Before anything else: confirm the problem is yours. If every site in your niche fell together, the category was re-rated and there is nothing on your pages to fix. That check is the subject of reading core update winners and losers properly, and skipping it leads to months spent rewriting content that was never the cause.
What follows assumes the check came back the other way: comparable sites held, yours did not.
Recovery is mostly subtraction
The instinct after a drop is to publish and improve. Across sites that actually recover, the dominant action is the opposite — removing content rather than adding it.
The reason is structural. Core updates assess sites as well as pages, so a large body of weak material affects how everything else is judged. Improving twenty pages while four hundred thin ones remain changes very little.
That makes the first task an inventory rather than an edit: find what should not exist, and be honest about how much of it there is.
The order of operations
| Step | Action | Why it comes here |
|---|---|---|
| 1. Confirm | Compare against your niche, same window | Rules out a category-wide re-rating |
| 2. Segment | Group pages by topic cluster, not by date | Losses concentrate in clusters, not randomly |
| 3. Cut | Remove clusters the site has no claim to | The single action most associated with recovery |
| 4. Deepen | Make the remaining clusters genuinely complete | Depth is what the surviving sites share |
| 5. Wait | Do nothing further until the next core update | Re-evaluation happens then, not continuously |
Deciding what to cut
The test is not traffic. It is whether the site has any genuine claim to the subject — whether a reader would think you were the right person to ask.
A gardening site that published laptop reviews because the difficulty scores looked favourable has a cluster with no claim. It may still bring visits. That is exactly what makes the decision hard and does not make it wrong.
- No claim to the subject — remove the cluster entirely, not page by page.
- Claim, but thin execution — keep and rebuild properly, or remove if you will not do the work.
- Claim and depth — leave alone. These are what you are protecting.
Partial deletion is the common failure. Removing half a weak cluster leaves the signal intact and the site smaller, which is the worst of both outcomes.
How long it takes
Longer than anyone wants. Recovery generally appears at a subsequent core update rather than in the weeks after changes are made, because that is when sites are re-assessed.
Agency post-update analyses — Amsive publishes winner and loser breakdowns after each rollout — consistently show movement clustering around update rollouts rather than spreading evenly between them. That pattern is the practical constraint on everything here.
The consequence is uncomfortable: changes made this week are graded in three months. That argues for doing the substantial thing you believe in, once, rather than a series of small experiments whose effects you will never be able to attribute.
Recovery is often partial
Sites that recover frequently return to a fraction of their former traffic rather than all of it, and that is a normal outcome rather than a failed one.
Some of the lost traffic belonged to pages that should not have ranked. Getting it back would mean restoring the thing that caused the problem. A smaller site holding steady through two subsequent updates is a better asset than a larger one that swings with every rollout.
The case for doing nothing
Sometimes the correct response to a drop is no response at all, and it is worth naming because it almost never gets recommended.
If the site is already narrow, already deep, and already written by someone with a real claim to the subject, there may be no weak cluster to remove. In that case the movement was a re-rating of the field rather than a judgement on you, and changing things introduces risk without addressing a cause.
The honest test is whether you can name what you would cut. If you cannot — if every cluster passes the claim-and-depth test — then waiting through the next rollout is a defensible decision rather than an absence of one.
What not to do
- Do not publish your way out. More pages on a site already judged to have too many weak ones compounds the problem.
- Do not change everything at once. If you alter the content, the structure and the linking simultaneously, you will not know which mattered.
- Do not refresh dates without changing anything. It is transparent and it wastes a crawl.
- Do not buy links. Core updates are not a link problem, and this reliably makes the position worse.
- Do not act on day three. Rankings move erratically during a rollout; wait for it to be confirmed complete.
Knowing whether it worked
Measure the same way you diagnosed: against your niche, through the same window, across the next rollout. Absolute traffic will mislead you, because seasonality and category movement are mixed into it.
Our update view charts before-and-after traffic for every confirmed update, which is what makes that comparison repeatable rather than a fresh research project each time. The wider method is in reading a Google core update, and the site-wide signal that makes subtraction work is covered in the helpful content update, four years on.
One limit worth stating: all of this rests on third-party traffic estimates, ours included. They are reliable for comparing your site against its competitors through the same window — which is the comparison recovery decisions actually need — and unreliable as absolute figures.
Frequently asked questions
How long does it take to recover from a Google core update?
Usually months, and recovery typically appears at a subsequent core update rather than in the weeks after changes are made. Sites are re-assessed when updates roll out, so improvements made now are effectively graded at the next one.
Should I delete content to recover?
Often yes, and it is the action most associated with recovery. Clusters the site has no genuine claim to are worth removing entirely rather than page by page, because partial deletion leaves the underlying signal intact.
Can I recover by publishing more?
Rarely. If a site was judged to carry too much weak material, adding to it compounds the problem. Depth within a defined scope is what the sites that hold through updates tend to share, not publishing volume.
Will I get all my traffic back?
Frequently not, and that can still be a successful outcome. Some of the lost traffic belonged to pages that arguably should not have ranked, and a smaller site that holds steady across several updates is worth more than a larger one that swings with each.
The takeaway Confirm the problem is yours, cut the clusters you have no claim to, deepen what remains, and then wait for the next update to grade it. Recovery is subtraction, and it is slower than the advice suggests.
