From AVM to AI: Three Phases of Innovation Maturity
AI-based real estate valuation encompasses a spectrum of technical approaches ranging from classical automated valuation models (AVMs) augmented with machine learning to fully multimodal deep learning architectures that fuse structured and unstructured data. Patent filings in this dataset span from 2008 to early 2026, revealing a clear three-phase maturation arc — from foundational rule-based systems through ML displacement of those rules, to today’s generative and multimodal frontier.
The Foundational Phase (2008–2016) established the AVM paradigm with limited ML sophistication. CoreLogic Solutions, LLC introduced systems for automated valuation of real estate developments (2013) and a method for creating real estate pricing indicators and predicting trends (2010), anchoring time-consistent AVM valuations aggregated across geographic areas. ZipRealty LLC pioneered user-activity-signal-augmented AVM in 2015, incorporating behavioral signals alongside traditional property attributes.
The Development Phase (2019–2022) saw machine learning models displace rule-based AVMs. Opendoor Labs Inc. filed its first Siamese network AVM in 2022; S&P Global Inc. introduced demographic-augmented property valuation and visualization in 2021; Skyline AI Ltd. disclosed a system for generating value predictions of commercial real estate in 2019. Academic literature from this period explored neural networks, genetic algorithms, and GIS-driven valuation across Cyprus, Armenia, Saudi Arabia, and Ghana — signaling genuine global adoption of ML methodology beyond the US.
The Acceleration Phase (2023–2026) represents the most active innovation stratum in this dataset, with at least 12 filings dated 2023 or later. These reflect integration of large language models, generative AI, blockchain, explainable AI (XAI), and multimodal sensor fusion. According to WIPO, PropTech broadly has seen rapid patent activity growth in this period, consistent with the acceleration observed in this dataset.
This landscape is derived from a targeted set of patent and academic literature records retrieved across focused searches spanning 2008 to early 2026. It represents a snapshot of innovation signals within this dataset only and should not be interpreted as a comprehensive view of the full industry.
The Four Technology Clusters Defining Property AI
Four distinct technology clusters emerge from analysis of the retrieved patent records, each representing a different architectural approach to the AI property valuation problem. These clusters are not mutually exclusive — the most sophisticated 2026 filings draw on elements from all four simultaneously.
Cluster 1: Comparable-Based Siamese and Deep Learning AVMs
This is the most densely populated cluster in the dataset, dominated by Opendoor Labs Inc. The core mechanism uses a trained Siamese (dual-branch) neural network to jointly model a subject property alongside a ranked set of comparable listings, establishing weighted relationships between comparable attributes and value adjustments. Unlike traditional AVM methods that apply fixed depreciation adjustments, the Siamese approach learns adjustment weights end-to-end from training data. Opendoor holds 5 retrieved filings, all centred on this architecture, making it the most concentrated active patent portfolio in the dataset.
Opendoor Labs Inc. holds the most concentrated active AI property valuation patent portfolio in this dataset, with 5 retrieved filings all centred on Siamese neural network AVM architecture — a dual-branch approach that learns comparable property adjustment weights end-to-end from training data rather than applying fixed depreciation rules.
Cluster 2: Ensemble and Hybrid ML Models with Explainability
This cluster encompasses gradient boosting frameworks including XGBoost, LightGBM, and CatBoost, LSTM networks for time-series price trends, and Transformer architectures for multi-horizon prediction across short, medium, and long timeframes. A differentiating signal in recent filings is the integration of explainable AI (XAI) via SHAP values and Integrated Gradients to produce interpretable valuation reports — addressing a critical barrier to adoption in regulated financial contexts. Yangbong Co., Ltd.’s 2025 Korean filing and Christ University’s 2025 Indian filing both explicitly incorporate transparent model structures, indicating that XAI is becoming a first-class technical requirement rather than an afterthought.
