About PropertyIQ
66% of all civil cases in India are related to land or property disputes.** A significant part of this problem arises from transactions where buyers and tenants are unable to clearly understand the legal requirements, costs, records, and contractual risks involved in a property transaction.
One of the challenges is that property laws and requirements vary significantly across Indian states. The rules governing who can purchase property, the permissions required, applicable taxes and stamp duty, land records, units of measurement, and tenancy can all change from one state to another. For someone purchasing property outside their home state, understanding these differences can be difficult, particularly when the relevant information is spread across different laws, government sources, and languages.
PropertyIQ is a cross-state property copilot designed to simplify this process. It helps users understand three fundamental questions before entering into a property transaction:
- Who is legally allowed to buy or rent the property?
- What will the transaction actually cost?
- What legal risks or important terms are hidden in the agreement?
PropertyIQ currently covers Himachal Pradesh, Maharashtra, Karnataka, and Punjab, with support for sale deeds and rent agreements.
The platform begins by identifying the rules that are relevant to the user's specific transaction. Users can select where they currently live, where the property is located, and the type of transaction they are considering. PropertyIQ then highlights the legal and financial requirements that apply to that particular situation, such as purchase restrictions, permissions, applicable rates, concessions, and relevant property records.
The objective is to answer the key questions that arise before a property transaction, in simple language and with references to the underlying law. This allows users to understand the legal position before committing their money or signing an agreement.
How PropertyIQ Works
PropertyIQ is built around a continuously updated legal database that uses authentic legal and government sources. Since property laws and rates can change, the system is designed to regularly refresh its legal information rather than relying on a static collection of rules.
When a user performs an eligibility or property check, PropertyIQ does not simply provide an AI model with a broad property-law question. Instead, it first identifies the specific legal information relevant to that transaction.
For example, the system can take into account:
- the state where the user is located;
- the state where the property is located;
- the type of transaction;
- the user's role in the transaction; and
- the particular legal question being asked.
It then retrieves the relevant provisions from its legal database using a contextual retrieval approach, drawing on techniques such as those described in Anthropic's work on contextual retrieval.
Only this relevant legal context is provided to the AI system. The AI is instructed to base its response on the retrieved material rather than relying on general information or unsupported assumptions.
This approach is important because property law is highly dependent on jurisdiction and context. A general answer about property law may not apply to a particular state, type of property, or transaction. By first retrieving the relevant law and then generating an answer from that material, PropertyIQ aims to reduce irrelevant answers and the risk of unsupported AI-generated information.
Each finding is also accompanied by a source reference, allowing the user to see the legal provision or source on which the answer is based.
Changes during the Vibeathon window
As of 21 August, PropertyIQ had a primitive Eligibility Check and Decode Contract tool. The Eligibility Check relied on a largely hardcoded static database combined with a rule-based engine and an LLM, with the rule engine serving as a fallback when the AI failed, while Decode Contract relied almost entirely on the LLM. Since then, the system has been substantially re-architected around a dynamic legal database that is updated daily using Indian Kanoon and relevant state-specific sources. This database now powers a contextual RAG system using improved chunking to provide the AI with the most relevant legal provisions, significantly reducing hallucinations and improving the quality of analysis. Decode Contract has similarly evolved from basic paragraph-level LLM analysis into a more robust playbook-backed system incorporating applicable legal provisions and contractual best practices, along with redlining and safer alternative clause suggestions. Alongside these substantive improvements, the UI/UX has been redesigned for easier and more intuitive navigation, while the Compare and Toolkit pages have also undergone significant upgrades.