7 minutes reading time (1401 words)

The Risks and Rewards of Building an AI-Powered Property Search Platform That Relies on External Portals

General

Saturday, 10 May 2025

32 Hits

As artificial intelligence (AI) tools such as ChatGPT, Claude, Gemini and Perplexity continue to evolve, many innovators in the property technology (PropTech) sector are exploring how to use them to deliver a new generation of property search experiences. One attractive idea is to build a platform where AI retrieves listings from existing property portals such as Rightmove, Zoopla or OnTheMarket, and presents them in a more intelligent or conversational format.

However, how feasible and sustainable is it to build a platform that depends on AI retrieving content from third-party property websites?

This article explores the potential advantages, risks, legal considerations and alternative strategies for building a reliable, long-term AI-powered property search platform.

Advantages of Using AI to Retrieve Listings from External Portals

1. Faster Development of a Minimum Viable Product (MVP)
Using AI tools to summarise or extract listing data from publicly available websites could make it possible to launch a functioning prototype quickly, without needing to build or host a large property database. This is particularly appealing for early-stage start-ups.

2. Improved Search Experience

Traditional portals tend to rely on keyword and filter-based searches. AI introduces new capabilities, such as:

Conversational or natural language search
Intelligent filtering based on lifestyle needs or commuting times
Preference learning over time
Auto-flagging of potential issues or opportunities

3. A Way to Bypass Portal Limitations

Some agents and users are dissatisfied with the current limitations of established portals. By using AI to provide a richer user experience, a new platform could stand out by offering more innovative services.

4. Reduced Infrastructure Costs (Temporarily)
By using external sources for listings, you may initially avoid costs related to data storage, licensing, onboarding agents, or maintaining live feeds. This may seem cost-effective in the short term, especially for proof-of-concept platforms.

Risks and Limitations of This Approach

1. Legal Restrictions on Third-Party Content
The listings on portals such as Rightmove and Zoopla are protected by copyright and database rights under the Copyright and Rights in Databases Regulations 1997 (UK). Even if AI can technically access and summarise this content, using it without permission could violate:

Copyright law
Database rights
Terms of service for those portals

This can result in takedown notices, legal claims, and significant reputational damage. Additionally, if you have permission now, it could be withdrawn at any time, putting your business venture at risk.

2. Portals Can and Do Block AI Access
Portals can block AI tools through several methods:

Disallowing crawlers like GPTBot via their robots.txt file
Detecting and blocking bot activity via IP ranges or browser fingerprinting
Preventing scraping through JavaScript rendering or CAPTCHAs
Requiring login or authenticated sessions to access listings

For example, OpenAI's GPTBot respects the robots.txt protocol, meaning if a portal blocks it, ChatGPT cannot use that site's content. Similarly, platforms such as Google Search are introducing meta tags like <meta name="robots" content="noai"> to prevent AI training or reuse.

3. Dependency on Unstable Third-Party Structures
Building your platform on other people's data and site structures is inherently fragile. If the portal changes its layout, restricts access, or discontinues a service, your AI-based solution may break with little warning, and your business could effectively die overnight.

4. Lack of Data Control and Accuracy
You have no control over whether the listings are up-to-date, accurate or compliant. This exposes your platform to risks under the Consumer Protection from Unfair Trading Regulations (2008) and could mislead consumers.

5. User Experience Risks
If your platform presents a listing that has been removed or redirects users to a login page, it undermines trust and creates a poor user experience. You also lose the ability to ensure consistent formatting or presentation of listing data.

Legal Implications

Using AI to reproduce, summarise, or index third-party listings may trigger legal issues related to:

Copyright infringement
Breach of contract (if terms of use are violated)
Violation of database rights
Misrepresentation or misleading omissions under consumer protection law
Data protection and GDPR, especially if personal data is mishandled

Even if the AI "summarises" rather than copies content directly, courts may still consider it a derived or reproduced work depending on context.

Artificial Intelligence Is Still in Its Infancy – And That's a Business Risk

While the hype around artificial intelligence is justified in many respects, it is important to recognise that AI technology, particularly generative AI like ChatGPT, is still in a relatively early stage of maturity. Businesses considering using AI to power core functionality - such as live property search based on third-party listings, must understand the following limitations.

1. AI Models Do Not "Browse the Web" Like a User
Even when tools like ChatGPT or Perplexity appear to search the web, they are often using:

Restricted, cached datasets that may be out of date
Selective access to specific sites that allow AI crawling
Summarisation models that are probabilistic, not deterministic

This means the answers may lack up-to-date accuracy or miss listings entirely, especially if portals block AI access.

2. Data Extraction by AI Is Unreliable and Inconsistent
AI models are not perfect at parsing structured property data from websites. A listing with dynamic content, embedded JavaScript, or non-standard layouts may:

Be misunderstood by the model
Result in missing or incorrect fields (e.g. price, postcode)
Omit critical material information required by law

For a product dealing with legally regulated information like property listings, this level of uncertainty is unacceptable.

3. Large Language Models Are Prone to Hallucination
AI models frequently hallucinate information - that is, they produce facts that sound plausible but are entirely fabricated. This is especially problematic in property, where incorrect information about tenure, leasehold terms, or pricing could:

Breach consumer protection laws
Mislead buyers or sellers
Create financial liability for your business

Unless you're working with verified data from a known source, the output cannot be trusted at face value.

4. AI API Access Terms Are Subject to Change
Many companies building around AI APIs (like OpenAI, Anthropic, or Google) do not have guaranteed long-term access to the models or pricing. Terms of use can change suddenly. If your platform relies entirely on an external AI API to function, you risk:

Increased operating costs
Usage limits or throttling
Revoked access based on policy shifts

You are essentially renting your core technology from a third party that may not share your commercial interests.

5. Regulatory Uncertainty Around AI Use of Web Content
There is currently significant debate in the UK and internationally about whether AI models can legally use web content for training or output generation. Legal frameworks are evolving rapidly. Future AI regulations may:

Prohibit certain use cases (e.g. reproducing third-party listings)
Require content provenance tracking
Limit model capabilities in regulated sectors

Building a business on a shifting legal and ethical foundation exposes you to sudden compliance or commercial shocks.

A More Sustainable Approach: Build on Direct or Open Data

If you want to build a property search platform with AI at its core, consider the following best practices:

1. Secure Listings Directly from Agents

Use a platform model that allows agents to upload listings directly
Integrate with feeds such as RETS, XML or via APIs
Form partnerships or white-label agreements to distribute listings legally

2. Use Public and Open Data Sources There is a wealth of public datasets available in the UK to build contextual property insights:

HM Land Registry: ownership, price paid data
EPC Register: energy performance ratings
Office for National Statistics: crime, income, employment, demographics
Environment Agency: flood zones, contamination risk
Local Authorities: planning applications, conservation areas

3. Enrich, Don't Republish Use AI to enhance property data that you own or have permission to display. For example:

Rewrite listing descriptions
Summarise key information
Auto-generate guides and other useful content
Predict renovation costs or ROI

This builds value without relying on fragile or unauthorised data feeds.

Conclusion: Don't Build a Business on Sand

While building an AI-driven property search experience based on retrieving listings from existing portals may seem like a clever shortcut, it poses substantial legal, technical and commercial risks. It is ultimately not sustainable.

The smarter strategy is to build your platform around data you control or licence, and then use AI to enhance that data - not replace it. This approach is legally compliant, scalable, and offers far greater long-term opportunity.

References