Cluster 3: Geospatial and Multimodal Integration
These systems fuse GIS data, satellite imagery, aerial photography, zoning maps, and proximity metrics with structured property records. The most advanced recent filing — Landlogic Solutions Inc.’s multimodal spatio-temporal AVM (WO, 2026) — layers large language models over unstructured narrative descriptions and integrates official planning and GIS data layers. Knowlvers Consulting’s web-based scorecard system computes a Neighborhood Index (NI) and Property Attribute Index (PAI) from geospatial APIs, providing financial credit risk evaluation outputs. According to research published by Nature, geospatial data integration consistently improves predictive accuracy in urban property models.
Cluster 4: Blockchain-Augmented and Decentralized Valuation
A smaller but strategically significant cluster integrates distributed ledger technology to ensure data provenance, immutable transaction records, and decentralized consensus for appraisal reports. Applications include storing valuation outputs as NFT assets for future model training, enabling blockchain consensus among broker nodes, and combining historical transaction data via smart contracts. Doorlight Inc.’s 2025 US filing converts completed property evaluation reports into NFT assets recorded on blockchain for future model training and multi-broker consensus verification — a direct convergence of PropTech and decentralized finance infrastructure.
Map the full AI property valuation patent landscape with PatSnap Eureka’s innovation intelligence tools.
Explore Patent Data in PatSnap Eureka →“Innovation is moderately concentrated: Opendoor Labs and CoreLogic collectively anchor the AVM methodology, but a long tail of smaller assignees and university filers suggests the field is democratizing — particularly in India and the PCT space.”
Geographic Concentration and the Emerging-Market Surge
The United States is the dominant AI property valuation patent jurisdiction, accounting for approximately 30 or more retrieved filings in this dataset — including CoreLogic Solutions, Opendoor Labs, S&P Global, Jones Lang LaSalle, Newmark, Landclan, Wells Fargo, Bank of America, Reali, and Domidocs. However, the geographic distribution is shifting materially, with India and Korea both recording substantial increases in filing volume in the 2023–2026 window.
India is a rapidly growing secondary AI property valuation patent jurisdiction, with filings from Christ University, Bluest Mettle Solutions, Knowlvers Consulting, Proplegit Global, SY Interiors, and Mesbro Technologies — representing at least 8 new filings in the 2023–2026 window alone.
India’s emergence is particularly notable: filings come not only from commercial entities but also from universities such as Christ University (Bengaluru), signaling that academic innovation is feeding directly into the patent pipeline. Korea demonstrates a similarly active trajectory, with Tanker Co., Ltd. (2021), Togetherapps Co., Ltd. (2024), Yangbong Co., Ltd. (2025), and Moon Hye-jeong’s auction risk system (2024) all representing distinct commercial applications of AI valuation.
China shows multiple filings from Ping An Bank Co., Ltd. (AI-based property valuation using weighted model ensemble), Hunan Shujie Technology Co., Ltd. (AI for remote sensing data asset valuation), Chengdu Acalin Technology Development Co., Ltd. (big data real estate value prediction), and Chongqing Huijiyuan Technology Co., Ltd. (commercial real estate AI valuation system). The OECD has noted that China’s overall patent application volume in AI applications has grown significantly since 2019, consistent with the pattern observed here.
PCT (WO) filings from Aicloud Pty Ltd (Australia), Landlogic Solutions Inc., Real Estate Equity Exchange Inc., CoreLogic, and Metrostudy/Zonda Intelligence indicate active international extension strategies. Japan also appears through SoftBank Group Corp.’s 2025 JP filings describing AI systems that learn from historical fluctuation data and future development plans for investment promotion purposes. Standards bodies such as the ISO are also beginning to address data quality requirements for AI-generated valuations — a regulatory development that will further shape IP filing strategies.
India and Korea are the fastest-growing filing jurisdictions in this dataset, with at least 8 and 4 new filings respectively in the 2023–2026 window. These markets represent both high-growth IP battlegrounds and underserved commercial opportunities for international players seeking to establish regional footholds in AI property valuation.
