You hit publish at 8:04am. At 8:09 you notice the number in your hook is wrong — you wrote 40% when your own spreadsheet says 14%. The post already has three comments.
Now what?
Every founder we work with has had this exact five minutes. And the answer they find when they search for it is one of the most confidently wrong things on the internet about the LinkedIn algorithm. There is a widely-circulated decay table that tells you exactly how much reach an edit costs, bracketed by minute. It is not evidence. We're going to take it apart, tell you what's actually defensible, and then give you the decision framework — which, usefully, doesn't depend on winning the argument.
The table everyone quotes, and why we don't repeat it
If you search this question you will land on some version of the following: edit within 10 minutes and you're fine, 10–30 minutes costs you 10–15% of impressions, 30–90 minutes costs 30–40%, after 90 minutes you're down 40–50%. Sometimes there's a second table breaking it out by edit type — typo fixes safe, content rewrites −38%, adding a link −42%, hashtag changes −12–18%.
It reads like data. Here's what's underneath it.
The sample is 340 posts across 52 accounts, published by a company that sells LinkedIn tooling. That's roughly six and a half posts per account. To detect a 12% difference in impressions — against a metric that routinely swings 3x between two posts from the same account on the same topic in the same week — you would need vastly more than that, and you'd need a control for topic, format, time, and account baseline. Nothing in the methodology suggests any of that happened.
Then there's the part that should end the conversation. That table typically cites Richard van der Blom's Algorithm Insights research for its mechanism — the "60–90 minute momentum window." Van der Blom's work is the real thing in this space; the 2026 edition analyzed roughly 1.3 million posts across 50,000 creators. We went looking for the edited-post finding in it.
We can't locate it. Every published summary of that research covers the same ground — reach decline, format performance, link penalties, follower growth. None of them mention edited posts. If the finding is in there, nobody citing it has ever pointed to where. A real researcher's name is doing authority work for a claim nobody can source, and it's been repeated downstream until it acquired the texture of established fact.
We're not saying editing is free. We're saying nobody has published credible evidence of what it costs, and a number invented to fill that gap is worse than admitting the gap. If you make posting decisions off that table, you are making them off a blog post.
What actually is defensible
Strip out the folklore and a few things survive, because they're supported by large samples or by LinkedIn's own product behavior.
A first-hour test pool exists. Your post goes to a limited slice of your audience first, and how that slice responds shapes what happens next. We've written about this mechanic before and it isn't controversial.
Reach is genuinely down and the bar is higher. Van der Blom's 2026 figures put average post reach at 8–12% of followers, down from 15–20% a year prior, with active creators seeing roughly 60% less reach over two years. Whatever an edit costs, it's happening on top of a squeeze that's already on.
External links in the post body cost reach — measurably. The same research puts one link in the body at roughly −18.8% median reach. Note what this does to the "adding a link post-publish costs you 42%" claim: the better-sourced finding is that the link itself is expensive, whenever it arrives. The edit isn't the villain in that scenario. The link is.
Two ledgers, not one. A post gets graded. Your account also carries a baseline that consistency raises and gaps lower. This matters enormously for the delete-and-repost question, and it's the thing the decay tables completely ignore.
That's the honest inventory. Notice there's no minute-by-minute penalty curve in it, because nobody has credibly produced one.
The three options, priced properly
You have exactly three moves. They have genuinely different cost profiles, and only one of them is well understood.
Leave it. Cost: whatever the error costs you in credibility, permanently, on the highest-traffic asset you'll publish this week. Zero algorithmic risk. This is correct far more often than founders think — see the triage below.
Edit in place. Cost: unknown, plausibly small, and it does not touch your account baseline. The post keeps its comments, its reactions, its URL, and its existing distribution. Worst realistic case is that it disrupts momentum on a post that was mid-flight. Best case is nothing happens. The honest position is that this is a low-cost move with an unquantified downside — which is a very different thing from a 40% penalty.
Delete and repost. This is the one founders reach for and the one we push back on hardest. You lose every reaction and comment — including the early engagement that was doing work for you. You lose the URL, so anyone who sent it in a DM sent a dead link. And unlike editing, you're now asking the account-level ledger to absorb a second post on the same topic within a short window — the same day, in the same lane, to substantially the same audience. That's a real cost, it's paid at the account level rather than the post level, and it's the one nobody prices.
