Weed management decision aids have proliferated in recent years, but none of them have been rigorously compared with actual farmer weed management on farm fields. This research compares the Michigan WEEDSIM/GWM bioeconomic model and the CORNHERB and SOYHERB herbicide selection models with farmer weed management in Michigan. In 19 site-years of research in corn and soybean during 1996 to 1997, we found that crop yield, weed control costs, and gross margin over weed control costs (profitability) with the computerized decision aids were not statistically superior to the farmer treatments, even at a one-sided threshold of P = 0.10. In corn the gross margin of the farmer treatment ranked highest in both years. In soybean the gross margin of the SOYHERB treatment ranked highest in 1996 and that of the WEEDSIM/GWM treatment was highest in 1997. Overall, the farmer treatment had the highest gross margin 5½ times, the CORNHERB–SOYHERB treatment 6 times, and the WEEDSIM/GWM treatment 7½ times (where ties were divided equally to give 1/2 to each). However, none of these rank differences corresponded to a statistically significant gross margin gain over the farmer treatment.