors.

Pricing

Published price lists reviewed for this report fall into three broad tiers, with significant variation by vendor:

Human-made or human-reviewed staging: commonly about $15 to $75 per image, with one- to two-day turnaround and limited revisions. As one reference point, BoxBrownie listed standard 2D virtual staging at $24 per image in mid-2026.
AI virtual staging software: typically sold by subscription or credits, with effective per-image costs ranging from well under $1 on high-volume plans to around $15 on smaller or premium plans, and turnaround measured in seconds.
Add-on services: decluttering, virtual renovation, twilight conversion and rush delivery, typically priced above basic staging.

Physical staging provides the comparison point. NAR’s 2025 Profile of Home Staging reported a median cost of $1,500 when sellers used a professional staging service and $500 when the listing agent staged personally. Physical staging also involves scheduling, delivery and monthly rental terms.

For a typical listing with five staged rooms, the cost difference is stark. At $24 per image, human-made virtual staging costs about $120; AI staging on a subscription plan may cost a few dollars; physical staging at the median professional cost is $1,500 before any extensions. This cost structure is the core of virtual staging’s commercial appeal and explains its spread from vacant luxury listings to everyday mid-market homes and rentals.

Evidence of effectiveness

The largest recurring dataset on staging effectiveness is NAR’s Profile of Home Staging. The 2025 edition, based on a survey conducted in February 2025, reported:

83 percent of buyers’ agents said staging made it easier for buyers to visualize a property as their future home
49 percent of sellers’ agents said staging reduced time on market
29 percent of agents reported staging led to a 1 to 10 percent increase in the dollar value offered
17 percent of buyers’ agents reported a 1 to 5 percent increase in the value offered compared with similar unstaged homes, while 41 percent said staging had no impact on the value offered
31 percent of buyers’ agents said buyers were more willing to walk through a home they had seen online
32 percent of buyers’ agents said staging would positively affect value if the home were decorated to buyers’ tastes

Three caveats apply. The data reflects agent perceptions rather than controlled experiments. It covers staging broadly and does not isolate virtual staging. And it shows mixed effects on price, with a meaningful share of agents reporting no price impact.

The most defensible interpretation is that staging reliably improves buyer comprehension of a space, which can increase engagement and showings, while effects on sale price vary by property and market. Virtual staging plausibly delivers much of the comprehension benefit at the online browsing stage, where buyers increasingly make shortlisting decisions, but it cannot carry that benefit into the physical showing.

Which rooms matter most

NAR’s staging research consistently places the living room among the most important rooms to stage, with the primary bedroom, kitchen and dining areas close behind. The guest bedroom ranked least important at 7 percent in the 2025 report. For virtual staging vendors and brokerages, this supports a selective approach: staging the rooms that appear in the first listing photos and rooms whose function a buyer might misread, rather than staging every image.

The expectation gap

NAR’s 2025 staging report also found that 58 percent of agents said buyers were disappointed when homes did not match what they had seen on television. The finding has a direct parallel in virtual staging: when online images present an idealized interior and the buyer walks into an empty room, the gap between expectation and reality can erode trust. Disclosure, side-by-side originals and realistic, correctly scaled staging are the commercial responses to that risk, and increasingly the regulatory ones as well.

How AI Virtual Staging Works: A Technical View

Modern AI virtual staging systems are not a single model. They are pipelines that combine several computer vision and generative components, each solving a different part of the problem: understanding the room, deciding what may change, generating new content, and verifying that the result is faithful to the property. Implementations differ between vendors, but the architecture below describes the common building blocks.

Stage 1: Ingestion and normalization

The pipeline begins with a listing photograph, typically a high-resolution JPEG captured with a wide-angle lens and often produced by blending bracketed exposures. Pre-processing steps commonly include:

Metadata parsing to read camera, lens and orientation information
Lens distortion correction to straighten lines curved by wide-angle optics
Resolution normalization, since generative models operate at fixed working resolutions and outputs must later be restored to listing quality
Exposure and color normalization, so lighting estimation downstream is not confused by inconsistent white balance

Input quality has a direct effect on output quality. Converging verticals from a tilted camera, very little visible floor, or heavily mixed light sources all degrade the downstream geometry and lighting estimates.

Stage 2: Scene understanding

Before anything can be added to a room, the system must understand its structure. Three families of models typically do this work.

