TrafficTriageOrganic Traffic Flat Despite Publishing? Split It by Page
By Ugur Saritepe · October 28, 2026
Organic traffic flat despite publishing new content? Split the flat line by page age before you publish more. On game-scout.app a fortnight that read 1,188 clicks against 1,255 was, page by page, 203 clicks lost and 270 gained. The 44 pages published since September added 50 of them. The lists that carry the site netted minus 13.
Where this came from
game-scout.app is my other site, a catalogue of curated PlayStation lists and game guides. Its weekly Google clicks climbed from 361 in early August to 891 in the week of 2026-08-31, then settled: 562, 618, 599 and 611 in the four full weeks from 09-14. Over the four weeks to 10-16 the repository took 89 merged pull requests and 44 new list and article pages. On 2026-10-18 I asked the assistant I work with what was wrong, and this is what I typed:
I need to focus on our seo guide from scratch again. What is missing right now in our flow ? Why did we get stuck ? What to do for next steps ? Should I try seo apps ? How can we improve our app right now ?
The answer was that the total was two flows cancelling out: new lists maturing, older head pages decaying. That is a claim with no windows attached, so the next morning I re-pulled it on two fixed fortnights and dated every page by the commit that created it.
The flat line
sc-domain:game-scout.app, dimensions ["page"], 2026-10-03 to 10-16 against 2026-09-19 to 10-02, rowLimit 250
clicks 1,188 > 1,255 (+67) impressions 75,096 > 73,959 position 8.27 > 8.53 rows 291 (truncated)
Plus 67 clicks on 1,188 is under 6%. Read as a total, nothing happened. The totals are exact even though the rows are cut at 250 per window; the rows below therefore undercount the long tail of every group, and the gross figures are lower bounds.
The same fortnight, by page age
Each list and article page was dated by the first commit that added its content file. Catalogue pages, hubs and the homepage are kept apart because they have no publish date of their own.
created pages clicks change losing gaining before August 68 969 > 956 -13 28 28 August 19 114 > 141 +27 8 8 Sept 1 to Oct 15 44 62 > 112 +50 5 24 hubs and indexes 57 29 > 18 -11 catalogue pages 71 0 > 5 +5 all rows 291 1,188 > 1,255 +67 57 78
The established lists did not decline as a group. They churned: 28 lost 147 clicks and 28 gained 134. Across every row, 57 pages lost 203 clicks and 78 gained 270, three times the net change in each direction. The new pages are the one group that moved one way, and they moved slowly: 112 clicks from 44 pages is 1.3 clicks a page a week, and 17 of the 44 earned nothing in the later fortnight. Forty more list and article pages created in the same period have no row at all.
What the losers lost
The eight biggest losers were all lists created between 2026-06-06 and 07-05, the oldest on the site. Together they went from 421 clicks to 319 and from 15,439 impressions to 13,275.
list clicks position impressions space 82 > 52 7.50 > 7.45 3,131 > 2,607 trucking 90 > 71 6.74 > 6.99 4,360 > 4,198 detective 67 > 51 7.69 > 8.03 2,039 > 1,894 medieval fantasy 21 > 11 8.32 > 8.29 1,708 > 1,257
Five of the eight held their position within half a place. What they lost was impressions and click-through, spread thin: the space list's 30 lost clicks came from about twenty queries of one to seven clicks each, its biggest, "space games ps5", going from 7 clicks to 2 at a reported position that improved from 7.3 to 6.1. A position that holds while clicks fall points away from the page and toward the results page around it, so I read that live:
"space games ps5", US, mobile
Reddit, PlayStation Store, GamesRadar, YouTube, YouTube, Reddit, Facebook group, GameFAQs, PlayStation Store, space.com game-scout: not in the top 10
Four forum threads, two videos and two store pages. A reported average of 6.1 and an absence from one US mobile read are not a contradiction; the reported figure is an average over every location and day the page did appear. It is one query on one morning, and the device split of the same site four days earlier found game-scout absent for the space and detective head queries then too. It does not tell me why these lists lost, only that the page itself did not move.
What the ranked pages say
I read the results for this question on 2026-10-19. The exact phrasing returns forum and Quora threads asking it, a LinkedIn post saying most content gets no traffic, and two publisher pages about AI search. On "organic traffic plateau" the top result, dated 2025-08-26, gets the method right: look at how recently published pages perform, and watch your top pages for decline. It shows no numbers. A framework post from 2026-10-07 carries one client's year-on-year totals and no page split. No ranked page I assessed on either phrasing puts the gross loss and gain next to the net, or measures what a new page adds in its first weeks.
What we have not done
We have not worked the losers. The test decided on 2026-10-03 for exactly this group, a page-format change on the co-op RPG list with the copy held fixed, has not shipped, so its 2026-10-27 readout cannot land. The comparison routine graded failed the same day still runs. In the four weeks this post measures, the effort went into the side of the sum that adds 1.3 clicks a page a week. This read also stops at two fortnights on one site; with a truncated page pull, a longer window would need the full export rather than our tool's 250-row cap.
How to split your own flat line
- Pick two equal windows inside the plateau. Not one that straddles a peak, or the split will read as a drop. Ours were the two fortnights to 10-02 and 10-16.
- Export the Pages table for both and add a publish date per page. Your CMS or your repository has it. Leave hubs, tag pages and the homepage in their own group.
- Per age group, sum the change, the losses and the gains separately. Good: the oldest group is flat with few losers. Bad: the gross losses are several times the net, as ours were.
- Divide the newest group's clicks by its pages and weeks. That is what one more page buys you right now. Ours was 1.3 a week.
- For the five biggest losers, read position and impressions. Position steady with impressions down: check the live results page for the head query. Position falling: read pages lost rankings but nothing changed for the per-query check. Plenty of impressions and no clicks at a first-page position: ranking on the first page but no traffic.
The rest of these measured reads are in the AI-assistant SEO series.
FAQ
Why is my organic traffic flat even though I publish every week?
Usually because two things are happening at once and cancel out: new pages are adding a little each, and some older pages are losing. On game-scout.app a fortnight that read 1,188 to 1,255 clicks was 203 clicks lost and 270 gained across pages. The 44 pages published since September added 50 clicks between them. The established lists netted minus 13 on 28 losers and 28 gainers.
How many clicks does a new page add in its first weeks?
On our site, 44 list and article pages created from 2026-09-01 to 10-15 earned 112 clicks in the fortnight to 10-16, about 1.3 clicks a page a week, and 17 of them earned none. Another 40 pages from the same period had no row in the pull at all. A new page is a small, slow addition, so a few older pages losing ground can erase a month of publishing.
How do I find which pages are holding my traffic back?
Compare two equal windows in the Search Console Pages table, export it, and add a column for when each page was published. Sum the clicks change per age group, and separately sum the losses and the gains. If the gross losses are several times the net change, the flat line is hiding churn, and the losers in your oldest group are where to look first.
Should I publish more content to break a traffic plateau?
Not as the first move. If new pages add about one click a week each and your older pages are losing more than that, more publishing raises the floor slowly while the losses keep running. Find out what the biggest established losers lost: impressions at a steady position point at demand or the results page, a falling position points at the page or its competitors.
The split above is one call of one tool and a join. Connect Search Console to the TrafficTriage MCP server and ask your assistant for the page dimension over two equal windows. It returns exact totals and up to 250 rows per window; your assistant can then join them against your publish dates. It does not know your publish dates on its own, and it does not store results pages over time.
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.