AFAIK there is no build-in option to do this.

Yet, since the output of dotplot is a ggplot2-object, it is relatively easy to tweak the resulting plot.

For example, if you would like to have:
a) the color ranging from 0 to 0.5,
b) with a break at 0.25,
c) and using a color-gradient from green-white-red,
d) together with a new label ("adjusted p-values") for the legend,

you could do/add this to p1 (= the 'original' output from dotplot):

p2 <- p1 + scale_fill_gradientn(name = "adjusted p-values",
           colours = c("green", "white", "red"),
           limits= c(0, 0.5), 
           breaks=c(0, 0.25, 0.5) )

Full code chunk:

> library(clusterProfiler)
> library(enrichplot)
> library(ggplot2)
> library(org.Hs.eg.db)
> 
> data(geneList, package="DOSE")
> genes <- names(geneList)[abs(geneList) > 2]
> 
> 
> ego <- enrichGO(gene          = genes,
+                 universe      = names(geneList),
+                 OrgDb         = org.Hs.eg.db,
+                 ont           = "ALL",
+                 pAdjustMethod = "BH",
+                 pvalueCutoff  = 0.01,
+                 qvalueCutoff  = 0.05,
+                 readable      = TRUE)
> 
> p1 <- dotplot(ego)
> print(p1)
> 

> ## customize plot
> p2 <- p1 + scale_fill_gradientn(name = "adjusted p-values",
+            colours = c("green", "white", "red"),
+            limits= c(0, 0.5), 
+            breaks=c(0, 0.25, 0.5) )
Scale for fill is already present.
Adding another scale for fill, which will replace the existing scale.
> print(p2)
>

I think p-values colour scale should be discrete by default (significant or not). In statistics, a p-value is used to reject a null hypothesis or not. If a p-value is smaller than the threshold value, the test is significant because the null hypothesis is rejected. If it is bigger than the threshold, it is not significant. There should not be a colour scale ranging from blue to red with a numeric range such as 0.001 to 0.01 because it encourages people to say "significant" and "very significant", which is not right.

color: variable that used to color enriched terms, e.g. 'pvalue', 'p.adjust' or 'qvalue'

This should have a value isSignif and be the default to be more statistically rigorous.