Can AI help Ghana catch double land sales before buyers pay?

Ghana's land administration faces a persistent problem: two buyers can each hold documents, receipts and convincing accounts for the same plot, and the conflict often surfaces only when one of them starts building. Writing in a Modern Ghana feature, technology writer Philipa Serwaa asks whether artificial intelligence could detect competing transactions before buyers commit their savings.

The strongest case for AI, she argues, is its ability to compare large volumes of information, including survey plans, earlier transactions and supporting documents, within a connected digital system. Overlapping boundaries could be flagged by comparing the coordinates on a submitted survey plan with existing parcel maps, generating an alert before an application proceeds further.

Serwaa draws a distinction between ordinary mapping tools and AI: geographic information systems can already detect geometric overlaps, while AI could add value by prioritising alerts, identifying patterns across applications and helping officers where older records contain inconsistent descriptions. She cautions that an overlap alone does not prove a double sale, since it could stem from an inaccurate survey, a mapping error, an outdated record or a legitimate transaction affecting an existing parcel.

Duplicate applications present another opportunity, with a system comparing parcel references, coordinates, parties' details and supporting documents, while ownership records could be arranged into a sequence to flag apparent inconsistencies. Such flags, she writes, are reasons to ask questions rather than grounds for declaring someone dishonest, because a difference in names could reflect a spelling error, a marriage or incomplete record-keeping.

The feature stresses that reliable records are the foundation of any credible AI project. Digitising documents does not correct their contents, and a transaction that has never entered the available records leaves nothing to check against, so a result showing no detected inconsistency should never be presented as an unconditional assurance that a purchase is safe.

Serwaa recommends that Ghana begin with a limited pilot using records that have been checked for quality, judged on whether it identifies conflicts officers would otherwise miss, how often it wrongly flags legitimate applications, whether investigations become faster and whether applicants receive clearer explanations. She argues that preventing harm will still require accurate surveys, dependable records, responsible officers and effective action against proven misconduct.