This is close to being opinion-based, by here is the summary that should help to make a choice.
Why would you use an ID attribute:
- this is a common and familiar to everybody doing test automation way to locate elements
- this is generally the fastest way to locate elements on a page because selenium gets it down to executing
document.getElementById() which is optimized by the modern browsers (though, usually performance of the end-to-end UI tests is not critical)
- it is a built-in locator in every selenium language binding
- if you would use Firebug or Chrome Developer Tools - the CSS selector and XPath generation tools would generally provide more robust locators using the
ids of the element whenever possible
- you would build shorter CSS selectors and XPath expressions. E.g.
#myid .someclass as opposed to [automation-id=myid] .someclass.
Why would you use a custom attribute:
- if you would add, say,
automation-id attributes to all the desired elements, you would somewhat namespace/scope it to the test automation - everybody would know what is this for just from the attribute name. Meaning, you would dramatically decrease chances of a developer changing the attribute intentionally as opposed to an id attribute, which can and is usually used for application client-side logic as well (reference to this and this answer)
Also, here are some relevant threads:
- Is adding IDs to everything standard practice when using Selenium?
- Which is the best and fastest way to find the element using webdriver? By.XPath or By.ID or anything else? And why?
- Something Better than IDs for Identifying Elements in Selenium Tests
