Transparent case study
Seobase SEO recovery: from a GSC baseline to controlled tests
Seobase had pages earning impressions but appearing deep in search, while the product it sells—AI-assisted SEO content automation—was not yet earning enough qualified organic demand. This is the operating record of the recovery work, not a claim that the work has already produced a ranking win.
Observed window: August 15–September 11, 2026 · Last updated September 13, 2026
The diagnosis
The first problem was visibility depth, not a shortage of URLs. The complete Search Console window averaged position 73.18. The backlink guide alone earned 1,891 impressions at position 72.17, and the title checker earned 542 impressions at position 62.63. Those pages had discovery signals, but their answers, headings, workflows, and internal next steps were not yet strong enough to turn visibility into clicks.
The measurement system also had to distinguish a provider outage from a genuine zero. The weekly snapshot now keeps raw GSC rows, conversion attribution, URL audits, and Semrush report health separate, so a quota error cannot become a false keyword opportunity or ranking result.
What changed
| Canonical page | Impressions | Position | Controlled test |
|---|---|---|---|
| /blog/find-backlinks-in-google | 1,891 | 72.17 | Clarified the Search Console workflow, limits, evidence, and next-step tool. |
| /tools/headline-checker | 542 | 62.63 | Aligned the existing title checker with the queries already earning impressions. |
| /blog/moz-vs-ahrefs | 357 | 79.36 | Added direct comparison workflows and the adjacent SE Ranking vs Moz intent. |
| /blog/shopify-seo-experts | 277 | 86.44 | Added a hiring checklist, practical service detail, and an automation decision path. |
| /blog/how-to-find-low-hanging-fruit-keywords | 198 | 86.08 | Connected Search Console evidence to a repeatable refresh and prioritization workflow. |
Update, September 18, 2026: four of the five pages above (find-backlinks-in-google, moz-vs-ahrefs, shopify-seo-experts and how-to-find-low-hanging-fruit-keywords) were removed from the site. The controlled tests recorded here did run, and the impressions and positions are what Search Console reported for the observed window — but the full history then showed those impressions never converted: 12,385 impressions produced 8 clicks in 96 days, at an average position in the seventies. The pages were also topically unrelated to the product, which is the pattern the August 2026 spam update penalises. Optimising them further would have improved a metric with no commercial value, so they were pruned rather than iterated on. This table stands as the record of what was tried.
The low-KD tool strategy
We reused canonical pages instead of multiplying near-duplicates. The strongest product-fit opportunities were an international SEO audit at 1,300 searches and KD 1, a search engine listing preview at 260 searches and KD 2, an hreflang generator at 1,900 searches and KD 8, and an SEO checklist PDF at 170 searches and KD 4. These are prioritization estimates, not traffic guarantees, so each tool states exactly what it checks and what it cannot prove.
What remains unproven
- Google has not yet recrawled or indexed several revised tool and comparison URLs.
- Semrush organic and backlink reports are currently quota-limited; those failures are recorded as unavailable.
- No ranking win over Outrank, traffic lift, or qualified-trial lift is claimed from this window.
- The next decision gate is a later equal complete GSC window with clicks, query/page rows, and site-connection attribution.
Use the same operating loop
Start with the page already earning impressions, verify the query and intent, make one controlled change, validate the live URL and links, then compare the next complete window. The point of AI SEO automation is not to publish more pages blindly; it is to reduce the time from evidence to a reviewed, measurable improvement.