A blockchain does not have to just confirm the timestamp. It can record a consensus of facts about the page as well. E.g. have multiple parties run crawlers and confirm that a majority of them agree about the content of the page to some delta (and include the deltas) at time X.

Do it right and the blockchain can provide trust by ensuring that the record demonstrates that a sufficient number of other parties have confirmed each part of the record.

You have a point in that if an archive is trusted the motivations for doing this largely falls away. The problem is of course that we don't know if the archive will always be trustworthy (e.g. at some point they may accidentally hire someone who is not trustworthy into a position where they are able to do damage), and if/when they're not is when they're likely to be most resistant to putting in place means to prove they are trustworthy.

If "n parties confirm at time X that the page said something" is good enough, then "n parties confirm that the page said something at time X" should also be good enough. If you trust that collection of parties to not have colluded at time X, you should also trust them to not be colluding now.

But this requires the record to be maintained at a sufficient number of those parties, and that's a problem - maintaining records over long periods of time is hard. This is why creating a cryptographically secured record is worthwhile - it does not require the parties to still be around to vouch for the record.

But that doesn't require a blockchain: each party can individually sign it with their own key, and some third party can collect all the records and the signatures.

Now you have to trust that the third party don't omit anything or add entries from other parties.

There are certainly alternative means to guarantee the same properties, depending on your trust model, but they will quickly start acquiring the same complexities.