Rank Tracker: the Claude skill that tracks rankings honestly
Rank Tracker is a free Claude skill that turns two Google Search Console exports into a rank tracking system. It compares periods, classifies every mover, weights by impressions, and explains the average-position artifacts that fool people. Claude cannot see live SERPs, so GSC position data is the truthful source. One curl command installs it.
Install it
One command, then paste two equal-length GSC query exports.
curl -o ~/.claude/skills/rank-tracker.md https://raw.githubusercontent.com/lhitches/claude-seo-skills/main/skills/rank-tracker.md
Then: "Track my rankings, here are two GSC exports." No install? Paste the skill file contents into any Claude chat and the same process runs.
What it does, and what you get
Rankings tell you where you stand. Two periods tell you where you are heading, and heading is what you actually want to know.
- Two-period compare: current and prior GSC query exports of equal length, every query matched across both.
- Seven movement classes: big win, win, flat, slip, big slip, new, lost, each with a clear threshold.
- Impression weighting: a 5-position jump on 12 impressions is a footnote, a 1-position slip on 4,000 impressions is the headline.
- Priority keywords first: name the queries you care about and they open the report, even the flat ones.
- The caveats section: average-position artifacts, unequal periods and low-data queries called out honestly, never buried.
Real sample output
This excerpt is a genuine run against lawrencehitches.com using real SEOtesting query data, two equal 28-day periods, 4,076 queries matched. It caught an artifact most tools would report as a loss.
RANK TRACKING REPORT: lawrencehitches.com
Periods: two equal 28-day windows · 4,076 queries matched · impressions grew 33,040 to 113,338.
- Biggest clean gain: "enterprise seo platform" up 11.0 positions on steady impressions (341 to 393). The whole enterprise cluster moved together, which is how you tell a real rank change from noise.
- The artifact it caught: the top "slip" was fake. A referral query at 11,097 impressions "dropped" 1.1 positions only because its prior average was built on exactly 2 impressions. Visibility actually grew 5,500x. A win wearing a slip's clothing.
- The real story: 74% of the site's impression growth was claude.ai referral-operator noise. The underlying site was close to flat. The skill said so plainly instead of reporting a 3.4x win.
- On-topic movers: "claude seo skills" up 8 positions with impressions up 6x and 4 clicks from position 54, the highest click count of any mover.
Excerpt from a real run, 11 July 2026. Full format: priority keywords, top movers up and down, new and lost, watchlist, caveats, and the next recheck date.
When to use it, and when to use a sibling skill
Use Rank Tracker for the trend. Pair it with its siblings for the fixes.
| You want | Use | Why |
|---|---|---|
| To know if your rankings are heading up or down | Rank Tracker (this skill) | Two-period compare with artifacts explained, no live SERP scraping |
| The pages one small push moves onto page one | Striking Distance Finder | Deep dive on the positions 8 to 20 watchlist |
| Whether a Google update helped or hurt | Winner or Loser | Built for the update window specifically |
| Which pages to fix first, right now | GSC Quick Start | Ranks the highest-value pages to fix first from one export |
FAQ
Does Rank Tracker scrape live search results?
No, and any skill that claims to is selling noise. Claude cannot see live SERPs. Rank Tracker uses Google Search Console average position, which comes from real impressions shown to real searchers, so it is the most truthful rank data a site owner has. It only covers Google, and it says so.
Why weight everything by impressions?
Because position change without volume is a mirage. A five-position jump on twelve impressions changes nothing, while a one-position slip on four thousand impressions is the day's headline. The skill ranks every mover by impressions so the report opens with what actually matters.
What is an average-position artifact?
When a query starts showing on more results, or in more locations, its average position can fall even as total visibility grows. The prior average may also have been built on a handful of impressions, making it a fake baseline. The skill flags these every run, so you never chase a loss that is really a win.
How often should I run it?
On a fixed cadence with equal periods: weekly for fast-moving sites, every 28 days for most. The skill ends every report with the next recheck date. A baseline with no recheck date is a screenshot, never a tracker.
Created by Lawrence Hitches
Chief of Staff at StudioHawk. I built this library from the processes our 120-person team runs across 500+ client campaigns. Every skill is free and MIT licensed.