Monocular depth estimation. Depth models infer the distance of every pixel from the camera using a single image. Open research models in this family, trained on large and diverse image datasets, have made robust indoor depth estimation widely accessible. The resulting depth map tells the system where the floor recedes, where walls stand and how large objects should appear at each position.

Semantic segmentation. Segmentation models label each pixel by category: floor, wall, ceiling, window, door, cabinetry, fixtures, existing furniture. These labels are essential for deciding which regions may be modified. Windows, doors and built-ins must be preserved; open floor and wall areas are candidates for furniture.

Layout and geometry estimation. Layout models estimate the room’s planes and vanishing points, approximating the box-like structure of the room and the camera’s position and focal length within it. This geometry anchors furniture placement in correct perspective.

Together, these outputs form a structured representation of the room that can guide generation far more precisely than a text prompt alone.

Stage 3: Defining what may change

The most important design decision in an AI staging system is the edit mask: the set of pixels the generative model is allowed to modify. A conservative mask limits generation to the floor area and the lower portions of walls where furniture plausibly stands, and explicitly protects windows, doors, fixtures, flooring texture outside furniture footprints, and ceiling features.

Poorly designed masks are the root cause of many staging failures. If the mask includes a window, the model may redraw it. If it includes the floor, the model may change the flooring material. Systems that rely on unconstrained whole-image generation are the most likely to produce unintended alterations that create legal exposure under disclosure rules.

For decluttering, the process runs in reverse: segmentation identifies existing furniture, and an inpainting model fills the masked area with plausible floor and wall surfaces. This is technically harder than staging an empty room, because the model must reconstruct surfaces it cannot see, such as the floor beneath a sofa, and errors there can misrepresent the condition of the property.

Stage 4: Conditioned generation

The core generation step usually relies on latent diffusion models. These models generate images by starting from random noise in a compressed latent space and iteratively denoising it over many steps, guided by conditioning signals. Each denoising step is a full pass through a large neural network, which is why image generation is computationally intensive.

For staging, several types of conditioning are combined:

Text conditioning describing room type, style and specific furniture, such as “Scandinavian living room, light oak floor, gray three-seat sofa, wool rug”
Structural conditioning, using control networks that feed depth maps, edge maps or segmentation maps into the diffusion process so generated furniture follows the room’s geometry
Inpainting conditioning, so the model only generates inside the edit mask and blends smoothly with the preserved image
Style conditioning, using reference images or fine-tuned adapters trained on interior design imagery, to produce consistent looks across a listing

Vendors often fine-tune base models on curated interior datasets, sometimes using lightweight adaptation techniques, to improve furniture realism, scale and style consistency. The number of denoising steps, the output resolution and the strength of each conditioning signal all trade off quality, speed and cost.

Stage 5: Lighting, shadows and compositing

Generated furniture must match the scene’s light. Advanced pipelines estimate light direction and color temperature from the original image and use them either as conditioning inputs or in a post-generation relighting step. Contact shadows, where furniture meets the floor, are especially important for realism. Without them, furniture appears to float.

The final composite merges generated regions with the preserved original image, often with blending at mask boundaries to avoid visible seams.

Stage 6: Upscaling and output

Because diffusion models typically work at lower resolutions than listing photos require, outputs are upscaled with super-resolution models and then sharpened and compressed for delivery. Care is needed to ensure upscaling does not introduce artifacts or alter fine details of preserved regions.

Stage 7: Automated verification

The most important and least discussed part of a responsible staging pipeline is verification. Because generative models can alter property features unintentionally, leading systems compare the staged output with the original image in protected regions. Techniques include:

Masked structural similarity checks, measuring whether pixels outside the edit mask remain effectively unchanged
Edge and line comparison, detecting whether window frames, door openings, ceiling lines or cabinetry edges have moved
Segmentation re-analysis, confirming that the same windows, doors and fixtures are present after staging
Scale plausibility checks, comparing furniture dimensions inferred from depth with typical real-world sizes

Images that fail verification can be regenerated automatically or routed to human review. This step directly addresses the risk that AI staging becomes undisclosed virtual renovation.

Stage 8: Labeling and provenance

Finally, compliant systems attach disclosure labels, such as a “Virtually Staged” watermark, and retain the unaltered original alongside every output. Some platforms are beginning to explore content provenance metadata standards, such as those developed by the Coalition for Content Provenance and Authenticity (C2PA), which record how an image was created and edited in cryptographically verifiable form.