Delete-and-repost feels decisive. It is almost always the most expensive option available.
The triage: decide by what's broken, not by the clock
The decay tables ask when did you notice. That's the wrong first question. The right one is what's wrong, because severity determines the move and the clock only breaks ties.
Factual error in a claim — fix it, always, immediately. A wrong number, a misattributed quote, a stat you got backwards. Fix it the second you see it, at any hour, and don't think about reach for even a moment. Your reputation is the asset; the post is inventory. An operator who reads your wrong number, acts on it, and finds out later has learned something about you that no amount of reach recovers. This is not an algorithm decision.
Broken or wrong link — fix it. A dead link converts at zero. There is no version of leaving it that wins.
Typo, one wrong word, a missing line break — leave it, mostly. Nobody has ever declined to hire an operator over a missing comma. If it's genuinely invisible to a reader who isn't you, the edit is serving your discomfort, not the post. Founders edit far more often for anxiety than for the reader, and the tell is that you're the only person who's noticed.
Wrong hook, weak opening, "I should have written this differently" — leave it and write the better version later. This is the case where delete-and-repost is most tempting and least justified. You don't have a defect, you have a preference that arrived late. Let it run, watch what it does, and put the improved framing into a post in three weeks — where it gets a fresh test pool instead of asking the same audience to read your second draft.
Genuinely wrong post — wrong client named, confidential number, something that shouldn't be public. Delete it. Don't repost. Don't replace it. Take the account-level cost and move on, because the alternative is worse than any reach question.
The pattern: edit for accuracy, leave for aesthetics. That rule needs no decay table to work, which is exactly why we like it.
The real fix is upstream
Here's the uncomfortable part. Every founder asking this question is asking it at the wrong point in the process.
The edit-versus-repost decision is a symptom of publishing without a gate. Founders who run a pre-publish pass — read it cold, check every number against its source, click every link, look at it on a phone — hit this decision a handful of times a year. Founders who write and publish in the same motion hit it weekly, and spend real cognitive budget on a question that shouldn't recur.
Four checks, ninety seconds, before you press publish:
- Every number traced to where it came from. Not remembered. Traced. This single check eliminates the only edit that's genuinely mandatory.
- Every link clicked. In the actual post preview, not in your draft.
- Read the first three lines cold, as a stranger. This is where hook regret comes from, and it's cheap to catch before publishing and impossible to fix after without cost.
- Look at it on your phone. Formatting breaks that are invisible on desktop are the second-most-common reason founders reach for the edit button.
Ninety seconds beats any recovery move, because the best-case outcome of every option in the previous section is you get back to where you would have been anyway.
FAQ
Does LinkedIn officially penalize edits? LinkedIn has never confirmed an edit penalty. That's not proof there's no cost — platforms confirm very little — but it does mean anyone quoting you a specific percentage is quoting something they made up or something someone else made up. Be as skeptical of the confident number as you are of the platform's silence.
So is it safe to edit or not? Editing in place is a low-cost move with an unquantified downside, and it doesn't touch your account baseline. Delete-and-repost is a high-cost move with a well-understood downside. If you must do something, edit.
What about editing hours or days later, once the post is done running? This is the case with the least plausible risk and the least benefit. Distribution is largely over, so there's little momentum left to disrupt — and also little audience left to serve. Fix factual errors for the record and the archive. Leave everything else.
Can I repost a good post later? Yes, and that's a different question with a better answer. A genuinely strong post reworked and republished months later, to a partly-turned-over audience, is a legitimate move. Deleting and immediately republishing the same post to the same people is not the same thing.
My post is doing well but has a small error. Do I risk it? If the error is factual, fix it — a post that's reaching more people is a post spreading the error further, which makes fixing it more urgent, not less. If it's cosmetic, leave it and enjoy the run.
The reason this question generates so much confident nonsense is that it's asked at a moment of panic, and panic is a good market for certainty. The honest answer is that the cost of editing is unknown and probably small, the cost of delete-and-repost is known and larger than it looks, and the cost of a wrong number sitting under your name is the only one that compounds.
We build the pre-publish gate into every client's content system, because the founders who never face this decision are the ones who ship the same volume with fewer of these five-minute mornings. If you're publishing consistently and still relying on catching things after they're live, that's the part worth fixing.