The problem with update commentary is sampling. Someone looks at six sites that dropped, notices they all had programmatic content, and concludes the update targeted programmatic content. Nobody checks the programmatic sites that went up, because nobody posts about those.
It is survivorship bias with the survivors removed instead of the casualties — the same error, upside down. And it is self-reinforcing, because the people with the largest audiences during an update are the people it hurt most.
Start with the winners, not the losers
This inverts how most people do it, and it is the single biggest improvement you can make. Losers tell you what stopped working, which is a long and ambiguous list. Winners tell you what Google chose to promote, which is far more specific.
Pull every site in one niche that gained meaningfully — say 25% or more — across the update window. Now you have a cohort that the algorithm deliberately rewarded. Whatever they share is your signal.
There is a practical reason too. You cannot copy a site’s way out of a penalty, but you can look at what a winner does and decide whether to do the same thing. One list is actionable and the other is a post-mortem.
Stay inside one niche
Core updates do not hit every vertical the same way. A single update can be brutal in health and almost invisible in B2B software. If you mix niches in one analysis, you average away the effect you are trying to see.
Pick the niche you actually work in. Twenty sites from your own vertical will teach you more than two thousand spread across the web, and the same cohort supports validating whether that niche is still enterable.
Building the cohort properly
The analysis is only as good as the set you assemble, and there are four ways to assemble it badly:
- Too few sites. Under about fifteen you are reading noise. If the niche cannot supply fifteen, widen it by one adjacent category, not by five.
- No floor on size. Small sites swing wildly for reasons that have nothing to do with the update, and traffic estimates are noisiest at exactly that scale. Exclude anything under a few thousand monthly visits before you start.
- Mixed business models. An ecommerce site and a publisher in the same vertical are playing different games. Separate them or note which is which.
- No control group. Pull the flat sites too. A trait that every winner shares is only interesting if the sites that did not move mostly lack it.
That last one is the step almost everyone skips, and it is what separates an analysis from a hunch. If eight of your ten winners publish original photography, but so do eight of ten sites that went nowhere, photography is not the signal.
Write the cohort down before you start looking for patterns, too. The temptation once you have a theory is to quietly drop the two winners that contradict it, and a list fixed in advance is the only real defence against doing that without noticing. If a site has to be excluded, note the reason next to it — the exclusions are frequently more interesting than the pattern they were getting in the way of.
The four things worth checking
Once you have your cohort of winners, there are four attributes that consistently separate them from the flat and the falling:
- Content origin. Is there anything in these pages that could not have been written from other pages? Original testing, proprietary data, photographs, first-hand accounts.
- Site focus. Winners are frequently narrower than losers. A site covering one subject thoroughly tends to outperform one covering nine subjects adequately.
- Publishing pattern. Look at whether they grew by adding pages or by improving existing ones. Since roughly 2023 the second has outperformed the first more often than not.
- History through prior updates. A site that has gained through three consecutive core updates is doing something structurally right. One that zigzags is riding volatility.
Losers tell you what stopped working. Winners tell you what Google chose to promote.
What the comparison usually shows
Run this a few times and the contrast tends to sit in the same places:
| Attribute | Typical winner | Typical loser |
|---|---|---|
| Topic range | One subject, covered deep | Whatever had search volume |
| Growth method | Fewer pages, better maintained | More pages, rarely revisited |
| On-page evidence | Something only they could publish | Well-written synthesis of others |
| Update history | Gained across several in a row | Alternates up and down |
| Authorship | Named people with a record | Staff bylines or none |
Treat this as a prior to test, not a conclusion to apply. The point of the method is that your niche gets to disagree with it.
Distinguish a hit from a drift
Not every drop is an update. Before you conclude anything, check the timing against the rollout window on Google’s Search Status Dashboard and check your peers. If everyone in the niche moved the same direction at the same time, it was the algorithm. If you alone moved, look at your own site first — a template change, a crawl issue, a lost link, or seasonality.
This is the question the update view exists to answer, and it is worth answering before you rewrite anything. A surprising share of “we got hit by the core update” turns out to be a broken canonical.
Wait for the rollout to finish
Core updates take one to three weeks to roll out, and movement during that window is genuinely unreliable — sites commonly drop in week one and recover by week three. Acting on day-four data is how people panic-delete pages that were about to come back.
Snapshot before the rollout starts, snapshot again a week after Google declares it complete, and compare those. Anything in between is weather, not climate.
If you were the one who lost
Two things are worth saying plainly. First, core updates are reassessments rather than penalties, so there is no specific thing to remove and no reconsideration request to file. Second, recovery generally arrives at a subsequent core update rather than continuously, which means the feedback loop on any fix you make is measured in months.
That timeline is the real constraint. It means you get very few attempts per year, so the work should go into the one or two changes your winners-versus-flat comparison actually supports — not into a list of thirty tweaks where you will never learn which one mattered.
What to do with the conclusion
The output of this exercise should be one sentence, not a strategy deck. Something like: “In this niche, sites with original testing photography and a narrower topic range gained; aggregator-style roundups lost.” That is actionable. “The update was about quality” is not.
Then check it against the next update. If the same pattern holds twice, it is a trend worth building on. If it reverses, you learned something more valuable — that this niche is volatile, and you should weight your bets accordingly.
Frequently asked questions
How long does a Google core update take to roll out?
Typically one to three weeks. Rankings move unpredictably during that window, and sites often drop in the first week then recover before it finishes, so any analysis done mid-rollout is measuring noise.
How do I know if a traffic drop was a core update or something else?
Check whether your peers moved at the same time. If the whole niche shifted together it was the algorithm; if only you moved, look first at your own site for a template change, a crawl issue or a broken canonical.
Can a site recover from a core update?
Yes, but recovery usually arrives at a later core update rather than gradually. There is no penalty to remove and nothing to file, so the practical constraint is that you get very few attempts per year.
The takeaway One niche, winners first, a flat control group to test against, and no conclusions until the rollout finishes. The output is a single sentence you can check against the next update — which is more than any take published in week one can offer.