Infrastructure and unit economics

AI staging runs on GPU infrastructure, typically rented from cloud providers. Per-image cost depends on model size, output resolution, number of denoising steps, the number of variations generated per finished image, and utilization of GPU capacity. Vendors manage costs through batching, caching, lower-resolution previews followed by full-quality renders for selected images, and distillation techniques that reduce the number of denoising steps. These efficiencies underpin the sub-dollar per-image pricing available on high-volume plans.

Regulation, Disclosure and Risk

AI’s spread through listing marketing has created a set of legal and reputational risks that brokerages and vendors must manage deliberately. Virtual staging sits at the center of several of them.

Disclosure of altered listing images

California has established the clearest legal standard to date. Assembly Bill 723, signed in October 2025 and effective January 1, 2026, added Section 10140.8 to the state’s Business and Professions Code. In summary, it requires a real estate licensee, or a person acting on the licensee’s behalf, who includes a digitally altered image in advertising or promotional material for the sale of real property to include a disclosure that the image has been altered, and to provide access to the original, unaltered image.

The statute reaches changes made with photo editing software or artificial intelligence that add, remove or change elements of the image, which captures virtual staging, decluttering and virtual renovation. It excludes routine photographic adjustments that do not change how the property is represented, such as lighting, white balance, color correction, cropping and straightening.

The practical implications for the market are significant even outside California:

Workflow design. Systems must retain originals and make them accessible from the listing, not just store them internally.
Labeling. Disclosure must accompany the altered image, which favors in-image labels that travel with the photo when it is syndicated or shared.
Vendor accountability. Brokerages need assurance that staging tools do not alter fixed features silently, because the licensee bears responsibility for what is published.
National precedent. California’s large market and the clarity of the statute make it a likely template for other states, MLSs and listing platforms.

Professional ethics and MLS rules

For REALTORS®, Article 12 of the NAR Code of Ethics requires members to present a true picture in their advertising and other public representations. Staged images that mislead buyers about a property’s condition, size, features or surroundings conflict with that obligation regardless of state law.

Many MLSs maintain their own rules on altered photos, commonly requiring virtually staged images to be labeled and, in some cases, requiring unstaged versions to be included. Rules vary by MLS and change over time, and syndication to portals can add further requirements. For brokerages operating across multiple MLS jurisdictions, a single internal standard set at the strictest common level is often simpler than tracking variations.

Fair housing in AI-driven marketing

The Fair Housing Act prohibits discrimination in housing-related advertising and transactions based on race, color, religion, sex, national origin, familial status and disability. AI introduces fair housing risk in two areas relevant to listing marketing.

Ad targeting and delivery. In April 2024, HUD issued guidance on how the Fair Housing Act applies to housing advertising delivered through digital platforms that use algorithms and AI for targeting. HUD later withdrew that guidance as part of a deregulatory review; a notice published in April 2026 listed it among eight withdrawn documents, effective September 17, 2025. The withdrawal removes HUD’s interpretive guidance but does not change the underlying statute, which continues to prohibit discriminatory advertising. Brokerages using AI-driven ad targeting therefore still carry legal exposure if targeting or delivery produces discriminatory outcomes.

Content generation. Generative AI can produce listing descriptions or images that signal preferences for particular households, for example by describing a property as ideal for a specific family type or by staging imagery with religious symbols or depictions of people. Brokerage policies commonly prohibit people, religious imagery and symbols tied to protected characteristics in staged images, and require review of AI-generated text for exclusionary language.

Consumer protection and misrepresentation

Beyond specific statutes, general consumer protection and misrepresentation law applies to listing marketing. Altered images that conceal defects, misstate the size of rooms, change views or remove neighboring structures can expose agents and brokerages to claims from buyers. The risk is heightened with AI tools because alterations can occur without the user’s awareness. This is why automated verification, human review and retention of originals are increasingly treated as risk controls rather than optional quality measures.

Data, privacy and intellectual property

AI deployments also raise data governance questions:

Client data. Uploading client information into general-purpose AI tools may conflict with privacy obligations or brokerage policies, particularly where data may be retained or used for model training.
Photo rights. Listing photographs are often licensed from photographers under specific terms. Brokerages should confirm that altering or uploading images to third-party AI services is permitted under those licenses.
Model provenance. Generative models trained on large image datasets are the subject of ongoing legal debate over training data rights. Enterprise buyers increasingly ask vendors about training data sources and indemnification.

