In case you are wondering: Our marketing division loves stock images and I try to keep them away from our GUIs.
I've never thought about exactly WHY we hate stock photos, but I think it's related to the concept of the uncanny valley.
Most cheesy business-centric stock photos look almost real, but there's always just something that makes them clearly unnatural. Is it the perfect mix of skin colors amongst the group? Is it the fact that they seem WAY too happy to be typing on a keyboard or talking on the phone? Is it that they all shop at the same business-casual outfitters?
In other words, a lot stock photography is simply bland. Unless 'bland' is the brand message the company wants to communicate, it's best to avoid the stuff.
Custom photography is ideal, but, to be fair, there are good stock photography resources. However, they tend to cost a bit more and do require the talents of a graphic designer to use them properly.
We tested at Intuit fairly extensively on the use of photos of people on the websites. In every case, the winner of the test was not a person, but a picture of the BOX, like the one you would get in the store.
Executives CONSTANTLY tried to get us to put in happy people and we kept saying, "It won't beat the box. The box is the champ."
The lesson is: People have a mental model of what they want and it is rarely a happy person. Show them what they picture in their head and you will get them to convert.
This answer is predicated on the idea that stock photos are almost always people.
All of the above
Usability tests I have run the experience was once the user became familiar with page layouts (usually after 2-3 page visits and 5 minutes of testing) they quickly became frustrated with stock images "they push content down the page" "they say nothing useful" "they're not relevant".
What I'm trying to say, is that they suck.
This link is to a study: "Photos as Web Content"
The findings of which are: "Users pay close attention to photos and other images that contain relevant information but ignore fluffy pictures used to "jazz up" Web pages."
I believe it's a common sense among UX gurus that stock photos should be avoid (unless there is real need for them), but there are a huge gap between 'sense' and a fact. So, a good way for you to keep your Marketing people away from stocks photos, is showing to them that those photos don't work. A way to do that is making an A/B test.
Even though I am personally against the use of stock images and advocate for solid testing with users, I can see that there might be an argument for using them. This is the case when a particular stock image has been used so much that it becomes a recognized image for something.
The best example I can think of is the picture of a woman with a headset that many sites use as their default support icon or link. Of course, there are other alternatives to this, but I think many first time visitors would instantly recognize this more than anything else (just try searching 'customer support' on ShutterStock and similiar websites).
So in considering the difference between pleasing first time versus returning visitors, you do have to take into consideration what the motivation and rationale for using stock photos are in some of the arguments and testing used to justify their use on websites.