When I look at scientific papers, bar charts I usually come to see are shown with mean +/- sem (standard error of the mean).
As stated here,
When standard error (SE) bars do not overlap, you cannot be sure that the difference between two means is statistically significant
and
If 95% CI error bars do not overlap, you can be sure the difference is statistically significant (P < 0.05).
Even if the reverse is not true, why don't we always use the confidence interval instead of the sem in bar charts, since it seems more informative ?