How Long SEO Takes to Work

Where the familiar answer comes from

Ask how long search work takes and you will be told three to six months. It is repeated so consistently that it sounds like a finding. It is not. The figure traces to a poll of 3,680 respondents on LinkedIn and X — search practitioners, answering a question about their own field, in public, where the answer reflects on them.

That is a self-selected opinion survey, not a measurement of outcomes. Nobody tracked a cohort of sites, recorded interventions and dated the results. The publisher's own framing of the piece is that "there's no definitive time frame," which is considerably more honest than the number that got extracted from it and repeated ever since. The original article says so plainly if you read past the headline.

None of that makes three to six months a bad rule of thumb. It makes it a rule of thumb, and worth knowing as such, because a buyer who thinks it is a measurement will treat month seven as a breach rather than as information.

What Google says, which is less specific and more useful

The most defensible source on this question is deliberately unbounded. On changes made after a traffic drop, Google states that "some changes can take effect in a few days, while others could take several months." On outcomes it goes further, and this sentence is the one every buyer should read before signing anything: "there's no guarantee that changes you make to your website will result in noticeable impact in search results." The same page adds that more deserving content will continue to rank well regardless.

Google's own guidance page on hiring a search provider — the document that warns about ranking promises and secretive providers — states no timeframe at all. That silence is not an oversight. Any provider quoting a schedule with more precision than the company operating the ranking systems is quoting a sales figure.

One further data point from Google is worth having, because it reframes expectations for new sites: John Mueller has said it can take up to a year for Google to work out where to rank a new site. Not to rank it at all — to settle where it belongs. Google's documentation on traffic drops is the primary source for the quotes above.

Different work runs on genuinely different clocks

Treating all search work as one timeline is the reason the question has no good general answer. Break it apart and each piece becomes forecastable.

• Technical fixes: days to weeks, gated by recrawl. A corrected directive, a fixed status code or an unblocked template can take effect as soon as the affected URLs are crawled again. The variable is not the fix, it is how often your site gets crawled, which is observable in Search Console's Crawl Stats and in server logs.
• Migrations: weeks, by Google's own estimate. Google states a small to medium-sized site can take a few weeks for most pages to move, and larger sites take longer. That is reprocessing, not recovery, and it is why migration monitoring runs against a pre-move baseline rather than against a feeling.
• Manual action recovery: unpredictable, and staged. Reconsideration reviews take several days or weeks, and link-related cases take longer. Nothing can be committed to beyond that, because the review is performed by people.
• Content and authority work: months, with no guarantee attached. The longest clock, the least controllable, and the one most often sold with the shortest promise.

What actually sets the timeline on your site

Six variables do most of the work, and all six are knowable before anyone starts.

How often your site is crawled. Nothing changes until a change is seen. A site whose key templates are fetched daily has a fundamentally shorter clock than one where they are fetched monthly, and this is measurable rather than a matter of opinion.

Whether the change is at template level. A fix applied to a template that governs forty thousand URLs propagates at a different scale from an edit to one page. Template-level work has both the longer lead time and the larger effect.

Whether anything actually ships. The most common cause of a slow result is not search engine latency. It is a specification sitting in a backlog for two release cycles. The consultant's clock and the developer's clock are different clocks, and only one of them appears in proposals.

The competitive position you are moving into. Moving from position 40 to 15 on an uncontested query is a different exercise from displacing an established result on a commercial one.

Site history. A new domain, a recently migrated one, or one recovering from a penalty each carry their own delay before Google settles on a position.

Seasonality. If your category has an annual cycle, a change made in the trough will not be readable until the cycle returns. Compare year over year, not month over month, and know that before the third week.

The order things change in, and how to tell progress from noise

This is the part that turns waiting into observation. Results do not arrive all at once; they arrive in a sequence, and each stage is a prerequisite for the next. Knowing the sequence means you can tell in week three whether something is happening, without pretending that clicks should have moved.

1. Crawling and indexing behavior. First and fastest. Are the affected URLs being fetched again? Has the last-crawl date moved? Are pages leaving the excluded states they were stuck in? This is visible in Crawl Stats, the Page Indexing report and URL Inspection within days, and if it does not move, nothing downstream will.
2. Impressions. The page is now being shown for queries it was not shown for, or shown more often for ones it was. Impressions move before position does, and they are the earliest evidence that a search engine has reassessed the page rather than merely refetched it.
3. Average position. Slower and noisier. Read it knowing what the metric is: Search Console reports the average position of the topmost result from your site, not the average of all your results, which is why it moves in ways that surprise people.
4. Clicks. Last, and gated by things outside the page — what else is on the result page, whether an AI Overview is answering the query in place, how compelling the title is against its neighbors.
5. Conversions and revenue. Lagging further still, and mediated by everything that happens after the click.

The practical rule: if stage one has not moved, the work has not been seen and the timeline has not started. If stage one and two have moved but stage four has not, the work is being seen and is not yet winning — a different problem with a different remedy. Conflating the two is how engagements get abandoned one month before they resolve, and also how they get extended a year past the point of hope.

The things that will fool you

Three confounders account for most misread timelines, and all three are avoidable with dated records.

Ranking updates landing mid-window. Google publishes a dated history of confirmed updates and the durations are not trivial: core updates in this period have run from under a day to more than eighteen days, and a spam update can begin while the effects of a core update are still settling. In late March 2026 a core update completed in under twenty hours and a spam update began three days later. Anyone diagnosing a traffic change in that window without dated data is guessing which event they are looking at, and being confidently wrong is worse than being uncertain.

Changes nobody logged. A developer ships a template change, a marketer swaps a tracking container, an agency updates titles across a section. Without a change log with dates, none of it is separable from the work you are trying to judge.

Measurement changes. A tracking alteration, a consent banner change or a reporting switch can produce a step change in the data that has nothing to do with search at all.

What to agree before the work starts

Three things, and they cost nothing except the discipline of writing them down.

A dated baseline, captured before anything changes, from Search Console and analytics — remembering that Search Console retains sixteen months, so history has a horizon. A change log, maintained by whoever ships, recording what changed and when, including the things that seem irrelevant. And review points matched to the clock of the work being done rather than to the billing cycle: a technical fix reviewed at recrawl, a migration reviewed against the baseline at intervals after launch, a content program reviewed on a horizon long enough for the answer to mean anything.

Agreeing the measures in advance also removes the retrospective metric selection that makes so many reports useless. If the number to be judged on is chosen after the results are in, it will be chosen to suit whoever is presenting it.

When to conclude it is not working

The honest exit criteria are about the leading indicators, not the lagging ones, and they can be applied far earlier than most engagements allow.

If the affected URLs are being recrawled and reindexed and impressions have not moved at all over a period long enough for the change to be assessed, the work is being seen and is not changing how the page is understood. That is a real signal, and it arrives well before any revenue number. If impressions have risen but position has not followed across several months, the site is being considered and found less deserving — which is Google's own framing, and it points at content and authority rather than at more technical work.

What is not a signal: flat clicks in week six, a single bad month, or a drop that coincides with a dated ranking update nobody has separated out yet. Panic on those and you will stop work that was about to be read, which is the most expensive mistake available on this timeline and the most common one.

Frequently Asked Questions

Published 2026-08-11