Skip to content
Doctor SEO
Free SEO audit

SEO Trends

Google Read Out Its Own Crawl and Index Timings

Gary Illyes presented Google internal crawling, indexing and serving timings in Barcelona. Google has published none of the figures, so we checked which rows a dated Google document corroborates and which appear nowhere at all.

4 October 2026 15 min read
Three-step diagram of how Google's crawl and index timings reached the reader. One, the slide: Gary Illyes presented typical and worst-case durations in Barcelona on 2 October 2026, and what it does not show is anything published, because Google published no blog post, no documentation page and no changelog entry. Two, the recap: one attendee, John Campbell of the agency ROAST, recapped it on LinkedIn, and what it does not show is a first-hand reading, because LinkedIn blocks automated retrieval and this article has not read it directly. Three, two reports: PPC Land published eight rows on 3 October and Relevant Audience sixteen on the 4th, and what it does corroborate is that the eight shared rows match exactly, which validates the transcript and not the data. None of the sixteen figures carries a methodology, a sample size, a date range or a definition of typical.

Google has numbers for how long its own pipeline takes, and on 2 October 2026 it read some of them out loud. Gary Illyes, an analyst at Google, presented a set of typical and worst-case durations for crawling, indexing and serving at Search Central Live Deep Dive Europe in Barcelona. A new URL is typically discovered in about 20 hours. A URL Google already knows is typically refreshed in about 30 days.

The surprising figure is not the 20 hours. It is the 30 days, which says the slow part of getting a change seen is almost never the discovery of a brand-new page.

Google has published none of these figures: no Search Central blog post, no documentation page, no changelog entry. They reached the industry through a conference recap — and the one row a dated Google document corroborates at the same magnitude is the least interesting row on the slide.

The slide, one attendee’s recap and the two reports that transcribe itThree-step diagram of how Google’s crawl and index timings reached the reader. One, the slide: Gary Illyes presented typical and worst-case durations in Barcelona on 2 October 2026, and what it does not show is anything published, because Google published no blog post, no documentation page and no changelog entry. Two, the recap: one attendee, John Campbell of the agency ROAST, recapped it on LinkedIn, and what it does not show is a first-hand reading, because LinkedIn blocks automated retrieval and this article has not read it directly. Three, two reports: PPC Land published eight rows on 3 October and Relevant Audience sixteen on the 4th, and what it does corroborate is that the eight shared rows match exactly, which validates the transcript and not the data. None of the sixteen figures carries a methodology, a sample size, a date range or a definition of typical.None of these figures is published, and this is how theyreached youThree hops between a slide in Barcelona and the number you are reading.1The slideGary Illyes presented typical andworst-case durations in Barcelonaon 2 October.WHAT IT DOES NOT SHOWGoogle published no blog post, nodocumentation page and nochangelog entry.2The recapOne attendee, John Campbell ofthe agency ROAST, recapped iton LinkedIn.WHAT IT DOES NOT SHOWLinkedIn blocks automatedretrieval, so this article has not readit directly.3Two reportsPPC Land published eight rows on3 October and Relevant Audiencesixteen on the 4th.WHAT IT DOES CORROBORATEThe eight shared rows matchexactly, which validates thetranscript, not the data.None of the sixteen figures carries a methodology, a sample size, a date range or a definition of typical.Conference recap of 2 October 2026, via PPC Land and Relevant Audience. No sample size.doctor-seo.net
The three hops between Gary Illyes’s slide at Search Central Live Deep Dive Europe (Barcelona, 2 October 2026) and this article: John Campbell’s recap (ROAST) on LinkedIn, and the reports by PPC Land (3 October, eight rows) and Relevant Audience (4 October, sixteen). Google has published none of the figures, and none of them carries a methodology or a sample size.

What was presented, and how it reached this article

The talk was given by Gary Illyes at Search Central Live Deep Dive Europe, held in Barcelona from 30 September to 2 October 2026. The timings were recapped by John Campbell, Head of Innovation and AI at the agency ROAST, who attended. Campbell’s recap was published on LinkedIn, which blocks automated retrieval, so this article could not read the recap directly and does not claim to have.

Two independent reports carry the figures. Luis Rijo at PPC Land published eight rows on 3 October 2026. Relevant Audience published sixteen rows on 4 October 2026. The eight rows that appear in both are identical in both, which corroborates the rendering of the recap — not the data behind it.

No methodology, no sample size, no date range and no definition of “typical” was disclosed with any figure. Relevant Audience puts it plainly: treat them “as an attendee’s account of a conference slide, not as a service-level promise”.

The sixteen timings as reported

These are the durations as rendered by the two reports named above. They describe Google’s pipeline, not any individual site, and none carries a sample size.

