Does AI-generated listing language affect property market outcomes for investors?

Abstract

This study investigates whether residential property listing descriptions continue to function as credible signals in the era of generative artificial intelligence (AI). It examines the linguistic differences between human-authored and AI-generated text and evaluates how these differences relate to three transaction outcomes: sale price, time on market and investor acquisition. The paper is concerned with market pricing dynamics rather than formal valuation in the International Valuation Standards (IVS) or Royal Institution of Chartered Surveyors (RICS) sense, and considers the implications of AI-mediated listing language for the residential investment market and the boundary between owner-occupier and investment-grade stock. The research analysed 6,376 residential sales in South Australia between June 2023 and January 2025, drawn from an initial pool of approximately 10,000 listings. The study employed the DeskLib AI detection tool and computational linguistic analysis to construct continuous measures of AI-likeness. A subset of 5,368 cases was linked to a major rental platform to identify properties acquired by investors within twelve months of sale. Ordinary least squares regressions were used to test the relationship between AI-likeness and sale price and time on market, and a logistic regression was used to examine investor acquisition, controlling for standard property attributes. AI-generated descriptions are systematically longer, more formulaic and more reliant on emphatic phrases compared to human-authored text. AI-likeness is associated with a small but statistically significant 2% reduction in sale price, properties selling approximately 9% faster, and 21% higher odds of investor acquisition within twelve months of sale. These results suggest that listing language is shifting from a costly signal of quality to a coordination device that improves market efficiency, and that the standardised character of AI-generated text appears to particularly resonate with investor buyers who prioritise transactional clarity over narrative distinctiveness. The study is limited to South Australian residential transactions between June 2023 and January 2025. AI detection remains probabilistic and vulnerable to obfuscation and subgroup bias, and the time-on-market model has modest explanatory power. Agents who adopt AI writing tools may also be more technologically sophisticated in other dimensions of marketing, which cannot be ruled out as a contributing factor. The investor flag is conservative, and findings may not extend to commercial markets where lease structures and income-based valuation dominate. Theoretically, the research updates signalling theory by showing that as production costs fall, listing language migrates from costly signal toward coordination device. Generative AI offers efficiency gains, reducing time on market by approximately 9% with only a small effect on sale price. Agencies can use these tools to accelerate workflows but should avoid over-reliance to prevent linguistic homogenisation, supplementing AI outputs with bespoke, locally grounded detail to maintain differentiation. The association between AI-generated listings and higher odds of investor acquisition has implications for portfolio managers and analysts assessing the composition of residential demand, and may signal a quiet shift in the boundary between owner-occupier and investment-grade stock. Platforms may need to innovate via visual and interactive tools as textual variety diminishes. The paper provides the first quantitative test of AI-likeness in relation to residential housing market outcomes, including investor acquisition as a distinct buyer segment relevant to property investment research. It advances signalling theory by demonstrating how the reduced cost of persuasive text erodes signal credibility while enhancing transaction speed through standardisation, and contributes methodologically by treating probabilistic AI detection scores as continuous research features for analysing market communication rather than as binary classifiers of authorship.