In T428311 and T431376, we – staff, in collaboration with experienced volunteers across a range of Wikipedias – will have assumedly determined the initial batch of LLM-generated MoS suggestions to be reliable.

This task tracks the work of running an A/B experiment to evaluate this set of suggestions on newcomers, and the edits they do/do not publish.

Decision(s) to be made

• 1. Learn what (if any) impact LLM-generated MoS suggestions have on newcomers likelihood to publish constructive edits on mweb so that we can decide: Will we enable these suggestions as default-on for newcomers, across projects?

Experiment timing

Early Q2 (FY 2026 - 2027)