Is It Normal for Google Rankings to Fluctuate? Three Tests

By Ugur Saritepe · October 28, 2026

Is it normal for Google rankings to fluctuate? For one page, yes, and the graph still carries information. On 9 October 2026 I was asked about one page on my own site whose line climbed in August and died in September with no edit behind it. The graph held three movements, and each needed a different test.

The question, as it was typed

game-scout.app is my own site, a catalogue of curated game lists. The owner attached a Search Console export for one list page, filtered to queries containing "detective" over the last three months, and typed this:

The session read the export, the git history and the live results page in three minutes. On 10 October I re-pulled every number below through the Search Console API for the page and for the whole site, so the figures here are the page's full series, not the "detective" subset the export showed. Data ends 6 October.

The graph, by week

Daily clicks on this page ran from 0 to 44 across the window, and the filtered view the owner looked at ran from 0 to 13. That is too small to read by day. By week, with the whole site beside it, the shape is plain:

week of     page clicks   page impr   site clicks   page share of site clicks
2026-07-06       25           661         245          10%
2026-07-13       20           664         187          11%
2026-07-20       38         1,134         230          17%
2026-07-27       54         1,457         288          19%
2026-08-03       60         1,568         361          17%
2026-08-10       69         1,694         416          17%
2026-08-17      107         2,332         520          21%
2026-08-24       95         1,884         641          15%
2026-08-31      171         2,868         891          19%
2026-09-07      170         2,937         785          22%
2026-09-14       84         1,797         562          15%
2026-09-21       31         1,013         618           5%
2026-09-28       24           848         599           4%

Three movements. A climb from July into August, when the page went from 20 clicks a week to about 100. A surge at the end of August that doubled it again for two weeks. A fall from mid-September to a quarter of the August level. The owner saw one line and asked one question. The honest answer is that the three pieces have three different causes, and the way to tell them apart is to test each against a different record.

Movement one: the climb, and the change log

Test one is your own change log."I did not change anything on page as far as I know" is a claim, and git can check it. The page's content file has eight commits. One of them is dated 17 July 2026: the H1 went from "Best Detective Games" to "Best Detective Games for PS5 & PS4", one of nine declining lists given the platform words because, as the pull request put it, 95 to 100 percent of each page's impressions came from queries that carried them.

The queries moved the way that change predicts. Two 28-day windows on the same page, both complete (73 and 120 rows, nothing truncated):

page queries, 28 days each        before (Jun 19 to Jul 16)      after (Aug 3 to Aug 30)
detective games ps5               128 impr, pos 9.9, 2 clicks     315 impr, pos 6.6, 18 clicks
best detective games ps5           65 impr, pos 8.2, 0 clicks     141 impr, pos 5.0, 15 clicks
ps5 detective games                95 impr, pos 8.7, 5 clicks     184 impr, pos 6.3, 15 clicks
whole page                      2,380 impr, pos 9.1, 64 clicks   7,478 impr, pos 7.3, 331 clicks

Position on the head query went from 9.9 to 6.6 and the page from 3 clicks a day to 12. So the owner's claim was half right. Nothing on the page changed in September. Something changed in July, and he had forgotten it, which is what a change log is for. Our tracker only started watching this page on 3 September, so for July the answer had to come from git. The attribution is by date and nothing else: the same page was broadened on 13 June (the story in the earlier post on this page) and renamed "Detective & Mystery" on 12 July, and there is no control page. The climb fits the retitle; it does not prove it.

Movement two: the surge, and the whole site

Test two is the property.The two weeks from 29 August to 13 September read like a second win for the page: 171 and 170 clicks a week against about 100 before. The site's own series says otherwise.

clicks a day            Aug 17 to 28    Aug 29 to Sep 13    Sep 15 to Oct 6
whole site                   75.2             120.9               83.1
this page                    14.7              22.9                5.5
page share of site           20%               19%                 7%

The page's share of the site's clicks held at 19 to 22 percent straight through the surge. Every list on the site rose together on 29 August and fell back together on 14 September, with no deploy on either date; the git log for that fortnight is new lists, articles and a slug migration on 12 September that changed no URL. Google's guide to traffic drops (updated 10 December 2025) says "Small fluctuations in position can happen at any time (including moving back up in position, without you needing to do anything)". This was a large one, and it was the site's, not the page's.

Movement three: the fall, and the results page

Test three is the page's own queries and the live results page. After the surge the site settled at 83 clicks a day, back where it had been. The page did not. Its share went 22 percent, 15, 5, 4. A 22-day compare of every page on the site, 15 September to 6 October against 3 to 24 August, returned 368 rows (marked truncated, but the rows sum to the click totals exactly):

clicks, 22 days each              Aug 3 to 24     Sep 15 to Oct 6
this page                              249               121
every other page together            1,118             1,707   (+53%)
pages with 20+ clicks before:  16     fell by more than half: 2 (this one, and coop-rpg 70 to 32)

Part of the other pages' gain is new supply (pages with at least one click went from 67 to 121), so the fair reading is that the site held and this page fell. The page's queries say what kind of fall. The head query "detective games ps5" held position 6.6 to 6.8 across the two windows while its impressions went from 315 to 121. The page as a whole held position 7.3 to 7.7 while impressions a day went from 300 to 127. That is not a lower rank on the searches it appears on. It is fewer searches it appears on.

