The last few days in Search Console change because the newest data can be preliminary: Google says it can still change within hours, and collected data is usually available in 2–3 days. None of the Google pages we read gives a figure for when a day becomes final, so we treat the newest 3 days as unsettled.
- Google says the newest figures in the Performance report can be preliminary, and a dotted line on the chart marks them (Google, Search Console Help, read 30 October 2026).
- Search Console Help puts the usual wait for collected data at 2–3 days, and Analytics Help puts it at 48 hours (Google, Search Console Help and Analytics Help, read 30 October 2026).
- The Search Console API can return finalized data only or data that includes fresh data, and it can name a first incomplete date after which values may still change noticeably (Google, Search Console API reference, read 30 October 2026).
- The Performance report labels daily data in Pacific Time in every view except the 24-hour view, so a Search Console day is a Pacific day (Google, Search Console Help, read 30 October 2026).
- Penlid treats the last 3 days of every Search Console export as possibly preliminary, because a file does not record when it was exported (Penlid, Search Console history merger, read 30 October 2026).
Why does the newest data in Search Console change?
The newest data in Search Console changes because Google shows it before it is final, and Google says such data might change in the next few hours
. Google's About the data page adds that the preliminary part is usually today's data and sometimes yesterday's.
When Google announced the 24-hour view in December 2024, it said Search Console shows a data point as soon as there is any data for it, even if collection for that point is unfinished, and marks it with a dotted line. Our reading: an early figure is a draft of the day, not a fault.
Google's pages give two timescales: hours for preliminary data, and a lag of days for collected data. They do not reconcile the two, and none states which way a preliminary day moves. Penlid's Search Console history merger uses the top of the 2–3 day range, 3 days, as its margin.
How can you tell which days are not final yet?
The dotted line on the Performance chart marks preliminary data and the API's first incomplete date marks it in code, but a downloaded file does not record when it was exported, so for files you must apply a margin yourself.
On the chart, Google says the report shows complete days by default. In the API, the dataState setting chooses between finalized data only and data that includes fresh data, and the response metadata can name a first_incomplete_date: Google says every value after that date may still change noticeably.
The history merger's page notes that a file does not record when it was exported, so nothing in it tells you which of its newest days were preliminary when you pulled it. That is why the last 3 days of every export are treated as possibly preliminary
there, and why the Search Console anomaly check never uses the newest day in a file.
When is a Search Console day final?
Google's pages that we read give no figure for when a Search Console day becomes final; they give lags and warnings, which the table sets beside the two margins Penlid uses.
| Page | What it says about timing | Whose rule |
|---|---|---|
| Search Console Help, overview of the report | Preliminary data can change within hours | |
| Search Console Help, data discrepancies | Collected data is usually available in 2–3 days | |
| Search Console API guide | Data is typically available after 2–3 days | |
| Analytics Help, linking Search Console | Data is available 48 hours after it is collected | |
| Search Console API reference | Values after the first incomplete date may still change noticeably | Google, no day count |
| Penlid history merger | The last 3 days of every export are possibly preliminary | Penlid's margin |
| Penlid SEO test calculator | Read a result 28 + 7 + 3 = 38 days after a change goes live | Penlid's margin; the 7 is ours |
These are lags and margins, not a promise that a day stops changing. Google's September 2019 announcement of fresher data in the Search Performance report said that each fresh data point will be replaced with the final data point after a few days
; none of the help pages we read repeats the line, and a few days gives no figure to count from. A day can also be corrected after it looks settled. Google's table reference for the bulk export to BigQuery says Google may have to revise previously saved data because of a processing error, though it adds that fixing old data does not happen often. Google's data anomalies page records missing data for 13 to 17 August 2026 (Generative AI in Search) that was restored in an update dated 21 August, four days after the last affected day (21 − 17 = 4). See also the guide to Search Console data anomalies.
Which time zone decides what a Search Console day is?
Pacific Time decides it. Google says the Performance report labels daily data in Pacific Time in every view except the 24-hour view, which shows your local time from your browser's settings, and the API reads a start date in Pacific Time too, given in its reference as UTC minus 7 or 8 hours.
Midnight Pacific is therefore 07:00 or 08:00 UTC (our arithmetic from the reference's figures), so a Search Console day runs from that hour to the same hour the next day. If you are east of the Pacific, the report's newest day ends hours after your own calendar day does, so a margin counted from your own midnight starts too early (our reasoning, not a Google rule).
What should you do with the newest days?
Keep them out of every comparison until they have settled, then read them again. Four habits help.
- End every window on a settled day. Use a margin: the history merger's is 3 days, and the anomaly check sets aside the newest day in the file. Both are our margins; Google gives no figure.
- Measure your own site's settling. Our suggestion: export the same date range twice, a few days apart, and compare the overlapping days. The history merger counts any difference in clicks or impressions between two exports as a conflict, which shows how far back your revisions reach. For keeping exports past the 16 months Search Console shows, see the guide to the 16-month limit.
- Wait before judging a change. Penlid's SEO test calculator reads an answer
28 + 7 + 3 = 38 days after the change goes live
: a 28-day after window, 7 days that are our own allowance for the newest days to settle (not a Google figure), and about 3 days of reporting lag. The guide to how long to wait after a content update covers the cases. - Give the did-it-work tool settled dates. The free did-it-work tool needs
the 56 days before each change and the 28 days after it
, and its page describes no handling of preliminary days, so end the later range on a day you trust. Because the tool sets your page's ratio against untouched pages from the same file (see the methodology), a shortfall shared by every page largely cancels and one on your page alone does not; that is our reasoning, not a measured result.
