After the fact (and limiting the answer to viable technical solutions), there is probably no way to prove you wrote something. I will assume the content is yet to be published.
There are two parts to a reasonable "proof", the first being association of the content with
a digital identity, i.e. signing the content so that it can be verified with a recognisable and easily available public key. The second (which may not be required, depending on how paranoid you are) is proving you created it first (i.e. to prove or disprove plagiarism).
A digital signature provides integrity and authenticity (proof of sender). Proof of authorship (related to non-repudiation) is a more difficult problem to solve.
The simplest way then is with a public key solution, as you suggest: before you publish you get a trusted third party who provides a secure timestamping service to timestamp and sign a digest of your signed content. The timestamping service will use X.509 certs, you are free to use PGP or X.509 to sign your content here.
Your "provable" published data should be: the content (signed by you) and signature (if not inline), your public key or its location, the digest method, and the third party's timestamping signature (of the digest).
This question on signing code commits pretty much covers the technical details:
A possible problem here is how to represent your content in a canonical format so that others can verify the signature. You should consider signing the content in a format which can be downloaded "intact" and not subject to unintended reformatting or alteration by the server side blog software. Offering a PDF version of the article would be a solution, and you also have the option of using an X.509 signature and timestamp within the PDF itself (PDF).
If you're old-school, there is a free 21st century equivalent to posting yourself a copy by registered mail, which may or may not be useful, and probably has about as much legal weight (i.e. negligible):