The Stranger Test for Repurposed Content

Someone waiting for a train opens Instagram and sees a clip from your webinar. It starts mid-sentence: "...and that's why, like I said earlier, the second option usually wins." They never saw the first option. They do not know who is speaking or why this person would know. They scroll on inside three seconds, and the clip goes into your content calendar as "published."

The webinar was good. The clip failed because it was an excerpt, a piece of something else that only made sense next to the rest of it. Most repurposed content fails the same way, and AI makes the problem bigger, because it can produce excerpts faster than anyone can check them.

There is a simple test that catches it before anything goes out, and it is the one habit I would put in place before building any repurposing pipeline, including the one in the free guide to AI content repurposing systems. Call it the stranger test.

The stranger test

Take any piece cut from a longer original: a clip, a quote card, a carousel, a newsletter section, a short post. Show it to someone who has never seen the original. Then ask them one question: what is this saying, and why should I care?

If they can answer in one sentence, the piece passes. If they hesitate, ask what they would need to know first, and that answer is the missing sentence. Most of the time it is small: who is speaking, or what "the second option" refers to. Add it, and the excerpt becomes a piece that can travel on its own.

You do not need a real stranger every time, though a colleague who skipped the webinar is ideal. The discipline is to read the piece as though you had never seen the source, which is hard for whoever made the original, because they cannot unknow the context. That is a good reason to have someone else run the check.

Run it before a piece is scheduled. Once it is live, a weak clip shows up only as a low number in a report, and a low number cannot tell you whether the idea was dull or the reader was lost. A stranger reading the piece in front of you can, because you watch exactly where they stop following.

The words that give an excerpt away

Some phrases almost always mean a piece depends on something the reader cannot see. Search every derivative for them before it goes out.

• "As I mentioned," "like I said," "earlier," "above" and "below."
• "This" or "that" at the start of a piece, pointing at nothing.
• Ordinal references: "the third point," "the second option," "step four."
• In this episode, in the session, on this call.
• An acronym the original defined twenty minutes earlier.
• A chart with no title or no units.
• A quote with no name, or a name with no reason to trust it.

None of these is wrong in the original. They are how people talk and write at length. They become a problem only when the piece is lifted out and asked to stand alone.

Cut at the seams

The deeper fix is choosing where to cut. A long piece has natural seams, places where an idea is complete on its own, and pieces cut along them need very little repair. Look for these.

• A claim with its reason. "Answer the phone before you buy more ads, because ads send people to the phone." Complete in one breath.
• A worked example. A before and after, a sample email, a calculation someone can redo with their own numbers.
• A question somebody actually asked and the answer that followed. Q&A sections of webinars are full of these.
• A mistake and its fix. People share these because they recognize themselves in the mistake.
• A disagreement. The moment the speaker said most people get this wrong, with the reason.

Pieces cut across the seams, like half an argument or the last step of a process, need so much repair that you are better off writing them fresh. When you notice yourself adding three sentences of context to a thirty-second clip, pick a different thirty seconds.

Let the AI do the first rewrite

AI is genuinely good at this repair when you ask for it plainly. The mistake is asking it to "turn this transcript into ten social posts," which invites it to chop and polish. Ask it to write for a stranger instead. A prompt along these lines works:

Rewrite this excerpt so it makes complete sense to someone who has never seen the original. Keep the one main claim. Add only the context the claim depends on, in one sentence. Remove every reference to other parts of the source. If the excerpt cannot stand alone, say so instead of rewriting it.

The last instruction matters most. A tool told it may refuse will flag the excerpts that were cut in the wrong place, which saves you publishing them. Then a person runs the stranger test on what comes back. The AI makes the repair fast, and the person decides whether it worked.

How many pieces is enough

The repurposing guide maps thirty output formats from a single long-form piece, and it is a useful menu. I would treat it as a menu rather than a target. I would bet on a small team that publishes six pieces that pass the stranger test over one that publishes thirty that do not, and the first team will have time left to reply to the comments.

A practical way to choose: count the seams first. If the original has five complete ideas in it, you have about five good pieces, possibly in two formats each. If it has one idea stretched over an hour, you have one good piece and a lot of filler, and no format library will change that.

I will concede one exception. A teaser, the kind of post whose whole job is to send people to the full piece, is allowed to be incomplete. It still has to make its one point completely, though. "Most repurposed clips fail one test: would a stranger understand them? The full checklist is in the episode" works as a teaser. "...which is why the second option wins" does not.

Where the system comes in

Once the stranger test is a habit, a repurposing system has something solid to automate. The free guide covers what repurposes well and what does not, a prompt library for each output type, orchestration options from Make and Zapier to the Claude Agent SDK, a three-tier quality review and how to feed the output into your content calendar. Put the stranger test inside that quality review as the first check, and have someone other than the original author run it.

Before you download anything, try the test on the last thing you repurposed. Pick one piece, send it to a colleague who never saw the original, and ask them the one question.

← Back to all articles