How long does SEO take — and why the first weeks do not count at all

The short answer, since that is what you came for: indexing takes from a few days to a few months. Ranking — from a few months to over a year. Two different things on two different clocks, and confusing them causes most of the disappointment three months into an engagement.

The longer answer is more interesting, because it explains why the counter so often never started.

Indexing is not ranking

This distinction is the foundation of everything else, in one sentence each.

Indexing is being written into the catalogue. The search engine learns the URL exists, fetches it, reads it and decides whether to keep it at all. That is the entry ticket.

Ranking is the contest for position among those who already hold a ticket. It begins only after indexing and cannot begin earlier.

A firm that says after three months that "SEO is not working" is, half the time, describing something else: its site was never indexed, so there was nothing to rank. The diagnosis and the cure differ entirely, and in visit statistics both look identical — as an absence of traffic.

Why the first weeks do not count at all

In the first weeks after launching a site or publishing a text there is usually nothing to measure — not pessimism, but a description of the mechanism.

The search engine first has to learn the URL exists. Then queue it. Then fetch it. Then read it and decide whether to keep it. Only after those four steps can the page appear anywhere, and only then can any indicator move.

The practical consequence: the first month's report says nothing about the quality of the work. It can show the sitemap was submitted and the URLs discovered — that is all you can expect. Judging a supplier by it is like judging a building site by the excavation.

What does change in that period, invisibly to the statistics: the number of URLs the search engine knows about. The only number worth looking at in the first month, and it is in the indexing report, not in visit statistics.

How long does indexing take in Google

A different question from "how long does SEO take to work", with a different answer, so we separate them straight away.

A new URL on a site the search engine visits regularly can be indexed in hours or days. That is the case the guides talk about — and it exists, but it applies to sites that already have trust and crawler attention.

Re-crawling an existing URL is a different scale entirely. Here we give our own measurement instead of borrowed ranges, because that is the only kind we can defend.

A new domain is the third case and the slowest. A URL with no history gives the search engine no reason to visit regularly — and visit frequency is built on how often something changes there and whether anyone links to it. So the first weeks after a new domain launches look identical however much work went into the content, and that is not a sign anything went wrong.

Three things follow for planning:

• Publication is not the start. The start is the crawl, and that can be weeks away.
• Updating an existing text is no faster than publishing a new one. Sometimes slower, because the URL is already known and the crawler has no reason to hurry.
• There is no button. Submitting a URL for re-indexing is a request, not an instruction, and does not change the queue.

Four gates before the real race begins

Search Console names these states outright, and it is the only place to check where you stand. Each leads to different work:

The second and third lines matter most. "Discovered" and "crawled" look similar and need entirely different things — in the first, the problem is how the search engine judges the whole site; in the second, what it found on one page.

How to check which gate you are at

It takes a quarter of an hour and it is the only thing that turns "it seems slow" into a specific diagnosis.

Open Search Console and go to the page indexing report. Two numbers: how many pages are indexed and how many are not. The second breaks down by reason, and those reasons are the names of the gates.

Check which reason dominates. If most URLs are "discovered, currently not indexed", you are at the second gate and content will not move it. If most are "crawled, currently not indexed", you are at the third, and that is a judgement of value to be taken seriously.

Check the last crawl date of a specific URL. The URL inspection tool gives it outright. The most concrete information in the whole report, because it says whether the search engine has seen the current version of the page at all — or is judging something that no longer exists.

That last one surprises people most often, which is why we show our own numbers below.

How long it really takes — on our own corpus

We will not quote ranges from other people's articles, because they cannot be checked. Here is our own measurement, with all its inconveniences.

On 8 September 2026 we reviewed our URLs. In the state "crawled, currently not indexed" there were 75 URLs. The last-crawl dates fell like this:

More than half of these URLs (53%) were last crawled five to six months before the reading. This despite our sitemap giving a fresh modification date, telling the search engine outright that the content had changed.

That is the number to base any planning on: re-crawling a page can take half a year, and there is no button for it.

The illusion of progress

Over the course of a week, across successive inspection runs, the number of rejected URLs fell from 83 through 77 to 72. It looks like an improvement.

It was not. The number of indexed URLs did not rise. Pages moved between buckets — "discovered" to "crawled", "crawled" to "unknown" — not into the search results. There was movement; there was no progress.

Two pages went backwards: they had been indexed and stopped being so.

Worth knowing before you read a report saying "the number of problems fell by 13%". Movement between states is not progress — the only progress is passing the fourth gate.

What genuinely speeds this up

Things with a documented connection to how quickly a search engine learns of a page:

• A sitemap with a real modification date. No guarantee of a crawl, but the only mechanism by which anything can be submitted.
• A link from an existing, already indexed page. A URL without a single internal link is in the worst possible position — including when it sits in the sitemap.
• Submitting a single URL in Search Console. It works, has a daily limit and does not scale to a hundred pages, but for the five most important it makes sense.
• Fewer pages about the same thing. If a site has twenty texts answering one question, the search engine spends attention on them that the important pages do not get. Slow, and the only lever that improves the whole site rather than one URL.

