At the autumn general assembly of the European real estate valuers association TEGOVA in Vienna, delegates approved a comprehensive position paper governing the deployment of artificial intelligence in property valuation. The umbrella body, which represents approximately 70,000 valuation professionals across 42 countries, drafted the framework in direct cooperation with the Austrian Real Estate Association ÖVI. The document sets a clear professional benchmark by defining strict operational boundaries for generative models and automated valuation tools.
A central pillar of the regulatory position is the classification of appraisers under the European Union AI Act. The association determined that real estate valuers operate strictly as deployers rather than developers. Consequently, full civil, professional and regulatory liability remains anchored to the human expert. Appraisers cannot outsource responsibility for flawed figures to commercial software vendors or generative algorithms under European law.
The urgency of these clear operating rules is reinforced by an empirical survey conducted by TEGOVA across member states. The data reveals that approximately 50 percent of European property valuers already integrate artificial intelligence into their regular workflows. Around 16 percent of respondents reported daily usage, with an overwhelming 96 percent relying on off the shelf consumer tools such as ChatGPT rather than audited, specialized property software.
However, the survey exposed a substantial transparency deficit across the valuation industry. Nearly half of the practitioners who leverage artificial intelligence admitted that they do not disclose this usage within their finished appraisal reports. Industry leaders warned that concealing algorithmic assistance undermines the credibility of reports and invites severe legal repercussions in disputed valuation disputes.
Empirical benchmarks presented at the assembly by Professor Sven Bienert of the IREBS International Real Estate Business School highlighted the dangers of uncritical automation. General purpose large language models fabricate valuation data or market parameters in roughly 10 percent of queries, with error rates spiking in illiquid and niche sectors. Furthermore, automated valuation models currently land within a margin of plus or minus 10 percent of certified human appraisals in only half of all cases.
The newly adopted standard clarifies that algorithmic systems are limited to supporting roles rather than autonomous decision making. While efficiency gains in document parsing and market research remain welcome, certified valuers across Europe must systematically verify all outputs and disclose every digital tool employed.