How long do Google Search changes take? The numbers, and where they actually come from

A table of Google timings has been going round for a fortnight. How long to discover a URL, how long to reindex one, how long to recover from a core update. It is genuinely useful and I have already used it twice in client conversations.

It is also not Google documentation, which is how most people are passing it on.

The figures come from a talk Gary Illyes gave at Search Central Live Deep Dive Europe in Barcelona on 2 October 2026, written up by an attendee and reported from there. That does not make them wrong. It does change what you can do with them, and the difference matters if you are about to quote one to a client.

Here is what the numbers say, where they came from, what Google itself publishes, and the one word in the table that I would actually plan around.

Gary Illyes presented them at Search Central Live Deep Dive Europe, Barcelona, on 2 October 2026. John Campbell of ROAST, a community speaker at the event, wrote up a recap. Search Engine Journal reported it from that recap. The slides were not on Google's own Search Central events page when SEJ published.

So the chain is: Google employee, on stage, no published deck, recapped by an attendee, reported secondhand.

To the recap's credit, it says the figures came from Google's internal analysis and calls them reference points rather than deadlines. It also admits what it does not have: no sample size, no measurement period, no definition of what counts as typical. It leaves out a third column, the fastest time for each process, which was on the slides.

I am labouring this because we published a piece three weeks ago arguing that an idea spread across our whole industry on the strength of sounding right, and that the test is who ran it, what happened, and what they found that they did not want to find. A table with no sample size fails the first part of that test. It would be convenient to skip the check because the numbers are useful and I want them to be true.

Much less, and much vaguer. Worth knowing, because this is what you can quote without a caveat.

On crawling, Google's own documentation says: "Crawling can take anywhere from a few days to a few weeks." On whether asking helps: "Requesting a crawl does not guarantee that inclusion in search results will happen instantly or even at all."

On site moves, Google says a small to medium site can take a few weeks for most pages to move, and larger sites take longer, with ranking fluctuations while it recrawls and reindexes.

On core updates, Google says some effects can appear within days, but confirming whether a site has actually improved can take months.

Notice the shape. Google publishes ranges with the uncertainty left in. The circulating table publishes point estimates with the uncertainty taken out. That is the difference between the two, and it is the reason the table travels further.

With all of the above attached, these are the figures I would keep. Source throughout is the Barcelona recap, not Google documentation.

Two of those surprised me.

Thirty days to recrawl a page it already knows. Most people assume updating an existing page is the fast path and publishing a new one is the slow path. On these figures it is the other way round: roughly 20 hours to find something new, roughly 30 days to come back to something old. If you have rewritten a key service page and nothing has moved in a fortnight, that is not a failure, it is a Tuesday.

Indexing end to end at about 1.5 hours. That only means every critical process completed successfully. It is the best case dressed as a typical case, and the honest number next to it is the one in the right-hand column.

Look at the right-hand column again. Four of those rows end in "never".

Not "eventually". Not "longer than you would like". Never. And in the recap, the reason attached to never is quality.

That is the most useful thing in the whole table and it is the part nobody is sharing. Google's own documentation says the same thing in plainer words: requesting a crawl does not guarantee inclusion "instantly or even at all."

So the question "how long will this take to index" has a hidden second question underneath it, which is "is this page worth indexing". If the answer to the second is no, the first has no answer. A page that is thin, duplicated, or built from the same material as the top ten results is not waiting in a queue. It is not in the queue.

This is not a reason to despair, it is a reason to stop refreshing Search Console and go and look at the page.

Stop promising dates for things you do not control. We do not guarantee anything, and this table is a good illustration of why. Every row has a tail, and some of the tails are a year long.

Separate what you control from what you wait for. You control whether the page is worth indexing, whether a crawler can reach it, and whether the redirect map is right. You do not control the queue. Put the effort in the first group.

Build in the waiting. If a site move is typically one to three months and can be a year, then a migration plan with a four-week measurement window is not a plan, it is an optimism. Our migration guide covers the mechanics and the redirect strategy is the part that decides which end of the range you land on.

Do not read a trend from a fortnight. We relaunched our own site on 3 October. Nine days later impressions had tripled and the headline looked excellent, and we published why that was the least interesting number in the set. The settling takes longer than the spike.

Nobody asks how long indexing takes. They ask why it has not worked yet, which is a different question wearing a technical costume.

Most of the time the honest answer has two parts. The first is that some of this is genuinely a queue and the timings above are the queue. The second is that if a page has been live for three months and is not indexed, the queue is not the problem.

And there is a third thing worth saying out loud, because it is ours and we keep saying it: indexed is not the goal. We have a page that ranks 3.3 for its main term, took 2,695 impressions across all its queries last month and got no clicks at all. It is indexed. It is ranking. It is doing nothing. Rankings stopped being the outcome, and so did being indexed.

Print it, and use it to set expectations rather than to make promises. Attach the caveat when you quote it, because a client who later discovers the source was a conference recap will discount everything else you told them.

And when somebody sends you a figure that is useful, flattering, or confirms what you already wanted to do, do the boring thing first and find out who ran it. I nearly did not, because the numbers are good and I wanted to publish them the same day.