Stage Typical Slowest
Discovery of a new URL ~20 hours Weeks to never
Refresh of a known URL ~30 days Weeks to never
Sitemap processing ~24 hours Up to 14 days, or never (quality)
Crawl capacity change 4 hours to 1–2 weeks 1–3 weeks (recovery)
Crawl demand change ~20 hours Weeks to months
robots.txt change ~24 hours 25 hours
Indexing, end to end ~1.5 hours Months, or never (quality)
Canonicalisation change 1–3 weeks Months (conflicting signals)
Site move 1–3 months 6 months to 1 year or more
Removal 1–3 weeks Months
Structured data update Hours to 1–2 weeks Weeks, or never
Removal via Search Console tool ~2 hours 24 hours
Snippet or title change 1–2 days Weeks to months
Manual action removal 1–2 weeks 4–6 weeks, longer for dormant sites
Core update recovery 3–6 months 6 months to 1 year
Spam update change 1–2 weeks Months

What a dated Google document actually corroborates

Six Google sources were read live on 4 October 2026 to see which of the sixteen rows exist anywhere in Google’s published record. The answer: one at the same magnitude, one at a different magnitude, two qualitatively, twelve not at all.

Google source Last updated What it contains
Large site owner’s guide to managing crawl budget 2026-07-22 UTC No discovery figure and no refresh figure. Google’s own document about crawling does not contain the two headline numbers.
robots.txt introduction and guide 2026-08-31 UTC “Google generally caches the contents of robots.txt file for up to 24 hours.” Matches the recap’s ~24 hours.
Site moves with URL changes 2026-08-20 UTC “for medium-sized websites, it can take a few weeks or more.” Faster than the recap’s 1–3 months typical.
Google Search core updates and your website 2025-12-10 UTC “it could take several months”, and if nothing changes “that could mean waiting until the next core update”. No 3–6 month figure.
Search Central documentation changelog Read 4 Oct 2026 One October 2026 entry, dated 1 October, about generative AI content. Nothing about crawling timings, indexing timings or Barcelona.
Search Status Dashboard Read 4 Oct 2026 September 2026 spam update, begin 2026-09-24T16:15:00+00:00, no end timestamp, “may take up to two weeks to complete”.

The site-move row is worth pausing on. Google’s documented guidance for a small to medium site is “a few weeks or more”; the recap’s typical figure is 1–3 months. A migration plan built on the published document is more optimistic than the slide.

The spam-update arithmetic, as of today

One row connects directly to a live event. The September 2026 spam update has a begin timestamp of 2026-09-24T16:15:00+00:00 on Google’s Search Status Dashboard, and when the dashboard was read on 4 October 2026 it still carried no end timestamp. That is day 10 of a window Google described as “may take up to two weeks to complete”.

The recap’s spam-update row says a change typically shows in 1–2 weeks and at worst in months. Both numbers point past the dashboard window rather than inside it, and neither tells a site owner whether their own traffic movement belongs to this update.

What to do this week

These figures change what counts as a problem, which is their practical value.

Discovery, refresh and site move: three durations and what each one changesTable of three stages with the typical reported duration and what each one changes in a plan. One, a new URL: about 20 hours to discovery, and weeks to never in the reported worst case, so the escalation threshold belongs at a week rather than the day after publication. Two, a known URL: about 30 days to the next crawl, and weeks to never in the worst case, so the levers are internal linking and sitemap lastmod accuracy, not patience. Three, a site move: 1 to 3 months, against the a few weeks or more that Google documents, so the published document is more optimistic than the slide and the plan has to be re-dated. No figure carries a sample size: record the gap between publication and first crawl on 20 of your own URLs.The slow part is not finding a new page, it is re-readingone that existsThree reported durations and what each one changes in this week’s plan.STAGETYPICAL, AS REPORTEDWHAT IT CHANGES1A new URL~20 hours to discovery; weeks to never in thereported worst case.Set the escalation threshold at a week, not the dayafter publication.2A known URL~30 days to the next crawl; weeks to never inthe worst case.The levers are internal linking and sitemap lastmodaccuracy, not patience.3A site move1–3 months, against the “a few weeks or more”Google documents.The published document is more optimistic than theslide: re-date the plan.Record the gap between publication and first crawl on 20 of your own URLs: that figure is dated.Conference recap via PPC Land, 3 Oct 2026, and Google’s site-move guide, 2026-08-20 UTC.doctor-seo.net
Three of the sixteen reported durations and what each one changes in a plan. Figures from the recap of the 2 October 2026 talk, via PPC Land (3 October) and Relevant Audience (4 October), with no methodology and no sample size. The site-move comparison uses Google’s site moves with URL changes guide, last updated 2026-08-20 UTC, read on 4 October 2026.
  • Stop diagnosing a publishing problem at 24 hours. If a new page is not indexed a day after publication, the reported typical discovery time has barely elapsed. Set the internal escalation threshold at a week, not a day.
  • Treat updated pages as the real latency problem. A ~30 day typical refresh means a price change, a disclaimer or a corrected statistic on an existing page can sit unseen for a month. If a change must be seen quickly, the lever is internal linking and sitemap lastmod accuracy, not patience.
  • Re-date your migration plan. If the plan was written against Google’s documented “a few weeks or more”, run the schedule again at 1–3 months and check what breaks — contract end dates, redirect retention, reporting comparisons.
  • Measure your own site rather than adopting the number. Record the gap between publication and first crawl for 20 recent URLs and keep the median. Your own figure is first-party, dated and citable; Google’s is none of those for your property.

