We present the longest record of surface afterslip on a continental strike‐slip fault for the 2004 M 6.0 Parkfield, California, earthquake, from which we can derive critical information about the duration and predictability of afterslip relevant to urban displacement hazard applications. Surface slip associated with this event occurred entirely postseismically along the interseismically creeping (0.6–1.5  cm/yr) main trace of the San Andreas fault. Using the first year of afterslip data, the program AFTER correctly predicted the cumulative surface afterslip (maximum ∼35  cm) eventually attained. By 1 yr postearthquake, observed afterslip had accumulated to only ∼74% of its modeled final value u_f in units of length. The 6‐yr data suggested final slip would be reached everywhere by ∼6–12  yrs.

Parkfield’s afterslip lasted much longer (∼6–12  yrs) than afterslip following a 2014 M 6.0 event in Napa, California, where no interseismic creep was known, and its afterslip neared completion (∼97% of u_f) by 1 yr. The uncertainty in u_f for the Napa event fell to ≤2  cm in only three months, versus in 2 yrs for the Parkfield event, mostly because duration of the power‐law stage of afterslip at Parkfield is much longer, ∼1000 (493–1666) days versus ∼100 (35–421) days for Napa. Because the urban Hayward fault near San Francisco, California, like the Parkfield section, exhibits interseismic creep in a similar geological regime, significant afterslip might last for up to a decade following an anticipated M≥6.7 earthquake, potentially delaying postearthquake recovery.