Three pages could use some attention. You have time to work on one.
One attracts most of the clicks. Another promises practical examples but stops short of showing one task all the way through. A third looks quiet in Search Console, but you recently rewrote it.
I would start with the second page. I can name what the reader is missing and keep the work small enough to finish.
I am using saved Search Console data from my own site, the articles and their change records. The figures cover August–October 2026; I checked the content again on 22 October. This is a new assessment of historical evidence. The proposed improvement has no measured result yet.
The useful part is getting from a report to a task you can explain. A spreadsheet can produce a surprisingly long backlog without establishing why any of it deserves your Thursday.
Decide what makes a page worth improving
Before looking at traffic, name the reader and the job the page should help them do.
For a B2B team, that might mean helping a buyer evaluate an integration. For my site, it means helping marketers do practical work and showing how I approach it. All three candidates fit that purpose.
Search Console helps me find candidates. It cannot decide which reader problem matters most to my business. I look for a relevant need, a specific weakness in the current answer and an improvement I can reasonably complete.
If you need to build the candidate list first, my Search Console query-to-backlog guide covers that process. Here, I am choosing between three existing pages and deciding how much work to take on.
Compare three candidates
The figures below come from a saved page-level Google Search Console (GSC) report for nclaursen.dk, collected on 10 October 2026. Each arrow runs from 8 August–6 September to 7 September–6 October 2026: two consecutive 30-day periods, using GSC’s Pacific dates.
| Page | Clicks | Impressions | Average position | My next step for this example |
|---|---|---|---|---|
| ChatGPT prompt guide | 12 → 95 | 745 → 4,532 | 16.91 → 8.04 | Monitor; inspect the growing query mix |
| AI marketing workflows | 0 → 3 | 385 → 302 | 16.61 → 10.48 | Add one worked SEO review example |
| Retrieval Stack | 0 → 1 | 145 → 128 | 10.72 → 8.44 | Observe the recent rewrite before editing again |
The saved report does not document search-type, country or device filters. It supports this dated comparison, but I would confirm those settings before using it as a repeatable baseline.
Average position summarises appearances across searches. The workflow page did not sit at position 10.48 for every reader. Google’s Performance report overview explains the metrics and dimensions.
The prompt guide: no specific editing task yet
The ChatGPT prompt guide has the most clicks and impressions of these three pages in the later period.
I would inspect which searches account for the growth and whether the guide serves them. Reading the page already rules out one lazy recommendation: “add examples”. It has a copyable template and examples for several marketing tasks.
Its change record also documents a September revision towards template and example searches. The comparison spans that work; it does not establish that the revision caused the growth.
For now, I would monitor. I would reconsider if query review revealed a relevant task the existing examples do not cover. More traffic makes the page worth watching, but it has not yet given me a better editing brief than the workflow page.
The workflow guide: a specific gap worth addressing
The AI marketing workflows guide has fewer impressions, a better average position and three clicks where it previously had none. Those are mixed signals on small counts.
The saved query examples include “meta ads agency workflow automation tasks list”, with seven impressions in the later period, and “how to use codex for marketing”, with three in the earlier period. They suggest interest in practical tasks. They do not establish demand for an SEO tutorial.
Only five query examples are displayed per period. The later five account for 39 of the page’s 302 impressions, so they explain only a small part of its search exposure. Google also notes that filtering can omit anonymised queries and affect totals through truncation.
Reading gives me the stronger reason to act. The guide promises practical workflow examples. Its SEO section names inputs and stages, but does not take one small task from supplied files to an inspectable recommendation.
The page already explains workflow selection, tool evaluation and validation. The gap is narrower: show a marketer what to do with the files and what a defensible recommendation looks like.
That is an editorial finding. The search evidence gives it context; it does not prove that the missing example is costing clicks.
Retrieval Stack: give the rewrite an observation window
The Retrieval Stack article also fits the site’s purpose. Its low click count could tempt another round of optimisation.
