Search Console Audit, Pass Five: Why Page-1 CTR and Page-3 Relevance Need Different Fixes

By Rakshit Yadav (@yadavrakshit60)•Oct 2026•9 min read

Short answer: Pass five of our own Search Console audit surfaced a pattern we'd been missing by only looking at impressions and position: a page already on page 1 can still be losing, because 0% CTR at position 9 is a different problem than 0% CTR at position 67, and a single page can be quietly splitting its own ranking across two queries instead of consolidating them. Four rows this time — /skills (49 impressions, position 67, 0% CTR), /compare/cursor-rules-vs-skills (13 impressions, position 29.1, 0% CTR), /commands (6 impressions at position 13.7 for "cursor slash commands," plus a second, separate ranking at position 31.8 for "cursor commands"), and /blog/best-cursor-commands (already page 1 at position 9, 4 impressions, 0% CTR). The first two repeat the "deep position, zero clicks" shape from pass one and pass four. The last two are new shapes, and they're the actual point of this pass.

If you've only ever treated Search Console data as "sort by impressions, fix the title," this pass adds two rules that pass one through four didn't need: a page-1 position with 0% CTR is a listicle-promise problem, not a relevance problem, and a page ranking for two related queries at two different positions is often one page that should consolidate both intents instead of picking one.

The four rows, as pulled this time

PageQueryImpressionsPositionCTRProblem type
/skillscursor skills49670%Deep-position relevance gap (repeat shape)
/compare/cursor-rules-vs-skillscursor rules skills docs1329.10%Near-miss, docs-seeking intent (repeat shape)
/commandscursor slash commands / cursor commands613.7 (page 2) and 31.8 (separate query)0%Split ranking across two queries on one page
/blog/best-cursor-commandscursor commands490%Page-1 CTR problem, not a ranking problem

Why page 1 at 0% CTR is a different fix than page 3 at 0% CTR

Every prior pass in this series treated "0% CTR" as one problem with one fix: tighten the lede, add a table, match the query in the H1. That logic holds when the page is sitting at position 30-70, because Google hasn't fully committed to showing the page yet — relevance is still in question. Position 9 is a different situation. /blog/best-cursor-commands is already on page 1 for "cursor commands," which means Google has already decided the content is good enough to compete with the top results. Zero clicks on 4 impressions at that position means the searcher saw the result and chose something else in the list above or below it — a SERP-snippet decision, not a relevance one.

The fix for a page-1, 0%-CTR row is the title tag and meta description, not the body copy: give the result a reason to be clicked over its page-1 neighbors. We rewrote the title from a generic "Best Cursor Commands" framing to a numbered, dated listicle promise — "17 Best Cursor Commands to Save Time (2026)" — and sharpened the meta description to name a concrete benefit (saved keystrokes on repeatable workflows) instead of restating the title. We also added a comparison callout midway through the post pointing at the Claude Code commands equivalent, so a reader comparing the two ecosystems has somewhere to go without us diluting the Cursor-commands framing the page is actually ranking for.

Why /commands was ranking for two queries at two different positions

The Search Console row for /commands wasn't one line — it was two: "cursor slash commands" at position 13.7 (borderline page 2) and "cursor commands" at position 31.8 (page 4). Same page, two related queries, two different outcomes. That's a sign the page's structure answers one intent clearly and the other only incidentally. Our /commands page opens by describing the Cursor slash-command catalog, which explains the stronger position on "cursor slash commands," but it never explicitly separates out a "Claude Code commands" equivalent section, so a visitor (or a crawler) looking for the broader "cursor commands" framing has to infer the scope from context.

The fix: restructure the page so the H1 and opening paragraph answer "cursor slash commands" first — since that's the stronger-performing query — then add a clearly labeled "Claude Code commands" section directly below it with its own mini-table, rather than treating Claude Code only as an aside. One page, two explicitly labeled sections, each matching one of the two queries it's already ranking for. We also added internal links from that section into AgenticKit's Claude Code commands content so the two intents reinforce each other instead of competing for the same H1.

