Can an algorithm value property?
Automated valuation is fast, but speed is not judgement. Why AI should inform property valuation rather than replace it.
Can an algorithm value property?
Automated valuation models (AVMs) promise faster, cheaper and more consistent property valuations. In some contexts they deliver exactly that. But speed and consistency are not the same as accuracy, and an output is not the same as a judgement.
The question for investors is not whether AI belongs in valuation. It is how it should be used.
The myth of objectivity
A common argument is that algorithms are more objective than people because they apply the same rules every time. In practice, a model is only as neutral as the data it learns from. If historic transactions reflect gaps, distortions or uneven market coverage, the model absorbs those patterns and repeats them at scale.
The result can look consistent while being systematically wrong: liquid, well-documented assets are priced with confidence, while unusual or under-represented assets are mispriced, often without anyone noticing.
More data is not better data
A second assumption is that more data automatically improves accuracy. Real estate does not work like that. Markets are fragmented, illiquid and shaped by local factors and sentiment. Adding more transactions can add more noise, outliers and short-term volatility.
What matters is relevance and quality. A handful of well-understood comparables often says more about a specific asset than thousands of loosely related data points. Models can be disciplined with proper validation, but that requires deliberate design and ongoing testing, not volume alone.
An output is not a valuation
The professional standards are clear on this point. Under the RICS Red Book, an automated output does not constitute a valuation without professional judgement applied to it. In 2025, RICS also published its first global standard on the responsible use of AI in surveying practice, which came into effect in March 2026. It sets expectations for governance, transparency and human oversight.
What these standards define well is what must be achieved: accountability, explainability and objectivity. What they leave open is how firms should do it in day-to-day work.
From principle to practice: co-valuation
The answer is a structured hybrid model, where AI produces evidence and people draw conclusions. In practice, that means three steps:
Screen the inputs. Check data quality, coverage and reliability before trusting any output.
Review and interpret. A professional reviews each result in context, and adjusts or overrides it where local knowledge or lease detail points another way.
Record the reasoning. Keep a clear trail of what the model produced, what changed and why, so every conclusion can be explained and challenged.
This approach also guards against a quieter risk: anchoring. When an algorithm produces a number first, people tend to adjust around it rather than question it. A defined review step forces that question to be asked.
What this means for investors
For investors, the lesson is simple. Use AI to widen the search and speed up screening, but do not treat its output as the answer. The value lies in combining machine scale with human judgement, and in being able to see exactly how a conclusion was reached.
This is the principle AI CoVal is built on. AI screens the market against your mandate. Every result is then reviewed by real estate professionals, and every score shows the reasoning behind it.
This article is for information only and does not constitute financial or valuation advice.
Get early access
Join the waitlist to receive ranked, decision-ready opportunities matched to your mandate.
Get in touch for early access
AI CoVal © 2026. All rights reserved.
Not financial advice. Do your own due diligence.