Where AI Valuation Is Being Deployed — Application Domains
AI property valuation technology is being deployed across six distinct application domains, each with different accuracy requirements, data inputs, and regulatory constraints. Residential iBuying represents the largest and most commercially mature domain, while market surveillance and court auction risk analysis represent newer and more specialised applications.
Residential Property Transactions and iBuying. The largest application domain targets residential home buying and selling platforms. Opendoor Labs Inc.’s Siamese AVM portfolio is explicitly designed to support automated offer pricing in iBuying workflows, where real-time valuation accuracy directly determines commercial outcomes. Zonda Intelligence (formerly Metrostudy) targets new residential development pricing, predicting marketed, listing, and closing prices for new builds via its WO, 2024 filing.
Commercial Real Estate and Lease Benchmarking. Jones Lang LaSalle IP, Inc. focuses on commercial lease valuation via its 2024 US filing, training ML models on commercial property data and derived metric datasets. Evalyoo, Inc. applies machine learning to commercial real estate evaluation, valuation, and recommendation — including client-profile-weighted scoring in its 2021 US filing.
Financial Services: Lending, Mortgage, and Credit Risk. S&P Global Inc.’s property valuation model processes historic financial transaction data, characteristic data, demographic data, and unstructured market sentiment to support net asset value computation for real estate portfolios relevant to lending and investment management. Wells Fargo Bank, N.A. addresses the appraisal component of mortgage underwriting via its Virtual Property Appraisals filing (US, 2021). These applications intersect directly with financial regulation frameworks tracked by BIS.
Real Estate Investment and Portfolio Management. Repan Co., Ltd.’s WO, 2020 filing uses recurrent neural networks with individualized weighting for portfolio future value prediction. Realpha Tech Corp evaluates investment ranking across more than 25 factors in its 2023 US filing. SoftBank Group Corp.’s 2025 JP filings describe AI systems that combine historical fluctuation data with future development plans for investment promotion.
Market Surveillance and Bubble Detection. Newmark & Company Real Estate, Inc. applies big data distributed node analysis to real estate bubble prediction in its 2023 US filing, identifying historical peaks and generating predictive alerts for regulators and institutional investors.
Auction and Court Sale Risk Analysis. Moon Hye-jeong’s 2024 KR filing describes an AI-based system providing risk and value analysis for court auctions and public auctions using semantic network analysis of real estate appraisal reports — processing appraisal report text for automated auction risk scoring. This application domain is unique to the Korean dataset and represents a jurisdiction-specific innovation direction.
Yangbong Co., Ltd.’s 2025 Korean hybrid AI property valuation system predicts real estate prices at 3-month, 1-year, and 3-year horizons simultaneously, using over 50 structural and policy variables — representing a maturation from point-in-time automated valuation models to dynamic multi-horizon investment planning tools.
Five Emerging Directions Shaping the 2026 Frontier
Among retrieved filings dated 2024–2026, five distinct forward-looking innovation signals are apparent — each representing a meaningful departure from the prior generation of AVM systems and carrying strategic implications for IP positioning.
1. LLM and Generative AI Integration
The most significant emerging signal is the incorporation of large language models into valuation pipelines. Landlogic Solutions Inc.’s multimodal spatio-temporal valuation system (WO, 2026) explicitly employs LLMs to extract structured features from unstructured property descriptions and planning documents. AIVRE, Inc.’s real estate appraisal system (US, 2026) uses third-party generative AI tools to generate and populate comparable property appraisal data. Catalano’s 2025 US filing feeds LLM systems with sensor-derived visitor behavioral data from property tours — a sensor-to-language pipeline not previously observed in the dataset.
2. Explainable AI (XAI) for Regulatory Compliance
Yangbong Co., Ltd.’s 2025 Korean filing explicitly uses SHAP (SHapley Additive exPlanations) and Integrated Gradients to produce interpretable prediction reports, directly addressing regulatory and institutional demands for audit trails in financial valuation. Christ University’s 2025 Indian filing similarly emphasizes transparent AI via ensemble modeling with feedback loops. IP strategists should consider XAI components as essential claim elements for filings targeting financial services regulatory environments.
