My least favorite case for this is "a new poll shows people think X", with no link to the actual poll with the actual questions.
Given the many, many demonstrations of how a wording change can completely reverse poll answers, I'm inclined to view non-quoted, non-sourced poll results as essentially dishonest. It's the sort of thing that will make me think twice about going back to that source for anything I can't cross-reference easily.
It's not like having sources available makes the story worse for others.
They usually announce the source of the poll. Many polls don't have a freely-available link to the details, though you can sometines spend lots of money to buy that access.
I have never understood the business model behind releasing key figures from your poll (generally the crux of it) but gating the rest of it (including polling methods) behind a pricey paywall. Is it to deter the casual lookie-loo while gaining the attention of companies in the industries impacted by the poll results?
Don't know how long they've been doing this, but I noticed recently that the Washington Post is linking to primary sources, and the NYT seems to do it sometimes but not others. So at least some of the big ones are doing better.