The last point is the one that compounds. The analytics and measurement level of the course covers building that record from Search Console, and the technical SEO level covers the crawl and indexing mechanics these timings sit on top of. If the pipeline itself is unfamiliar, how Google works is the shorter route in.

Where the industry genuinely disagrees

Two questions behind this story are unsettled, and nothing in the evidence above settles either. Doctor SEO reports them rather than ruling on them.

Whether a conference figure is evidence about Google’s systems. One camp holds that conference talks are where Google discloses operational detail it will never write down, that Gary Illyes is among the few people who can state these numbers accurately, and that refusing the data means practitioners work with no numbers at all. The other camp holds that an unpublished slide relayed through one attendee cannot be re-read, re-checked or compared against a previous version; that no methodology, sample size, date range or definition of “typical” accompanies any figure; and that a number nobody can retrieve is not a measurement. Nobody can settle it, because Google publishes neither its conference slides nor any mapping from a stated figure to a documented one.

Where the boundary of an update goes. The spam-update row and the open dashboard window invite a site owner to date their own traffic movement to this update. One camp treats the dashboard timestamps as the only reproducible boundary; the other treats them as a publication date rather than a serving date. No published study pairs third-party volatility readings against confirmed rollout windows with sample sizes disclosed, so both argue from the same absence of data.

What Txema Hermoso thinks

This section is opinion, not reporting. The sections above are what the sources say; this is what I make of it.

The useful thing is not any single duration. It is that the shape of the pipeline is now public: discovery fast, indexing fast, refresh slow, recovery slower by an order of magnitude. That ordering matches what I see in client accounts, and it reframes most “Google hasn’t indexed my page” tickets as “Google hasn’t re-read my page” — a different problem with different fixes.

What I will not do is treat 20 hours or 30 days as a benchmark for anyone’s site. The figures carry no sample size and no definition of “typical”, and a median across Google’s entire crawl surface says little about a 200-page business site with a thin link profile. Nor will I say the conference figures are more or less reliable than Google’s documentation: that is the open question above, and it is not mine to close.

The one claim I will defend is the documentary one, because anyone can reproduce it: of sixteen reported rows, one is corroborated at the same magnitude by a dated Google page, one is contradicted in magnitude by a dated Google page, two are documented only qualitatively, and the two headline crawl figures appear in no Google document at all — including the guide Google wrote about crawling.

What is still unknown

Four things, and they are why none of this becomes a forecast.

  • Whether “typical” means a median, a mean or a modal band. No definition was disclosed, so the figures cannot be compared against anyone’s own distribution.
  • What population the figures describe. No sample size, no date range and no site-size segmentation. A number drawn from the whole web behaves differently on a small site.
  • Whether Google will publish them. The documentation changelog carried one October 2026 entry when read on 4 October, and it was about generative AI content.
  • Whether the figures are current. The core-update page dates from 2025-12-10 UTC and the crawl-budget guide from 2026-07-22 UTC; an undated slide cannot be placed relative to either.

The short version

  • Gary Illyes presented Google’s internal crawling, indexing and serving timings at Search Central Live Deep Dive Europe in Barcelona on 2 October 2026.
  • Typical figures as reported: new URL discovered in ~20 hours, known URL refreshed in ~30 days, end-to-end indexing ~1.5 hours, site move 1–3 months, core update recovery 3–6 months, spam update change 1–2 weeks.
  • Google has published none of them. The figures reached the industry through one attendee’s recap, rendered identically by two independent reports.
  • No methodology, sample size, date range or definition of “typical” was disclosed with any figure.
  • Of sixteen rows, one matches a dated Google document in magnitude (robots.txt, ~24 hours), one is slower than a dated Google document (site moves), two are documented qualitatively, and twelve appear nowhere in Google’s published record.
  • The September 2026 spam update began 2026-09-24T16:15:00+00:00 and still had no end timestamp when checked on 4 October 2026 — day 10 of a stated two-week window.
  • The actionable move is to measure your own publication-to-crawl gap rather than adopt a figure drawn from the whole web.

Frequently asked questions

Did Google officially announce these timings?

No. The figures were presented in a conference talk by Gary Illyes on 2 October 2026 and reached the industry through an attendee’s recap. As of 4 October 2026 they appear in no Search Central blog post, no documentation page and no changelog entry.

Does ~20 hours mean my new page will be indexed tomorrow?

No. The ~20 hours figure describes discovery of a new URL, not indexing, and it is a typical value across Google’s crawl surface with no sample size disclosed. The reported slowest case for discovery is “weeks to never”.

Can I use the 3–6 month core update recovery figure in a client report?

Only with its provenance attached. It is an unpublished conference figure with no sample size, and Google’s own core updates page, last updated 2025-12-10 UTC, says “several months” without a range. Reporting the range as Google’s published guidance would misstate the record.

Sources