How long does SEO take is the question every buyer asks in the first meeting and no agency answers honestly, because the honest answer is a range and ranges lose pitches. At Search Central Live Deep Dive Europe in Barcelona on October 2, Gary Illyes put Google's own ranges on a slide. John Campbell of ROAST published the tables the same day, and Search Engine Journal reproduced eleven of the twenty-three rows two days later. The numbers are not flattering to the way most programs are sold, scoped or reported.
How long does SEO take, according to Google's own numbers
Google's own estimates split SEO timing into eleven distinct processes that each run on a different clock, from roughly 20 hours for discovering a new URL to three to six months for recovering from a core update. There is no single SEO timeline, and treating one as though it exists is the root of most disappointed clients.
The provenance matters, so take it plainly. These figures come from a recap written by a community speaker at the event, not from a published Google document, and the slides were labelled as estimations based on internal analysis. Google did not define typical, did not give a sample size and did not state a measurement period. The recap also omitted the fastest column and made eleven of twenty-three rows public. That is thinner evidence than a study and much stronger evidence than the folklore it replaces, which was agencies quoting each other.
| PROCESS | TYPICAL | SLOWEST |
|---|---|---|
| Discovery of a new URL | ~20 hours | Weeks to never |
| Refresh crawl of a known URL | ~30 days | Weeks to never |
| Sitemap processing | ~24 hours | Up to 14 days, or never (quality) |
| Indexing, end to end | ~1.5 hours | Months, or never (quality) |
| Canonicalization change | 1 to 3 weeks | Months (conflicting signals) |
| Title change in results | 1 to 2 days | Several weeks to months |
| Snippet change in results | 1 to 2 days | Several weeks to months |
| Search Console removal by owner | ~2 hours | 24 hours |
| Manual action removal | 1 to 2 weeks | 4 to 6 weeks, longer if dormant |
| Site move | 1 to 3 months | 6 months to 1 year or more |
| Core update recovery | 3 to 6 months | 6 months to 1 year (next core update) |
Read down the typical column and the shape of the table is obvious. Everything mechanical is fast. Fetching, parsing, indexing and rendering a title are measured in hours and days. Everything evaluative is slow. Deciding a site deserves its old positions back, or that a migrated domain inherits what the old one earned, is measured in months. The engineering is not the bottleneck and has not been for years.
This is also why the stages compound rather than run in parallel. A page cannot be indexed before it is crawled, and a canonical change cannot settle while the signals pointing at it still conflict. A 30-day refresh crawl followed by a one-to-three-week canonicalization change is a seven-week wait before anyone can fairly judge the fix. We flagged the same compounding problem in Search Console's click data earlier this month: the lag is not one delay, it is several stacked on top of each other.
The 20-hour discovery number is the least useful one
Discovery at roughly 20 hours and indexing at roughly 90 minutes are the two numbers that will be quoted most and matter least, because almost nobody's SEO problem is that Google has not found their pages. The interesting figure sitting beside them is the 30-day refresh crawl for a URL Google already knows.
Think about what a 30-day typical refresh does to an optimization programme. You rewrite 40 product pages in week one. Google has known those URLs for years, so they sit in the refresh queue, not the discovery queue. The median page gets re-crawled about a month later, some considerably after that, and only then does evaluation begin. Report on that work at the 30-day mark and you are reporting on pages Google has in many cases not looked at since you changed them.
“A timeline you cannot measure inside your reporting window is not a timeline. It is a hope with a date on it.”
The practical move is to stop treating recrawl as free and start treating it as inventory. Know which of your URLs are high-demand and recrawled in days, and which sit in the long tail and go a month or more between visits. The second group needs its changes batched and submitted deliberately, because you get roughly one evaluation per month per page and wasting one costs you four weeks. Illyes' own slowest column for refresh reads weeks to never, and the never there is not a technical fault.
Core update recovery is 3 to 6 months
Core update recovery at three to six months typical, and six months to a year in the slowest case, is the single most commercially awkward number on the slide, because it is longer than the review cycle almost every SEO engagement is judged on. The arithmetic does not work.
