As it exists today, trendline will display it's output even when there are monstrous outliers / high levered data points / abnormal Cook's distances which make the trendline meaningless from a statistical perspective. While I agree that hiding the trendline by default would be cumbersome for many users (who may want the trendline even though it shouldn't be used), it might be more subtle to color the trendline red by default, or make it a dotted line, when it is statistically invalid. Perhaps this could occur by default whenever there are outliers, highly levered points, the regression slope is not statistically significant, the slope is less than the standard error, the residuals are heteroskedastic, etc.