Marketing · Fundamentals
Repurpose one long piece into channel-specific drafts
The research report took three months. It deserves more than one launch email. You need a newsletter section, a week of social posts, a one-pager for sales and a short version for the partner newsletter, and each one is a chance for a number to drift. "73% of respondents" becomes "73% of managers", and nobody notices until a customer does. This guide repurposes the piece channel by channel, with every claim tied to the paragraph it came from.
What AI is for here, and what it isn't
Repurposing is translation between formats. A report becomes a three-paragraph newsletter item, a 200-word social post, a one-page summary for sales. Each format has its own length, structure and reader, and writing all of them by hand is a day's work.
AI is quick at this, and good at fitting a piece to a format's constraints. Asked to keep a claim register, it will also tell you where each claim in each draft came from in the source.
The risk is drift. Every time a claim is shortened, it can change: a qualifier disappears, a sample becomes a population, a "could" becomes a "will". Sometimes the AI adds a claim the source never made, because it sounds like the kind of thing such a report would say. So the method requires a source location for every claim, and a check against the source before anything is published. What's worth saying in each channel, and the final edit, are yours.
Number the source and list the channels
Number the source's paragraphs (P1, P2…) or use its section numbers. This is what makes every claim traceable.
Then write a short spec for each output:
- Channel and reader: "Monthly customer newsletter, existing customers, mostly operations managers."
- Length and format: word or character limits, headline, call to action. Platform limits change, so check each one's current limit rather than relying on memory.
- Purpose: what you want the reader to do.
- Who posts it: your company account, or an employee. This affects disclosure (see Step 4).
Draft with a claim register
Draft one channel at a time, so each gets the attention a short format needs.
Repurpose the source below into the format specified. Use only claims
the source makes.
FORMAT: [channel, reader, length, structure, call to action]
SOURCE (paragraphs numbered): [paste]
PRODUCE:
1. THE DRAFT.
2. CLAIM REGISTER: a table of every factual claim in the draft
(numbers, findings, comparisons, quotes), with:
- the claim as written in the draft
- the source paragraph it comes from
- the source's exact wording
- CHANGED? yes/no: does the draft's wording mean anything
different (scope, qualifier, certainty, population)?
RULES: Don't add statistics, examples or claims that aren't in the
source. Keep qualifiers ("up to", "of respondents", "in our survey").
If the format is too short to carry a qualifier, choose a different
claim instead of dropping the qualifier.
The last rule is the one that matters most in short formats. A finding that needs a qualifier to be true shouldn't appear without it.
Check every claim against the source
Go through each claim register yourself:
- Find each claim in the source at the paragraph cited. If it isn't there, it comes out, however plausible it sounds.
- Compare the wording. Watch for the population ("respondents" vs "managers"), certainty ("may" vs "will"), scope ("in our survey" vs "in the industry"), and rounding.
- Check the "no" answers too. The register's "CHANGED?" column is a prompt for your attention, not a verdict.
Then check across drafts. The same number should appear the same way in every channel.
Edit for the channel
Now edit like an editor: is this the right finding to lead with in this channel? Will this reader care? Is the call to action clear?
If employees will share posts about your company's products or research, they should make their connection clear. In the US, the Federal Trade Commission's Endorsement Guides (revised in 2023) treat an employee's undisclosed connection as something that can make an endorsement misleading. A simple line such as "I work at [company]", or the posts coming from the company account, keeps it straightforward. Check your organization's social media policy.
Worked example: the operations survey
Fictional example. A software company publishes a 2,400-word report from a survey of 412 operations professionals: managers, analysts and directors. The marketer needs a newsletter section, three social posts and a sales one-pager.
The social post draft opens: "73% of operations managers rebuild the same reports by hand every week." The claim register cites P4 and marks it CHANGED: yes. P4 reads: "73% of respondents rebuild at least one recurring report manually each week." Two changes: "respondents" became "operations managers", and "at least one recurring report" became "the same reports". The post becomes: "73% of the 412 operations professionals we surveyed rebuild at least one recurring report by hand every week."
The one-pager draft contains "Automation cuts reporting time in half." The register can't cite a paragraph for it, because the report never says it. It comes out. In its place: the report's actual finding from P9, "Teams using automated reporting spent a median of 3 hours a week on reports, compared with 7 for teams that don't", with the survey caveat in the footnote.
Across all five pieces, the number 73% now appears with the same population every time.
Record the claims
Source: [title], [date], paragraphs numbered P1–P[n]
Editor: [name]
CLAIM REGISTER (all channels)
ID | Claim as written | Channel | Source para | Source wording | Matches? | Checked by
CHANNEL SPECS
[channel] — reader [ ] — limit [checked on date] — posted by [account]
DISCLOSURE
Employee posts include: [disclosure wording]
AI assistance: drafts and register by [tool]; every claim checked
against the source by [name], [date]
What stays with you: what's worth saying in each channel, and the final edit.