The page itself checked out. URL Inspection on 10 October: indexed, fetch successful, canonical matching, last crawl 3 October. The change tracker: zero edits since it started on 3 September, two drift events (a rotating links block), six sitewide events that touched every tracked page. And the live results page for "detective games ps5" on 10 October, US mobile:

1  reddit.com         Looking for the best detective game on PS5
2  youtube.com        I Played Detective Games on PS5 and Ranked Them
3  vgnysoft.com       Detective Collection [PlayStation 5]   (store page)
4  reddit.com         same thread, second subreddit
5  facebook.com       group post          6  facebook.com   group post
7  eneba.com          17 Best Detective Games in 2026
8  playstation.com    The best puzzle games on PS4 and PS5
9  nexigo.com         Top 9 Mystery Games     10  thegamer.com   Best Detective Games On PS4 And PS5
game-scout.app: absent

Two forum threads, two group posts, a video and a store page hold six of ten slots. The same ten URLs came back on 9 October in a different order (the store page was tenth, Eneba third), which is the day-to-day fluctuation the question was really about, visible on one query in one day. Our list used to sit in that top ten. It is still ranked, by Google's own position figure, on the searches where it still appears. It is just shown less often, and the slots went to formats a list cannot become.

So, is it normal?

Three answers for three movements, and a method that generalises.

1. Check your own change log first.Git, a deploy log, or a page change tracker. A climb or dip within two to four weeks of a dated edit is probably the edit. "I changed nothing" is a hypothesis; the log tests it. Ours said the owner had changed the H1 six weeks before the climb.
2. Then compare the page to the whole property.Pull both daily series over the same window and watch the page's share of the site's clicks. A share that holds while both lines move is a site-wide movement; Google's guide lists those causes and none of them lives on your page. Ours held at 19 to 22 percent through the surge.
3. Only then read the page's queries and the live results page. Position held with impressions down is a results-page change; position down with impressions held is a rank change; both down on an edited page points back to step one. Ours held position and lost 62 percent of its head-query impressions to threads, a video and a store.

The day-to-day wobble inside all three is normal and carries no decision. The week-to-week shape does, once you know which record to hold it against.

What we have not done

We have not found a record of the two-to-three-week read that the 17 July pull request told us to make; the first read of that retitle on record is the one above, twelve weeks after the merge. And we have not established why the page lost its slots in September; the thin-page incident of 12 to 17 September was rejected as the cause on its own evidence, the September 2026 spam update was released nine days after the fall began, and the results page shows forums, a video and a store where our list used to sit. A prediction logged on 3 October (clicks back above 150 in the fortnight to 27 October after a recrawl, with the page left untouched) is still open.

What this does not tell me

• One page on one site. The three tests are the method; the numbers are one worked example, not a rate.
• The climb's attribution is by date against a change with no control page, and two earlier edits sit five and thirty-four days before it.
• The surge has no cause on record, on the site or outside it.
• The live results page is a US mobile read on one day, and it moved between two reads a day apart.
• The page compare was truncated at 368 rows. Every click in both windows is inside those rows; 199 impressions are not.
• The tracker started on 3 September. It can say nothing about July, which is why git had to.

How to run this on your own site

With Search Console connected to your assistant through the MCP server, the three tests are four calls. Pick the page and the window, then ask for the page and the property on the same days:

sc-domain:game-scout.app, dimensions ["date"], 2026-07-01 to 2026-10-10,
page equals https://game-scout.app/best/best-detective-games

98 rows, not truncated: 964 clicks, 21,663 impressions, position 7.39
2026-09-05  44 clicks   528 impressions   pos 6.16
2026-10-06   1 click     79 impressions   pos 9.87

same call with no page filter (the whole property)

98 rows: 6,687 clicks, 353,947 impressions, position 9.19
2026-09-05  182 clicks   7,545 impressions
2026-10-06   66 clicks   3,909 impressions

game-scout.app, /best/best-detective-games, kinds edit drift sitewide

trackedSince 2026-09-03   edits 0   drift 2   sitewide 6
2026-09-28  content  drift     -3 words
2026-09-06  content  sitewide  +41 words

detective games ps5, us

1 reddit.com  2 youtube.com  3 vgnysoft.com  4 reddit.com  5 facebook.com ...
game-scout.app: not in the top 10

Roll both series up by week, divide the page by the site, and read the share. Then ask the three questions in order: did I change it, did everything move, did the results page change. The free triage report covers the on-page half of step one in thirty seconds from a URL; the rest needs your own Search Console data, which is the point of connecting it.

FAQ

Is it normal for Google rankings to fluctuate from day to day?

Yes. On one page of our site daily clicks ran from 0 to 44 over three months with no change behind most of it, and Google's own guide says small fluctuations in position can happen at any time. Read the page by week, not by day, and compare it to the whole site before reading anything into one day.

How do I tell whether a drop is my page or the whole site?

Pull the daily series for the page and for the property over the same window and compare the page's share of the site's clicks week by week. If the share holds while both lines move, the movement is the site's. If the share collapses while the site holds, the movement is the page's. On our page the share held at 15 to 22 percent for ten weeks and then fell to 4.

My position is the same but impressions dropped. What happened?

The page is shown on fewer searches, not ranked lower on the ones it is shown for. Check the head query's impressions across two windows at the same position, then look at the live results page for that query. On our page the head query held position 6.6 to 6.8 while its impressions fell 62 percent, and the live top ten had filled with forum threads, a video and a store page.

I did not change the page. Why did it move?

Check that claim first against a dated record: the git history or a change-tracking log. On our page the owner's claim was true for September and not for July, where a title change six weeks earlier explained the climb. Then test the movement against the whole site, and only then against the live results page.

In your AI assistant

Ask your AI about your own site

TrafficTriage plugs into Claude, Cursor, and ChatGPT. Your assistant runs the 8 checks, reads your Search Console data, and can even check your site every morning. Free account, no API key, nothing to install.

Sign up freeSee how it works