If you have the results, I do not understand the question. Your effect size was large enough to be discovered in the experiment, and you already calculated how much it is.
By the way. Can word count be a non-whole number? If not, the t test is inappropriate, you should rather use a nonparametric method.
I am still wondering about your question. You ran the experiment already and have the results, you detected a difference, meaning your power was enough to detect said difference. To decide whether such an effect size is enough, you must consider reasons of the field of research (measurement error, data in the literature etc.). This is less of a statistical question
Regarding word count, the average can be non-whole, but can you have a single data of 543.65 word count for example? Seems odd to me, please do clarify.