Duplicate post checker

Paste the versions of a post you plan to send to different groups. This tells you how similar they really are — and shows you the exact phrases they share, so you know what to rewrite. Runs entirely in your browser.

How the similarity score works

This is standard near-duplicate detection, run locally on your text. It is worth understanding, because the number only means something if you know what it counts.

Each post is normalised — lowercased, punctuation stripped, whitespace collapsed — and then cut into overlapping three-word sequences called shingles. "I help local businesses grow" becomes i help local, help local businesses, local businesses grow. Two posts are then compared by Jaccard similarity: the number of shingles they share, divided by the total number of distinct shingles across both.

Three-word sequences are the useful unit here. Single words match far too easily — any two English sentences share "the" and "a" — while five-word sequences miss posts that are obviously the same but have one word swapped in the middle. Three catches reworded-but-identical text while ignoring incidental overlap. For very short posts the tool drops to two-word sequences automatically, because a forty-word post does not contain enough trigrams to compare meaningfully.

The shared-phrase list uses a different method: a word-level longest-common-run search that finds every sequence of four or more consecutive words appearing in both posts, then keeps the longest non-overlapping ones. That is what turns the score into something you can act on. A percentage tells you there is a problem; a list of the exact phrases tells you which sentences to rewrite.

Reading the score

These bands are our practical guidance, not Facebook policy. Meta does not publish a duplicate-content threshold, and anyone who tells you the exact number at which posts get flagged is guessing. What follows is calibrated on how a reader experiences repetition, which is the thing you can actually observe and the thing that determines whether your posting works.

Why this matters more than the technical question

People usually arrive at this question worried about being flagged. That is a reasonable worry, but it is the second-order problem. The first-order problem is simpler and entirely within your control: in any tight niche, the same people are in several of your groups.

If you post to fifteen local business groups in one city, a substantial share of the members of one are members of three or four others. Those are exactly the engaged, well-connected people you most want to reach — and they are the people who see your message three or four times. Whether the second sighting said "Hi" or "Hey" makes no difference to them. It reads as a broadcast, and broadcasts get scrolled past.

Fix that and the platform-level concern largely takes care of itself, because the behaviour that reads as spam to a person is broadly the behaviour that reads as spam to a system. Optimise for the reader and you do not need to reason about thresholds you cannot see.

How to get your score down

In rough order of how much difference each makes:

1. Change the first sentence completely. The opening line is what most people read before deciding to scroll, and it is the highest-weighted part of any shingle comparison because it sets up everything after it.
2. Lead with a different point. If version one opens with the problem and version two opens with the result, you have two posts. If both open with the problem and differ in the middle, you have one post twice.
3. Change the length. A three-line post and a twelve-line post cannot score as near-duplicates however similar the vocabulary, and length is the difference a skim-reader registers first.
4. Rewrite the phrases the tool highlighted. Mechanical, fast, and it moves the number immediately — but do this after the three above, not instead of them.
5. Change the call to action. Lowest impact on the score, but it varies the part of the post people are most likely to have seen before.