How should I test HTML generated for a web page?

I have some HTML on a page that has a bunch of tables and data (it's a report page). This is all legacy code, so no harassment necessary on the use of tables.

Given that it is legacy code, it is fragile, and we want to confirm that the table looks like we want (number of columns, rows, and the data inside of them are accurate).

My first inclination is to use selenium web driver and run through everything that way (Page Object Pattern), but a co-worker suggested that I just view source of the page, copy the table in question, and then just use this to do a string comparison in the test.

My initial thoughts on his proposal is that it is not a good test because you're starting with the answer and then writing a test to make sure you get that answer (essentially non-TDD). But I'm not sure that's a good enough answer in this case.

How should I test HTML table to make sure all columns, rows are how we like, in addition to the contents of each cell?

Thanks, Sean