3. Hyperlocal and Satellite-Augmented Valuation
Christ University’s 2025 filing ingests satellite imagery alongside government databases, taxation records, and social media signals for hyperlocal Indian market valuations. Hunan Shujie Technology Co., Ltd. pursues AI-based remote sensing data asset valuation and pricing in two 2025 CN filings, with explicit task-utility framing for financial and insurance decision contexts. The integration of satellite and remote sensing data into valuation pipelines represents an important expansion of the feature space available to AVM systems.
4. Blockchain-NFT and Decentralized Consensus for Report Integrity
Doorlight Inc.’s 2025 US filing converts completed property evaluation reports into NFT assets recorded on blockchain for future model training and multi-broker consensus verification. Trete Inc.’s 2025 US filing uses AI modules for asynchronous valuation data collection with unique identifier matching. Early movers establishing IP in AI-generated appraisal records anchored to distributed ledgers may find defensible positions as regulators increase scrutiny of AI-generated financial assessments.
5. Spatio-Temporal and Multi-Horizon Price Forecasting
SoftBank Group Corp.’s 2025 JP filings describe AI systems that learn from historical fluctuation data and future development plans for investment promotion. Yangbong’s hybrid AI system predicts prices at 3-month, 1-year, and 3-year horizons in parallel, using over 50 structural and policy variables. This multi-horizon forecasting capability represents a maturation beyond point-in-time AVM to dynamic investment planning tools — and a significant expansion of the scope of AI property valuation claims.
“Multimodal data fusion — combining satellite imagery, GIS layers, LLM-parsed text, and behavioral sensor data — is the architectural frontier. The most technically ambitious 2026 filings integrate all of these modalities simultaneously.”
Track emerging AI property valuation patent signals as they file — PatSnap Eureka monitors the global landscape in real time.
Monitor with PatSnap Eureka →Strategic Implications for IP and R&D Teams
The patent signals in this dataset translate into concrete strategic considerations for R&D teams, IP counsel, and commercial leaders in the property technology and financial services sectors.
Design around or license the Siamese AVM architecture. Siamese network and comparable-weighting architectures have become the dominant residential AVM paradigm, with Opendoor Labs holding a concentrated active patent portfolio. Entrants to the residential AVM space must design around or license this architecture, or develop differentiated comparable-weighting mechanisms that fall outside existing claim scope.
Treat explainability as a first-class technical requirement. Filings from Korea (2025) and India (2025) explicitly incorporate SHAP and transparent model structures. Given the trajectory of financial regulation — including frameworks from BIS on AI in banking — XAI components should be considered essential claim elements for any filing targeting financial services or regulated lending environments.
Prioritize India and Korea as IP filing jurisdictions. India and Korea are the fastest-growing filing jurisdictions in this dataset, with at least 8 and 4 new filings respectively in the 2023–2026 window. These markets represent both high-growth IP battlegrounds and underserved commercial opportunities for international players seeking to establish regional footholds.
Invest in multimodal data pipeline IP. Multimodal data fusion — combining satellite imagery, GIS layers, LLM-parsed text, and behavioral sensor data — is the architectural frontier. The most technically ambitious 2026 filings integrate all of these modalities. R&D teams should prioritize data pipeline and feature engineering IP that generalizes across modalities, as this is where valuation accuracy differentiation will reside.
Establish early positions in blockchain-anchored appraisal records. Blockchain-NFT mechanisms for valuation report immutability and audit trails represent a convergence opportunity between PropTech and decentralized finance infrastructure. Early movers establishing IP in AI-generated appraisal records anchored to distributed ledgers may find defensible positions as regulators increase scrutiny of AI-generated financial assessments.
For IP professionals tracking this space, PatSnap’s IP intelligence platform provides citation network analysis and claim-level comparison across the AVM and multimodal valuation patent clusters described here. The innovation trend monitoring capabilities allow R&D teams to track emerging filing jurisdictions — including the India and Korea acceleration identified in this dataset — as they evolve in real time.