What to do when a page goes unindexed for months

If the diagnosis above came out badly, there is a path — walked in order, because each step only makes sense once the previous one came out well.

First rule out what blocks outright. A tag forbidding indexing, an entry in the robots file, a canonical pointing elsewhere, password protection. All four can be checked in the URL inspection tool in minutes, and all four are defects rather than a decision by the search engine. The only part of the list fixed for certain and immediately.

Then check whether the URL can be found inside the site at all. A page no other page links to exists only in the sitemap — and a sitemap is a suggestion, not a guarantee. Three links from related texts do more than ten re-indexing requests.

Only at the end, accept that this may not be a problem with that URL. When several dozen URLs hang in this state at once, the cause is a level up: the judgement the search engine passes on the whole site, not a verdict on one text. Google's position has been consistent for years and comes down to one sentence — mass non-indexing is a signal about the quality and usefulness of the whole, not a technical fault to be repaired with a button.

This is the hardest conclusion here to accept, because there is no quick action for it. It does have a practical consequence: adding more texts to a site the search engine does not show enlarges the problem instead of solving it.

What is sold as acceleration and is not

• Publishing more often. More pages in the same queue do not shorten the queue. On a site that is already waiting, it makes things worse.
• Changing the publication date without changing the content. A signal the search engine can verify by opening the page.
• Adding keywords to existing texts. No connection to whether the page gets crawled.
• Traffic from social media. Worth being precise, because a stronger claim circulates than can be defended: a link placed somewhere the search engine visits gets discovered — that is the linking mechanism, not the traffic. A thousand people clicking in the feed is not a documented accelerating signal.

Three situations, three different clocks

"How long does it take" has three answers depending on what you start from. Confusing them is the second-biggest source of misunderstanding, after confusing indexing with ranking.

A new page on a site that already works. The shortest clock. The search engine visits that site regularly, so a new URL is usually discovered quickly — provided a link from an existing, indexed page leads to it. Without one, the URL joins the queue on the same terms as a stranger's.

A new site on a new domain. The longest clock and the most often underestimated. A domain with no history gives the search engine no reason to treat it as a priority. This is counted in months, not weeks, and no optimisation shortens it.

A rebuilt site on the same URLs. The intermediate case, and the riskiest. The addresses are known, so discovery is immediate — but if the paths changed along the way and the redirects were not prepared, the whole history goes in the bin and the clock restarts. The only one of the three in which the rebuild itself can do you harm; described in redesign and optimisation.

What to do in the meantime

Waiting is not a strategy, but not every piece of work in that period is useful either. Three things that genuinely shorten the road to having something to measure:

Tidy up the internal linking. Every important page should have a link from another, already indexed one. The only item on this list that also works on pages that have existed for a long time and are stuck in the queue.

Remove or merge whatever repeats itself. A site with twenty texts about the same thing scatters attention the important pages do not get. The effect is slow and applies to the whole domain rather than one URL — but it is the only lever that works on the second gate.

Prepare the measurement before there is anything to measure. Record the baseline values, set up a correct contact event, exclude your own traffic. In three months it is too late: the data from this period will already have been collected — and it will contain your own testing.

What not to do: do not publish more in order to go faster. More pages in the same queue do not shorten the queue.

How to tell it is going well before the customers arrive

The signals come in order, each preceding the next by weeks. Looking straight at the last is the commonest reason projects that are going well get stopped halfway.

1. The number of indexed pages rises. The first signal, and the only one that says you have started at all.
2. Impressions rise in Search Console. The site starts appearing, though still low down. There may be no clicks at all, and that is normal.
3. The average position improves. A move from twentieth to twelfth is invisible in visit statistics and is real progress.
4. Clicks appear. Usually in a jump, when the position crosses the visibility threshold.
5. Enquiries appear. The last signal and the only one the board cares about.

Months pass between the first and the fifth. Anyone reporting only the fifth will have nothing to show for the first six months — even if everything is going well.

What affects the pace — and what you control

Not all the factors are equal and only some are on your side. Worth knowing what is worth fighting for.

The three items marked as yours cost least. No accident: the things you genuinely control are usually cheap and boring, while the expensive and impressive ones — like publishing more often — do not touch the problem.

Also worth noticing what is not on this list: keywords, text length and headings. All three matter for position, meaning what happens after the fourth gate. Before it they change nothing, and working on them in the first months is work done too early.

What not to promise and what not to buy

"Top 3 in three months" is a promise nobody can keep, because a third party decides the order. A supplier can promise work and its scope, not a result they do not control. An offer guaranteeing a position either concerns a phrase nobody searches for, or will not be honoured.

What to demand instead of a guarantee: a report showing which of the five signals you are at, and a reason why the next step is the one proposed.