According to a new study from DataTrace, a national provider of property and ownership data and automation solutions for title searches, public record searches performed solely by artificial intelligence (AI) are not sufficient to ensure reliable title decisions on a large scale. The company's analysis, titled "AI Title Search Tested in the Real World: What Accuracy, Risk and Readiness Really Look Like", compares exclusive AI searches with those supported by 'title plants' to assess completeness, accuracy, and insurability. A title plant is an index of all recorded events affecting a property, characterised by being recorded as the official legal description of the property.
Annette Cotton, Senior Vice President and Chief Data Officer at DataTrace, explained that the study represents a projection of potential risks and the opportunity to look at specific gaps where the AI solution was used alone. She added that the analysis forecasts what the consequences would have been if the property had been insured, for example, and certain gaps could not have been reported, remedied, or resolved.
DataTrace manages over 2,000 title plants nationally. The new study is a continuation of DataTrace's first white paper on its position regarding the value of title plants, according to Cotton. The current investigation was designed to find more detailed answers to these questions and to illuminate the implications of the early findings from the AI test compared to traditional examination.
DataTrace's analysis clarifies that AI performs best when it operates on a foundation of structured, validated title data, rather than fragmented public records at county level. The speed of AI creates value, but only in conjunction with the required reliability, completeness, and accuracy for insurable title decisions. The future does not lie in an opposition between AI and title examiners. It lies in AI powered by trusted title data and guided by experienced title experts. Only in this way can the industry scale automation without sacrificing trust or insurability.
Among the results of the current study is that, in a review of 200 randomly selected residential property files, title searches conducted solely by AI missed at least one significant fact in 40.8 percent of the searchable files. Compared to a title search supported by DataTrace title plant data, these omissions revealed significant accuracy gaps.
- —Of the 200 files evaluated, AI was unable to search 16 files (8 percent) because it lacked the title plant data or comparable normalised datasets to perform the search.
- —The largest gaps occurred in high-risk categories, such as involuntary liens, which showed an error rate of over 36 percent.
