What makes API testing different from web-page testing — understanding the contract, testing authorization on every endpoint, and why broken object-level access is so common.
APIs power most modern applications, and they fail in ways a browser-driven test often misses. Testing them well means thinking about the API's contract directly — the endpoints, the parameters, and above all who is allowed to do what.
Find the API's documentation or specification (such as an OpenAPI/Swagger definition) if it exists, and map the endpoints, methods, and expected parameters. If there is no spec, reconstruct one from observed traffic. You are testing the contract, so you need to know what it claims to be.
The most common and impactful API issues are authorization failures. The classic is broken object-level authorization: an endpoint that returns object 123 for you will happily return object 124 belonging to someone else, because it checks that you are logged in but not that the object is yours.
Rate limiting, input validation, and information leakage in verbose errors all matter for APIs too. But if you only have time for one thing, test authorization on every endpoint.