AI & Product Innovation · Global

    AI Announcements vs Real Portal Deployment

    Author: Fouad Bekkar
    Founder, Coraly (GPPI)
    Market commentary - Not investment advice
    Fouad BekkarMarch 21, 20269 min read

    Key Takeaways

    • 1.GPPI tracks AI feature deployment across 68 portal profiles globally. The majority of AI announcements in 2024–2025 have not translated into live, consumer-accessible features.
    • 2.Portals that have shipped concrete AI products fall into two categories: valuation intelligence (Bayut TruEstimate, Zillow AI Zestimate refinements, REA PropTrack) and listing integrity (Bayut TruCheck, Dubai DLD's AI advertising governance system).
    • 3.Portals that have announced AI search or recommendation features have largely not shipped accessible implementations. Natural language search remains a demonstration feature at most portals, not a production-grade consumer tool.
    • 4.For operators benchmarking their own AI roadmap against competitors, the honest picture is that most portals are in early-to-mid deployment phases for consumer-facing AI — with shipped features concentrated in valuation intelligence and listing integrity, not search UX or personalisation.

    What GPPI actually measured

    This analysis is based on GPPI's review of 68 portal profiles — specifically, product signals in the product-technology section of each profile, cross-referenced against publicly accessible product features as of early 2026. GPPI distinguishes between three states: Shipped (live, accessible to consumers or agents), In Beta or Limited Access (deployed to a subset of users), and Announced (press releases, investor presentations, or conference presentations with no publicly accessible implementation found).

    The finding is stark: across the 68 portals GPPI tracks, consumer-facing AI features that are demonstrably live — not announced, not in beta, but accessible to a typical user visiting the portal today — are concentrated in two domains: automated property valuation and listing integrity/verification. Everything else is largely in announcement or early beta state.

    What's actually shipped: valuation intelligence

    Automated valuation models (AVMs) are the clearest example of deployed AI in global portals. Zillow's Zestimate — a consumer-facing home value estimate present on essentially every US residential listing — is the benchmark product in this category, having been live since 2006 and continuously refined with machine learning. Bayut's TruEstimate (UAE), REA Group's PropTrack valuations (Australia), and Zoopla's Hometrack AVM (UK) are the most developed equivalents outside the US.

    The valuation intelligence category is mature because the business case is proven: AVMs drive return visits from homeowners tracking their property value, create a vendor acquisition funnel (homeowners receiving valuation alerts become sellers), and support B2B data licensing to lenders and insurers. The AI is the product, not the marketing.

    What's actually shipped: listing integrity

    The second category of deployed AI is listing verification and integrity — AI systems that classify, flag, or validate property listings at submission. This is less visible to consumers than AVM products but arguably more operationally significant. Bayut's TruCheck verified-listing programme uses an automated verification pipeline to confirm listing availability and authenticity. Dubai's Land Department has deployed an AI-powered advertising governance system that cross-checks portal listings against official permit registries. Abu Dhabi's Madhmoun government MLS requires permit-backed advertising across UAE portals.

    These integrity AI deployments are driven by regulatory pressure in markets where fake and expired listings have historically been a consumer trust problem — most acutely in MENA markets. They represent genuine AI infrastructure investment, not marketing positioning.

    What's announced but not yet shipped

    The largest category in GPPI's analysis is AI features that have been announced — in press releases, conference presentations, or investor communications — but for which GPPI could not identify a live, publicly accessible consumer implementation. This includes: natural language property search ('find me a three-bedroom near good schools with a garden under £500k'); personalised recommendation engines that surface listings based on inferred preferences beyond saved searches; AI-generated listing descriptions (present in agent tools but not in consumer-facing search results); and portfolio analytics for investor users.

    The announcement-to-deployment gap is most visible in search UX. Multiple portals have demoed conversational search interfaces at industry conferences. As of GPPI's 2025 assessment, publicly accessible, production-grade natural language search interfaces — accessible to a typical consumer today without a waitlist or beta signup — were not identifiable across the portals in our database. Portals with relevant deployments are encouraged to submit updates via the correction pathway on their GPPI profile.

    What this means for portal operators

    For operators benchmarking their own AI roadmap against competitors, the picture is more reassuring than the press would suggest: most portals are in early-to-mid deployment phases on consumer-facing AI. The portals that have genuinely shipped AI products have done so in two focused categories — valuation and integrity — where the business case is proven and the data infrastructure requirement is tractable. Operators should resist catch-up investment driven by press announcements and instead focus on the categories where AI is demonstrably delivering consumer value.

    The implication is that operators should resist 'catch-up AI' investment driven by press announcements and instead focus on the categories where AI is demonstrably delivering consumer value: automated valuation models if you have sufficient transaction data, and listing integrity automation if you operate in markets where fake or stale listings are a consumer trust problem. Both require real data infrastructure investment. Neither requires a foundational model.

    Portals with notable shipped AI features in GPPI's database

    The following portals have product signals indicating live, consumer-accessible AI features as of the GPPI 2025 assessment.

    Related Resources

    GPPI 2025 Report

    The full annual benchmark report

    Methodology

    How GPPI measures portal performance

    How to cite GPPI

    GPPI Research. "AI Announcements vs Real Portal Deployment." Coraly GPPI Signals, March 21, 2026. https://coraly.ai/signals/ai/ai-portal-gap-announcements-vs-deployment-2025