But the local record documents a substantial rewrite on 3 October 2026. The current source contains the clearer passage-formatting guidance, worked examples and fragment audit described in that record. The web version retrieved during this review contains them too.
The reporting period ends on 6 October, so it mostly describes the article before that rewrite. The exact hosting deployment date remains unverified.
I would start an observation window after the confirmed publication date. An obvious factual error or broken link should still be fixed promptly. Another broad rewrite needs a reason this comparison does not supply.
Give the chosen page one task
The workflow guide wins this comparison because I can name a useful addition. The prompt guide has no equally specific brief, and Retrieval Stack needs time after its rewrite.
The SEO section currently moves through keyword clustering, opportunity review, briefing, drafting, QA and a refresh recommendation. I would add one compact example beside that flow:
- Inputs: current and previous query exports filtered to one page, separate page totals, the article text and a short audience note.
- Review: compare one relevant reader need with the passage already addressing it.
- Output: one proposed edit, the evidence supporting it and a reason to leave the rest alone.
The completion check: can a marketer see which files to supply, what to inspect and how the recommendation follows from the page?
I would allow an initial two-hour editing slot. That is an estimate, not recorded effort. The task excludes a full automation tutorial, new tool setup and a rewrite of the other workflows. If it needs that much explanation, I would link to a separate tutorial from a short example in the guide.
The hypothesis is that this addition will help a reader turn the overview into a task they can perform. I would check that usefulness first. A search lift is something to investigate later.
Does the choice still hold?
On 22 October, the specific worked example was still absent from the local workflow guide and the web version retrieved for this review. The figures above remain the historical example.
A separate GSC check compares 19 September–18 October with 20 August–18 September 2026. The prompt guide has 50 clicks in each period, so its earlier click growth is not continuing in this comparison. The workflow guide still has three clicks. These sparse figures do not give me a stronger editing task elsewhere.
Record what would change your mind
Before implementing the addition, check the latest page and save a fresh baseline. The workflow guide has already had several edits; the historical comparison cannot isolate their effects.
For your own review, open Search results → Pages, filter to the candidate URL and inspect Queries. Compare completed periods of equal length and record the property, search type, countries, devices and other filters. Use a longer window where sparse data or seasonality makes a short comparison misleading.
Save page totals separately from the query group you intend to follow, including the rule for selecting that group. Record the exact edit and verified publication date.
I would set a first review roughly six weeks after publication. That is a diary reminder, not a promise of enough evidence. With three clicks in the historical period, I would expect to need patience.
At review, check whether the example is usable, the relevant search mix changed or other edits complicate the comparison. Movement alone will not show that the addition caused it. Review relevant on-site actions separately if analytics are available: Search Console and Analytics describe different parts of the journey. These query rows do not identify which visitor later submitted a form.
If the example is now covered elsewhere and a link serves the reader, reduce the task to that link. If closer review finds no useful gap, record no change. Neither decision needs an apologetic rewrite attached.
A worksheet for your next page review
Copy this once per candidate. Compare three completed sheets before choosing what gets your time.
| Field | What to record |
|---|---|
| Page and purpose | URL, intended reader and the task it should help them complete |
| Search evidence | Both periods, filters, page totals and relevant query examples; keep their scopes separate |
| Content finding | The passage you checked and what is missing, unclear or already sufficient |
| Change history | Recent edits and their verified publication dates, or what is unknown |
| Decision and reason | One bounded edit, investigation, monitoring or no change; explain why it deserves attention before the other candidates |
| Scope and effort | What is included, what is excluded and your time estimate |
| Baseline and review | Saved metrics, exact change, publication date and first review date |
| Result and next decision | What happened, what remains uncertain and what would make you revise the decision |
For this example, I would add the small workflow example, monitor the prompt guide and give Retrieval Stack a later observation window. Your choice may be different. The useful test is whether you can explain the next task, why it beats the alternatives and what would make you leave the page alone.
Recommended next
More writing connected by category or shared topics.