Operational risk

Finally, operational risks deserve attention: vendor reliability, model updates that change output quality without notice, inconsistent results across staff, and the reputational cost of a single poorly staged listing shared widely on social media. Mature buyers address these through vendor service-level agreements, change notification requirements, approved style libraries and periodic quality audits.

Implementation Playbook

The gap between AI adoption and AI impact in residential real estate is largely an implementation gap. The following playbook reflects common practices among brokerages and teams that treat AI, and virtual staging in particular, as an operational capability rather than an individual experiment.

For brokerages and teams

1. Start with a cost and workflow audit. Identify where money and time are already spent on tasks AI can address: listing copy, social content, photo editing, physical staging, lead response and transaction coordination. Prioritize applications that replace an existing cost and have verifiable outputs. Virtual staging typically qualifies on both counts.

2. Centralize procurement for visual tools. Allowing each agent to choose their own staging tool creates inconsistent quality, unpredictable compliance and no audit trail. A brokerage-approved staging provider, or a short list, allows volume pricing, consistent disclosure labels, approved style libraries and centralized retention of original images.

3. Write a listing media policy. A clear internal policy should define:

Which edits are permitted without disclosure, such as exposure and white balance
Which edits require disclosure, such as virtual staging and decluttering
Which edits are prohibited, such as removing defects, changing views or altering structural features
Labeling requirements and placement
Requirements to retain and provide access to originals
Prohibited content in staged images, such as people, religious symbols and personal items tied to protected characteristics

4. Build review into the workflow. Every staged image should be compared with its original before publication, either by automated verification, human review or both. Review should check windows, doors, flooring, fixtures, views and furniture scale.

5. Stage selectively. NAR’s staging research supports prioritizing the living room, primary bedroom, kitchen and dining areas, and rooms with unclear function. Staging every image adds cost, increases the chance of errors and can make listings feel artificial.

6. Train agents on prompting and judgment. Agent skill strongly influences results. Training should cover photography requirements for good staging, style selection for target buyers, reviewing outputs for scale and alterations, and explaining staging to buyers at showings.

7. Measure outcomes. Brokerages are well placed to measure what individual agents cannot. Useful metrics include listing page views, saves and inquiry rates for staged versus unstaged listings, days on market, showing requests, and cost per listing for marketing media. Comparisons should control for price band, location and property type as far as possible.

A simple ROI framework for virtual staging

A basic return model compares the cost of virtual staging with the alternatives and with plausible improvements in listing performance. For a single vacant listing:

Physical staging cost: median of $1,500 for professional staging, per NAR’s 2025 data, plus potential extension fees if the listing stays on the market longer than the initial term
Human-made virtual staging: about $120 for five rooms at $24 per image
AI virtual staging: a few dollars for five rooms on a subscription plan, plus internal review time

Even if virtual staging delivers only a fraction of physical staging’s benefits, its cost is low enough that a small improvement in buyer engagement or a modest reduction in days on market can justify it. Each additional week on market carries carrying costs for sellers and opportunity costs for agents.

The model breaks down when staging is used to compensate for poor photography or condition problems, when images mislead buyers and create disappointment at showings, or when compliance failures generate legal or reputational costs. Those risks are why governance belongs in the ROI calculation.

For technology vendors

Vendors selling AI virtual staging and related tools into brokerages face rising expectations from enterprise buyers. The following capabilities increasingly differentiate offerings:

Structural preservation guarantees: clear documentation of how the system prevents alteration of windows, doors, flooring and fixtures, supported by automated verification
Compliance features: automatic disclosure labels, original image retention, downloadable before-and-after pairs and audit logs
Brand and style controls: brokerage-approved style libraries, consistent outputs across a listing and the ability to restrict styles or decor elements
Integration: connections to listing management systems, MLS photo uploads, CRM and marketing platforms
Transparency: clear statements of model update policies, data retention, use of customer images for training and training data provenance
Human escalation: access to human editors for complex rooms, decluttering and high-value listings

For listing platforms and MLSs

Portals and MLSs shape industry practice through their rules and display formats. Practices that support trust include displaying staged and original images side by side, supporting standard metadata or labels that identify virtually staged images, and providing clear guidance to members on permitted alterations. As AI-generated images become harder to distinguish from photographs, platform-level labeling and provenance standards will likely become more important than individual agent disclosure alone.

Virtual Staging Use Cases by Segment

The economics and risks of virtual staging differ by property type and business model. The following segment profiles describe typical deployment patterns. They are illustrative rather than based on a single dataset.

