Nine daily numbers in order: how to tell an indexing event from a growth curve

Read as a total, our first nine days are 5,965 impressions. Read in order they are 5, 4, 6, 7, 19, 84, 1,100, 2,685 and 2,055, and that is not a growth curve. It is an indexing event with a visible before and after, and the two readings support different claims. The daily series is published on our proof page for aiknowsus.com, window 2026-10-04 to 2026-10-12, export taken 2026-10-14. Anyone can read it across rather than down, which is the only way to see what actually happened.

The three phases in nine numbers

A growth curve rises in proportion to effort. This does not. It has three distinct phases and the boundary between the second and the third is a step, not a slope.

Days	Impressions	What the shape is
4 to 7 October 2026	5, 4, 6, 7	Flat. A handful of pages are known and almost nothing is being served
8 to 9 October 2026	19, 84	The first real movement, still in double figures
10 to 12 October 2026	1,100, 2,685, 2,055	A step to a new level, then a settle just below the peak

Nothing about the publishing effort changed between 9 and 10 October. What changed was on Google's side: enough of the site had been crawled and processed to be eligible across a far wider set of queries, and eligibility arrived in one piece rather than in a trickle. That is a mechanical event, and it is the single most useful thing the daily view tells you.

Why the total hides it

5,965 across nine days averages to a tidy figure that occurred on none of the nine days. Four of the days are single digits. One day alone carries 2,685. An average over a series that contains a step describes neither side of the step, which is why a total is the wrong unit for the first weeks of a domain and the right unit later, once the daily figures are of a similar order.

The same arithmetic is why a chart of the first nine days looks dramatic and a chart of the first ninety would not. The step is enormous relative to the flat part before it and ordinary relative to anything that comes after.

What to read instead of the slope

Four readings come out of a daily series that no total produces:

• Where the step is. The date the series changes level is the date to line up against your own crawl and indexing records.
• How long the flat part lasted. Four days at single digits is the gap between publishing and being served, and it is the number worth shortening next time.
• Whether the new level holds. One day above the old level is an event. Three days above it is a level.
• What the peak costs you in interpretation. A series that ends on its highest day invites a reader to extend the line. Ours ends one day after the peak, which removes that temptation.

The queries behind a step like this

A new domain does not become eligible for contested queries in nine days. It becomes eligible for specific, lightly contested ones, which is where the first impressions of any new site come from. That is worth knowing before anybody reads the step as evidence about commercial terms, and it is the reason the first hundred pages on a new domain should be aimed at narrow questions with clear wording. The test for whether a phrase is narrow enough to be worth a page of its own is set out in what a long tail keyword is and how to tell one worth a page.

How to take this reading on your own site

Open the Performance report, set the date range to the whole life of the site rather than a rolling window, and export the daily rows instead of reading the chart. Then write the dates down the page and the impressions beside them. Four things to hold to while you do it: use days and not weeks, because a weekly bucket can hide a step entirely; keep the export date beside the series, since recent days are revised; do not smooth the line until the series is long enough that daily noise is the problem rather than the signal; and resist averaging across the step.

Questions about reading a daily impression series on a new site

How do I know a step is indexing and not something I did?

Check what you published in the days immediately before it. A step caused by publishing arrives gradually, page by page, because each page becomes eligible as it is processed. A step caused by the site crossing a crawl threshold arrives at once across pages that were already live.

Does the total belong in the report as well as the series?

Both, with the series above it. The total is the headline and the series is the evidence, and a reader who sees four single digit days before the step will believe the rest of the report more, not less.

Is a drop the day after a peak a problem?

Not at this stage. The 2,055 that follows 2,685 in our series is a new level settling rather than a decline, and a single day either side of a peak is too short a window to read as a direction.

What if the flat part never ends?

Then the issue is indexing rather than patience. Check coverage, check that the pages are reachable from somewhere other than the sitemap, and check the first response really contains the text a reader sees.

Is there an equivalent series for AI assistants?

There is no daily row to export, so the record has to be made rather than read. Our own readings of that surface cover ChatGPT and Gemini, 597 usable readings on 15 October 2026 and 1,260 on 16 October 2026, and neither run says anything about the Google side of the same domain.

Why we now export the daily rows on every new domain

This series changed how we start. On any domain in its first weeks we take the daily export rather than the chart, from day one, and we keep it even when the numbers are embarrassing, because the four single digit days at the front are what make the step legible later. Without them the step is just a high number with nothing to be high against. The practical rule we hold to is that the first export happens before there is anything worth exporting, since a baseline taken after the interesting thing has happened is not a baseline. That is also the honest reason our own first nine days are published at all: we are paid to take this measurement for other people, and the first set we had to take was ours.