Example (illustrative numbers)
These counts are invented to show the arithmetic; they are not data from any site. They assume the early read is low, as it would be if collection were unfinished. The restatements are made large enough to see in the arithmetic; Google's 2019 announcement expected fresh data to change a bit
before it is final. Google does not say which way a restated day moves, and two exports of your own will show it.
| Day | Previous week | This week, first export | This week, second export |
|---|---|---|---|
| Friday | 86 | 93 | 93 |
| Saturday | 61 | 66 | 66 |
| Sunday | 58 | 63 | 63 |
| Monday | 97 | 101 | 101 |
| Tuesday | 94 | 103 | 109 |
| Wednesday | 91 | 83 | 106 |
| Thursday | 85 | 58 | 98 |
| Whole week | 572 | 567 | 636 |
Read on the Friday, the week looks flat: 567 ÷ 572 = 0.991, which is −0.9%. Read the next Wednesday, the same week is 636 ÷ 572 = 1.112, which is +11.2%. Nothing happened to the page in between; Tuesday to Thursday were restated (103 to 109, 83 to 106 and 58 to 98).
Now set aside the last 3 days of the first export, as the history merger does. Four days remain: Friday to Monday, 323 clicks this week against 302 the week before, and 323 ÷ 302 = 1.070, which is +7.0% in both exports. That is not the final answer, which is +11.2%, but it is the part you can already trust, and it does not change when you export again.
In a 28-day after window the unsettled days weigh less. If a page earns the same clicks each day and 3 of the 28 days read 30% low, the total is short by 3 × 0.3 ÷ 28 = 0.9 ÷ 28 = 3.2%, before anything has really changed.
What this cannot tell you
- When your own days settle. Google gives no figure, and our 3-day margin is a convention built on Google's 2–3 days, not a measured property of your data. A day outside the margin can still be corrected, as the August 2026 restoration shows.
- Which way a restated day moves. The Google pages we read do not say, so the example assumes a rise only to show the arithmetic; your own exports will show the real direction.
- Whether two numbers for one day differ because of settling. A table that stops at 1,000 rows, anonymized queries and filters also make two views of one day disagree. The guide to the export row limit covers the first of these.
- How often unsettled days distort a real comparison. The error rates on our methodology page come from simulated sites where the true answer was known; nothing there measures what preliminary days do to a real export.
Questions people also ask
How often is Search Console data updated?
Google's pages that we read give no refresh schedule. They say figures for the newest days arrive while still preliminary and can shift within hours, and that collected data usually lands after two to three days. In December 2024 Google described the 24-hour view as appearing with a delay of only a few hours, which fits, in our reading, a pattern of early figures corrected as collection finishes.
How accurate is fresh data in Search Console?
No accuracy figure for fresh data appears on the Google pages we read. Google marks data that is still being collected as preliminary, and its API reference warns that anything after the first incomplete date may still change noticeably. To see how far your own recent days move, export one date range twice, a few days apart, and compare the days the two files share.
Why is the most recent day missing from Search Console?
The report shows complete days by default, so a day that is still being collected may not be drawn. In the API, days with no data are left out of the results. The report's days are also Pacific days, so if you are far ahead of Pacific Time, your newest local date may not have begun in its calendar. If a whole stretch is absent, check Google's anomalies page, or our guide to data anomalies, for a recorded issue before assuming a delay.
How long is the Search Console data delay?
Google's figures differ by page: 2–3 days in Search Console Help, typically 2–3 days in the API guide, and 48 hours in Analytics Help. Two of the three say usually or typically, and a recorded data issue can leave days incomplete for longer, as Google's August 2026 anomalies entry shows. For planning we use 3 days, which is our own margin rather than a Google rule.
Sources
- Performance report (Search results): Overview and basic setup, Google, Search Console Help, read 30 October 2026.
might change in the next few hours
- Performance report (Search results): About the data, Google, Search Console Help, read 30 October 2026.
- Performance report (Search results): Troubleshooting data discrepancies, Google, Search Console Help, read 30 October 2026.
- Performance report (Search results): Dimensions and data groupings, Google, Search Console Help, read 30 October 2026.
- Connect Search Console to Google Analytics, Google, Analytics Help, read 30 October 2026.
- Search Analytics: query, Google, Search Console API reference, read 30 October 2026.
- Getting your performance data, Google, Search Console API guide, read 30 October 2026.
- Search Analytics, Google, Search Console API reference, read 30 October 2026.
- Table guidelines and reference (bulk data export), Google, Search Console Help, read 30 October 2026.
- Data anomalies in Search Console, Google, Search Console Help, read 30 October 2026.
- An improved way to view your recent performance data in Search Console, Google, Search Central Blog, read 30 October 2026.
- Fresher data in your site's Search Performance report, Google, Search Central Blog, read 30 October 2026.
will be replaced with the final data point after a few days
might change a bit before being finalized
- Methodology: how we measure, and what we can't, Penlid, read 30 October 2026.
- Did that content change actually work?, Penlid, read 30 October 2026.
the 56 days before each change and the 28 days after it
- Search Console history merger, Penlid, read 30 October 2026.
the last 3 days of every export are treated as possibly preliminary
- Is this drop real? Search Console anomaly check, Penlid, read 30 October 2026.
- SEO test calculator, Penlid, read 30 October 2026.
28 + 7 + 3 = 38 days after the change goes live