Early lab experiments suggest AI makes freelancer proposals more alike and harder to tell apart

This matters for online labor platforms because their job is to pair employers with workers who suit each other. If proposals all read alike, employers have less information about what makes each worker distinct, which could weaken the quality of matches even when each individual proposal is clearer.

The work is still in progress. The researchers also report that GenAI affected proposals and the work tasks themselves differently, so a better proposal does not necessarily signal better work.

The researchers define matching effectiveness as a platform's ability to efficiently pair mutually satisfactory relationships between employers and workers. They argue that GenAI-assisted proposals could help by improving clarity and raising the likelihood of a successful match.

They also argue that the same tools can obscure workers' unique attributes. To test both effects, they varied GenAI access in the lab and followed what workers did, what they produced and how employers decided.

The preliminary results fit that two-sided argument: proposals looked better on the surface but were less distinguishable from one another, and AI's effects differed between proposal writing and the work tasks.