Content repurposing rules that stop a portfolio filling up with thin duplicates

An editor opens the content calendar for a group of six shows and finds the same 700 word preview scheduled four times in October, once per brand, with the venue name and the date swapped and one paragraph moved. Somebody wrote content repurposing rules for this portfolio two years ago. The rules said reuse wherever possible, and the team did exactly that.

The output looks productive. Four pages, four brands, one writer's afternoon. Six months later the same four pages are competing for the same searches, none of them ranks, and the archive has 300 more like them.

What repurposing turns into without a gate

Reuse is a good instinct badly served by most workflows. A publishing system makes creating a page cheap and makes deciding not to create one impossible, because there is no button for that. So the decision defaults to publish, every time, and the archive grows by the size of the calendar regardless of whether anything new was said.

Google's spam policies name the version of this that gets punished. The policy defines scaled content abuse as "when many pages are generated for the primary purpose of manipulating search rankings and not helping users", and lists among its examples "republishing content from other sites without adding any original content or value" and taking content and modifying it "only slightly (for example, by substituting synonyms or using automated techniques)" before republishing.

Most event portfolios are not doing that on purpose. They are doing it by accident, at low volume, with human writers, and the accidental version produces the same archive. The difference is that a deliberate spammer knows how many pages they made and an organiser usually does not.

A test that has to be passed before publication

The gate belongs before the page exists, and it works best as four questions with written answers, filed with the piece.

• Is the audience different? A preview aimed at exhibitors and a preview aimed at visitors are different pieces. The same preview posted to a second brand whose readers overlap with the first is one piece published twice.
• Is the format different? A session write up, a data table drawn from the same research, and a video explainer are genuinely different objects. A 700 word article and a 650 word article are the same object.
• Is enough of the material new? Set a share and hold it. Forty per cent is where I would start, measured on the body text.
• Does it target a query no existing page owns? If the answer is no, publishing it creates a competitor for a page you already have.

A piece that clears all four gets a page. A piece that fails any one of them gets a link from the original, or a new section inside the original, and the calendar slot goes to something else.

That last clause is the part teams skip, and it is what makes the rule survive contact with a content plan. A gate that only says no creates a hole in the schedule and gets overruled by the second week. A gate that redirects the effort into deepening an existing page keeps the writer busy and the archive small.

How do you know whether 40 per cent is new?

For a single piece at the point of publication, a reader can judge it. Put the two documents side by side, mark the paragraphs that carry information absent from the original, and count words. It takes five minutes and it is accurate enough for a publish or link decision.

The question that needs a machine is the portfolio one: how much of everything already published overlaps with everything else. That is a different job with a different method, and measuring duplication across the whole corpus belongs in its own workflow rather than in an editor's pre publication checklist.

Keep the two separate deliberately. The gate is a judgement made once, quickly, by the person who wrote the piece. The corpus measurement is a monthly number that tells you whether the gate is working.

Scoring twenty recent reposts

The way to find out whether your rules mean anything is to grade what you have already published against them.

Take the twenty most recent pieces that anybody would describe as repurposed. For each, find the original, mark the new material, and answer the four questions. This is a two hour exercise for one person.

Suppose it comes back like this. Six of the twenty pass all four. Five pass three, usually failing on new material. Nine pass two or fewer, and of those, seven have a new material share under 15 per cent, with a median across the whole failing group of 18 per cent.

That is 45 per cent of the sample, nine of twenty, that should have been a link. Extrapolated across a portfolio that publishes, say, 40 repurposed pieces a quarter, you are creating roughly 18 pages a quarter that dilute pages you already have, which is 72 a year.

The number is worth having because it converts a style argument into a volume. Nobody wins an argument about whether a preview is thin. Everybody understands 72 pages a year.

Two details make the scoring survive review. Mark the new material before you count it, in the document, so the judgement is visible to somebody who disagrees. And record which of the four questions each piece failed, because the pattern tells you what to fix. A sample that fails mostly on new material has a research problem and needs writers given more time per piece. A sample that fails mostly on the query question has a planning problem and needs the calendar checked against the existing archive before commissioning, which is a five minute step nobody currently owns.

Run the same twenty against the previous year if the archive supports it. A failing share that is stable tells you the rule was never applied. A failing share that is rising usually tracks a headcount cut, because the first thing a shrinking team does is reuse more and research less, and that decision is rarely made explicitly by anyone.

Why the test is about the reader first

The search argument is the one that gets budget, and it is the weaker of the two.

Google's guidance on creating helpful content asks publishers to self assess with questions including "Does the content provide original information, reporting, research, or analysis?" and "Does the content provide substantial value when compared to other pages in search results?" and, most directly, whether content that draws on other sources avoids "simply copying or rewriting those sources, and instead provide substantial additional value and originality".

Those questions describe a reader's experience before they describe a ranking system. A visitor who follows a link from a newsletter to a piece they have already read on a sibling site does not file a complaint. They stop opening the newsletter. That cost lands months later in a channel report and gets attributed to subject lines.

The four question gate is worth running even in a portfolio that has no search ambition at all, which is the test of whether a rule is about quality or about rankings.

What to do with the pieces that fail

Three destinations, and picking the right one matters more than the gate itself.

If the failing piece is genuinely the same page in two places, the pair needs one URL and a decision about which one survives, which is a canonical and redirect question with an answer that depends on how similar the two really are.

If the failing piece is a lightly different take that a second brand's audience does want, the honest move is a short editorial note on the second brand linking to the first, with the second brand's own framing in the note. That is the syndication pattern between sibling titles, and it works when the note is written for the second audience and fails when it is a stub.

If the failing piece has new material that did not fit anywhere, put the new material into the original and leave the piece unpublished. The writer's work is not wasted, the original gets better, and the archive does not grow.

Where this stops

The 40 per cent threshold is arbitrary and I have no source that says otherwise. It is a coordination device: a number everyone applies the same way beats a better number applied inconsistently. Anybody who tells you the correct figure is 30 or 55 is guessing too, and the useful response is to pick one, run it for two quarters, and see whether the pieces it rejected turn out to have been worth publishing.

The bigger limit is that the test measures difference and not quality. A piece can be 90 per cent new material, target an unclaimed query, reach a new audience in a new format, and still be worthless. The gate stops you filling the archive with duplicates and does nothing at all about filling it with original filler, which is a harder problem and one that no rule solves.

There is also a case the rule handles badly, which is deliberate annual repetition. A show preview genuinely should exist for each edition, and each one is 80 per cent the same as last year's. Handle those as a separate class with their own URL policy for editions rather than forcing them through a test designed for one off pieces.

Pull the last twenty repurposed pieces you published, answer the four questions for each, and count how many would have been a link. Do that before you write any rules down, because the count decides whether you need a gate or a nudge, and it gives whoever owns the editorial operation a number to argue from.