Vacant resale homes

Vacant resale listings are the core use case. Empty rooms photograph poorly, lack scale cues and give buyers little sense of function. Virtual staging of the main living area, primary bedroom and dining space is typically the highest-return application, especially in the mid-market where physical staging budgets are limited. The main risk is overstaging, where staged images create expectations the empty home cannot meet at showings. Clear labeling and inclusion of unstaged originals mitigate that risk.

New construction and builder inventory

Builders often hold completed but unfurnished spec homes, or sell from plans before construction is complete. Virtual staging allows builders to present multiple design directions without furnishing model units, and to update marketing quickly as finish packages change. In this segment, virtual renovation also plays a role, visualizing upgrade options. Because these images may show finishes or options a buyer has not yet selected, disclosure that images are illustrative is particularly important.

Occupied homes

Occupied homes present a different problem: the seller’s furniture may be dated, personal or cluttered. AI decluttering followed by restaging can transform listing photos without asking sellers to move out or remove furniture. This is technically the hardest use case, because the model must reconstruct floors and walls hidden behind removed items, and errors can misrepresent condition. Human review is especially valuable here.

Rental and multifamily portfolios

Property managers and multifamily operators list large volumes of units, often with identical floor plans. AI staging allows a single set of staged images per unit type, consistent branding across a portfolio and fast updates when units turn over. The volume economics strongly favor subscription-based AI tools. Fair housing considerations apply equally to rental advertising, so content policies on staged imagery should be consistent across the portfolio.

Luxury properties

Luxury listings present a more nuanced picture. Buyers in this segment often expect high-quality physical staging and in-person experiences, and the gap between a virtually staged image and an empty showing can be more noticeable. Many luxury agents use a hybrid approach: physical staging of key rooms, with virtual staging for secondary spaces or for alternative design concepts that help buyers imagine possibilities. Human-made or human-reviewed staging is more common here because accuracy and finish quality matter more.

Short-term and investment properties

Investors evaluating properties remotely, and owners preparing short-term rental listings, use virtual staging to test furnishing concepts before purchasing furniture. Here staging functions as a planning tool as much as a marketing tool, overlapping with interior design and furniture procurement.

Downsizing and estate sales

The aging of the seller population makes this segment increasingly relevant. NAR’s 2025 profile found the typical seller was 64 years old and had lived in the home a median of 11 years. Long-tenure homes often contain decades of furniture and personal belongings, and estate sales may involve properties being sold by heirs who live elsewhere. Virtual decluttering and staging allow these homes to be marketed without first clearing every room, which can shorten preparation time considerably. The same accuracy concerns apply as with any occupied home: removal of belongings must not conceal wear, damage or deferred maintenance that a buyer would reasonably expect to see. In practice, many agents in this segment combine a partial physical clear-out of key rooms with virtual staging of the rest.

Outlook: 2026 to 2030

Several trends are likely to shape AI in residential real estate, and virtual staging in particular, over the next several years.

1. From text to visual and spatial AI

The first wave of agent AI adoption centered on text: listing descriptions, emails and social captions. NAR’s 2026 survey shows continued interest in content creation, with 54 percent of members planning to use generative AI for content over the next year. But the higher-value frontier is visual and spatial. Image generation, video generation and 3D reconstruction are maturing quickly, and listing marketing is fundamentally visual. Expect more listings to combine AI-staged still images with staged video walkthroughs, interactive style switching and AI-generated floor plans.

2. Interactive and buyer-controlled staging

Today, most virtual staging is chosen by the listing agent and presented to all buyers. A logical next step is buyer-controlled staging, in which portal users switch styles, test their own furniture or visualize renovations directly on listing photos. This shifts staging from a seller marketing expense toward a buyer engagement feature, and it raises new questions about disclosure, since buyer-generated alterations may be displayed alongside official listing media.

3. Spatial capture and digital twins

3D capture through phone-based scanning, lidar-equipped devices and photogrammetry is making accurate room geometry cheaper to obtain. When AI staging is anchored to measured geometry rather than inferred from a single photo, scale errors fall sharply and staging can become a genuine planning tool for buyers. Digital twins of homes could eventually support staging, renovation estimates, insurance and moving logistics from a single captured model.

