In my experience it really depends on the nature of the websites due to the datasets they were trained on. A popular training dataset for summarization models i

In my experience it really depends on the nature of the websites due to the datasets they were trained on. A popular training dataset for summarization models is e.g. news articles, so while they can work well on other news sites, they might not work well on a SDK's docs page and so on.