Every site needs images, and most site owners get them from stock libraries without reading the licence attached. This site explains where free and low cost visual assets come from, what the common licences permit, and how to use images without slowing a page to a crawl.
The phrase free stock image covers several different arrangements. Some libraries release work under a public domain dedication, where the creator has waived what rights they can and no attribution is required. Others use a permissive licence that allows commercial use but still forbids reselling the file itself or using a recognisable person in a way that implies endorsement. A few offer a free tier that quietly requires a credit line. The differences matter most in exactly the situation where nobody checks, which is commercial use on a site that later earns money.
Model and property releases are the part most often missed. A licence from a library covers the photographer rights in the image. It does not automatically grant the right to use a recognisable face to advertise a product, or to imply that a person endorses something. Reputable libraries mark which images carry a model release, and that marking is worth reading before an image goes on a landing page.
Technical handling matters as much as licensing. An image straight from a stock library is usually far larger than any web page needs, both in dimensions and in file size. Serving a photograph at several times its displayed size wastes the visitor bandwidth and delays the largest element on the page. The fix is ordinary: resize to the largest size actually displayed, choose a modern format such as WebP or AVIF, and let the browser pick an appropriate variant through width descriptors.
Two attributes prevent most layout problems. Declaring width and height lets the browser reserve the right amount of space before the file arrives, which stops the page jumping as images load. A meaningful alternative text describes what the image conveys, which serves anyone using a screen reader and anyone on a connection where the image never arrives. Decorative images should carry an empty alternative text rather than a filename, because a screen reader announcing a string of digits is worse than silence.
Finally, keep a record of where each image came from and under which licence. A spreadsheet with the source URL, the licence name and the date is enough. It costs a minute per image and it is the only thing that answers the question properly if a rights holder ever asks.
It is also worth deciding when not to use a stock image at all. A generic photograph of people at a desk adds page weight and no information, and readers have learned to skip it. A diagram, a screenshot of the thing being described, or a plain table often communicates more and costs less to serve. Where a decorative image genuinely helps the page breathe, choose one that is small, compress it properly, and give it an empty alternative text so assistive technology passes over it. The goal is a page that loads quickly and reads clearly, and that goal is met more often by using fewer images carefully than by filling every section with something.