4. Agentic workflows

The next phase of AI in brokerage operations is likely to involve agentic systems: AI that executes multi-step workflows rather than responding to single prompts. In listing marketing, an agentic system might ingest photos from a shoot, select rooms to stage based on brokerage policy, generate staged versions in an approved style, run verification checks, apply disclosure labels, draft listing copy, and prepare the MLS upload for agent approval. The human role shifts toward supervision, approval and client relationships. This is consistent with Morgan Stanley’s view that operating efficiency through labor savings is the largest near-term opportunity for real estate companies using AI.

5. Regulatory diffusion

California’s AB 723 is unlikely to remain an outlier. Disclosure of altered images aligns with long-standing advertising principles, and the spread of AI image generation makes the issue more visible to consumers and legislators. Brokerages operating in multiple states can reduce future compliance costs by adopting California-style disclosure and original-image access as a national standard now. Provenance standards, such as content credentials that record how an image was produced, may eventually supplement or replace manual labeling.

6. Verification as a competitive differentiator

As generative tools become commoditized, the competitive advantage among virtual staging vendors is likely to shift from generation quality to trust: demonstrable preservation of property features, audit trails, compliance tooling and integration with brokerage systems. Vendors that can show, with evidence, that their outputs do not misrepresent properties will be better positioned with enterprise brokerages and MLSs.

7. Market structure

The virtual staging market today includes specialist AI startups, established photo editing services adding AI, listing photography companies bundling staging, and large portals and brokerage platforms building or acquiring capabilities. As AI reduces the marginal cost of a staged image toward zero, standalone per-image pricing faces pressure. Likely outcomes include bundling of staging into listing media packages, subscription models tied to brokerage seat counts, and consolidation as larger platforms integrate staging into broader marketing suites.

8. Infrastructure and energy awareness

AI image generation runs on GPU infrastructure in data centers, whose electricity demand is growing quickly. The International Energy Agency’s April 2026 report estimated global data center electricity consumption grew 17 percent in 2025 to about 485 TWh, with consumption by AI-focused data centers rising about 50 percent. Virtual staging is a small slice of this demand, and per listing it is very likely far less carbon-intensive than physical staging with its furniture manufacturing and trucking. But enterprise buyers increasingly ask vendors about infrastructure efficiency and energy sourcing, and sustainability disclosures may become part of vendor evaluation.

Risks to the outlook

Several factors could slow adoption or change its direction:

Trust erosion if high-profile cases of misleading AI-altered listings lead to consumer backlash or stricter regulation
Platform rules that restrict or downgrade virtually staged images in search results or listing displays
Legal uncertainty around training data and image rights for generative models
Market conditions, since prolonged low transaction volumes constrain brokerage technology budgets even as they increase the need for efficiency
Agent skills, as the gap between adoption and impact persists without training and process change

Conclusion

AI has become part of the everyday infrastructure of residential real estate. Even the Columbia University did a study about AI in the Real Estate Industry (find it here).

A majority of agents now use it in some form, with 41 percent of REALTORS® reporting current AI use in NAR’s 2025 survey and broader emerging-technology use climbing further in 2026. Industry analysts estimate tens of billions of dollars in potential efficiency gains across real estate by 2030. Yet the data also shows a persistent gap: nearly half of agents surveyed in 2025 had not seen noticeable business impact from AI.

Closing that gap depends less on access to tools than on choosing the right applications and governing them well. Virtual staging illustrates both the opportunity and the discipline required. It addresses the content buyers value most, listing photos, and replaces a cost sellers and agents already pay, at a small fraction of the price of physical staging. The evidence suggests staging reliably helps buyers visualize a home, with 83 percent of buyers’ agents in NAR’s 2025 staging research reporting that effect, even if its impact on price varies.

At the same time, AI virtual staging introduces a new class of risk: the silent alteration of the property itself. California’s disclosure law, professional ethics obligations, fair housing principles and consumer protection standards all converge on the same requirement. Staged images must help buyers understand a property, not misrepresent it.

The brokerages and vendors most likely to benefit from AI in the coming years will be those that combine generative capability with verification, disclosure and measurement. In virtual staging, that means pipelines that preserve architecture, label every altered image, retain every original, and track whether staging actually improves listing performance. Technology has made a staged image almost free to produce. Trust remains the scarce resource, and it will determine which AI applications create lasting value for the home-selling industry.

For brokerage leaders, the practical agenda for the next 12 months is clear: centralize visual AI tools, adopt disclosure standards at least as strict as California’s, build verification into every staging workflow, train agents in both prompting and judgment, and measure listing outcomes rigorously. Firms that do this will turn AI from an individual productivity aid into a durable competitive advantage.