Illustrative math: share of a 12-month engagement consumed by one typical wait, using Illyes' typical figures
A 12-month contract signed the month a site gets hit by a core update can spend half of its life inside the typical recovery window, and all of it inside the slow one. The quarterly business review at day 90 lands in the first third of that window. Nothing has resolved. The data looks flat because it is flat, and the honest report says so, which is a conversation most account teams would rather have at month nine than month three.
Worse, the slowest column attaches recovery to the next core update. That makes the clock external. You are not waiting on your own work to take effect, you are waiting for Google to run the next evaluation, and the gap between core updates is not yours to schedule. Google's most recent confirmed core update was in May 2026, and the September 2026 spam update took almost 14 days to finish rolling out against under three days for the three spam updates before it. Rollout length is drifting upward. Plan for the slow case.
How long does SEO take after a site migration
A site move typically takes one to three months to settle by Google's estimate, and six months to more than a year in the slowest case, which puts migration on roughly the same clock as core update recovery rather than the two-week clock most migration plans assume.
Most migration plans we review budget carefully for the cutover and almost nothing for the settling period. The redirect map gets three weeks of QA, the launch gets a war room, and then the plan ends. There is no owner, no budget line and no agreed metric for the eight weeks after launch when the domain is being re-evaluated and traffic is genuinely allowed to be down. That silence is where migrations get declared failures that were simply unfinished.
This is where our enterprise SEO work spends most of its first quarter, and where the MarketCheck engagement earned its result: not in the cutover, but in the unglamorous weeks afterwards when somebody has to keep checking that what was supposed to transfer actually did.
Never is a quality verdict, not a technical one
Five of the eleven published slowest-case outcomes end in the word never, and three of those are annotated with quality as the reason, which is the most important thing on the slide and the part least likely to be quoted. Some pages are not slow to index. They are declined.
Sitemap processing can end in never for quality. End-to-end indexing can end in never for quality. Refresh crawling can end in never. Read those together and the conclusion is blunt: Google is making a worth-it judgment at the crawl and index stage, before ranking is even in play, and no technical intervention reverses a worth-it judgment. You can fix a sitemap in an afternoon. You cannot fix thin pages with a sitemap.
It follows that the first diagnostic question on a stalled site is not how long SEO takes, it is which clock the site is on at all. A site waiting on evaluation and a site that has been declined produce identical flat lines in a rankings report, and the remedies have nothing in common. Telling them apart is a different exercise from ranking reporting, and it is the question our SEO audit services are built to answer before anyone writes a roadmap. John Mueller made the same point on Search Off the Record on October 1, noting that crawl demand often tracks a site's perceived quality.
There is also a measurement trap here worth naming. If you are diagnosing from third-party rank data, be aware the data itself has been unreliable this autumn: Nozzle and SEOmonitor both reported Google returning complete but incorrect result pages to automated requests from late September, which is a problem we went through in detail last week. Diagnosing a quality decline from corrupted rank data is how programmes end up fixing the wrong thing for a quarter.
What to change in your reporting cadence
The actionable consequence of Google's timing table is a reporting cadence matched to each process rather than to the calendar, because a monthly report measuring a three-to-six-month process will show noise eleven times before it shows signal. Three changes cover most of it.
None of this makes SEO faster. It makes the waiting legible, which is the part clients actually leave over. Our read, and we will say it plainly: the agencies that lose enterprise accounts in 2027 will not be the slow ones, they will be the ones that promised a 90-day answer to a 180-day question and had nothing honest to show at day 90.
Do this next. Take your three largest accounts, write the typical range from Illyes' table next to every open workstream, and find the ones whose review date arrives before their window closes. Those are the conversations to reset now, while resetting them is still your idea. If you are buying SEO audit services this quarter, make that the deliverable: not a list of fixes, but a verdict on which clock each part of the site is actually on.
Questions about SEO timelines
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Tyler leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.