What we actually changed, page by page

  1. /skills — Tightened the title and meta to a direct, numbered "Cursor Skills" framing matching the literal query, added an FAQ block answering "what are Cursor skills" (snippet-eligible, plain-language, no jargon), strengthened internal links from Best Cursor Skills and the ranked 25 Best Cursor Skills in 2026 listicle into the catalog page, and added a short comparison callout to Claude Code skills without displacing the Cursor-first framing.
  2. /compare/cursor-rules-vs-skills — Tightened the H1/title to closely match "cursor rules skills docs," confirmed the decision table sits near the top (it already did, per pass four), and extended the comparison with a "how Claude Code skills handle this" column, since the page already discusses the shared SKILL.md format both tools use.
  3. /commands — Restructured the H1 and intro to directly answer "cursor slash commands" first, added a clearly labeled Claude Code commands section with its own comparison table below it, and linked that section to AgenticKit's Claude Code commands content.
  4. /blog/best-cursor-commands — Rewrote the title to a numbered, dated listicle promise, sharpened the meta description with a concrete benefit, and inserted a comparison callout box midway through linking to the Claude Code commands equivalent content.

The two rules this pass adds to the series

  1. Position tells you which lever to pull. Below roughly position 20, pull the relevance lever (H1, lede, structure). Above roughly position 10 with 0% CTR, pull the SERP-snippet lever (title tag, meta description, listicle framing) — the content has already won the relevance argument, and the only thing left to fix is the one line of text the searcher actually reads before clicking.
  2. A page ranking for two related queries at two positions is a consolidation problem, not two separate pages to build. Before splitting content across two URLs, check whether one page can answer both intents explicitly — a primary section for the stronger query, a clearly labeled secondary section for the other — rather than hoping one H1 covers both.

A checklist to add to the series

  1. For any row at position ≤10 with 0% CTR: the fix is the title and meta, not the body. Add a number and a year to listicle titles; name a concrete benefit in the meta description.
  2. For any row at position >20 with 0% CTR: the fix is the H1, lede, and structure — the content hasn't yet convinced Google it's the best answer, let alone the searcher.
  3. If Search Console shows the same URL against two related queries at different positions, check whether the page currently serves one intent clearly and the other only by inference. If so, add an explicitly labeled section for the weaker query instead of building a second page.
  4. Re-pull the same rows in 2-3 weeks, same as every prior pass in this series — position and CTR at low impression counts stay noisy, and a page-1 title change can take a cycle or two to register in Search Console's own data.

FAQ

Is 0% CTR always a title-tag problem? No. Below roughly position 20, it's usually a relevance problem (Google hasn't fully committed to the page yet) and the fix is the H1 and lede. Above position 10, relevance is already settled, so 0% CTR is almost always the title tag and meta description failing to win the click against competing results.

Why would one page rank for two different queries at two different positions? Because the page's structure answers one query intent explicitly and the other only incidentally. The fix is usually to add an explicitly labeled section for the weaker query rather than assuming it needs a separate page.

Should /blog/best-cursor-commands and the Claude Code commands content be merged into one page? No — they target different ecosystems and different primary intents. The right move is cross-linking and a labeled comparison callout, not a merge, so each page keeps ranking for its own query while still serving the adjacent intent.

How is this pass different from pass four? Pass four's new row (/compare/cursor-rules-vs-skills) was a near-miss at position 29 — a relevance/framing fix. This pass's two new rows are a page-1 CTR problem and a split-ranking problem, which need a different fix (snippet copy, section structure) than anything the series had covered before.

Sources

A sixth pass re-pulled these same two rows and found they hadn't moved yet — the fix needs a full re-crawl cycle, not 48 hours — and shipped a separate cannibalization and internal-linking fix on two other pages. See pass six.

Written by Rakshit Yadav. Published 2026-10-31, last updated 2026-10-31.

Related Tools & Agents

🛠️ Free Tool: rules-grader🤖 Agent: seo-specialist⚡ Command: /seo-auditSkill: on-page-seoSkill: project-conventions

Run this checklist automatically

The on-page SEO skill and /seo-audit command apply this exact method to any page you point at.

See the on-page SEO skill →