Content distribution automation

Content distribution automation through people, not channels

One strong article can work in more places than your blog. This workflow takes a published post or an approved e-book and rewrites it as a first-person article for Medium, Hackernoon and LinkedIn Pulse, then re-voices it for each creator who will publish it under their own name.

It runs in Claude Code, and it is one of the ways one idea becomes several assets. The rules are specific enough that the output reads like the creator wrote it, and strict enough that nothing reaches their profile they wouldn't say.

Why reposting the same article doesn't work

Cross-posting a blog article word for word gives you a duplicate of your own page, written in a company's voice, on platforms that reward a person's. So the house rules for these articles are different from the blog's: open with a result, stay between 1,500 and 2,000 words, and write in the first person as someone who tried the method.

Rewriting by hand for every platform and every creator is where teams give up. Most distribution plans end at a LinkedIn post linking back to the blog.

How a post becomes a creator's article

The source can be a blog post or an e-book, and the steps differ only at the start:

1. For a blog post, fetch the article and extract everything verbatim: headings, body, lists, tables, quotes and links.
2. For an e-book, extract the text from the PDF itself, because the e-book's landing page only lists what's inside, and writing from that list would mean inventing the substance.
3. Rewrite it as a neutral first-person article following the editorial guidelines, and keep that as the base version.
4. Re-voice the base for one creator using their profile: tone, hook styles, their best-performing formats and their hard no's.
5. Run the mechanical check before anyone calls it finished.

The base version is never overwritten. Re-voicing an article for another creator later is a new pass over the base, not a fresh rewrite.

A creator's rules change the article, not just the tone

Every creator fills in a profile, and the most important part of it is the list of things they will not say. Those rules can change what an article argues:

• One creator never recommends LinkedIn as an outreach channel, so any sequence that used it is rebuilt around email and calls
• None of them knock competitors by name, so comparisons are rebuilt around neutral criteria
• Results are framed as one person's outcome on one list, with a suggestion to test on a small sample first
• Advice written for the wrong reader is re-angled, so a tip aimed at job seekers becomes one about sales calls

A persona that only changed the vocabulary would still put words in someone's mouth. The hard no's are what make the output safe to publish under a real name.

Formatting for the platforms

Medium doesn't render markdown tables, so every table becomes a list in one of three shapes: a numbered sequence for timelines, bullets pairing an option with its cost for comparisons, and a mistake followed by its fix. The original article is linked exactly once, in the introduction, and three to five more links from the source are spread one per section.

The check before delivery

A script reads every article and reports its word count, dashes, meta title and description lengths, banned words, and every link with the number of times it appears. The pass conditions are strict: 1,500 to 2,000 words, no em dashes, a title of 60 characters or fewer, and every URL appearing exactly once.

It checks the whole batch as well, because the common failure isn't inside one article. It is one link landing in every article while another lands in none.

Articles that pass go into Google Docs through the same delivery script as content creation automation, with each title prefixed by the creator's name so the folder sorts by author.

Numbers only from the source

An article written from an e-book carries no figure that isn't in the e-book. Where the source has no percentage, the article gets its specificity from named steps, structures and prompts instead of an invented rate.

That rule replaced an older one that tolerated plausible-sounding numbers, and it changed the openings too. They now lead with a structural fact from the source, such as how many parts a demo call has.

A content repurposing workflow for teams with voices

It fits a company whose people already have an audience, or want one, and a library of articles worth more than one appearance. It works best with two or three creators who have filled in a real profile, which is also most of what LinkedIn personal branding asks of a person anyway.

It is the wrong tool for automated posting. Nothing here posts anything: each creator reviews their article and publishes it themselves, or doesn't.

What distribution costs

There are no paid APIs beyond Claude Code itself, and fetching and PDF extraction run locally.

Distribution runs under execution capacity, the AI agents route at a flat $2,000 a month alongside a fractional Head of Content retainer. Allow one to two weeks to start, because each creator's profile and hard no's are written with them, not for them.

Start with one person and one article

Pick the person on your team most likely to publish, and one article of yours that deserves a second life. We'll draft their profile together on a call and run the article through it.