Method
The technical SEO audit method: one loop, run until it compounds
Search Console, Analytics, Keyword Planner and Trends, connected into audit systems I designed and direct. The loop is the same whether the site is a Shopify store, a WordPress catalogue or a custom app. What changes is where the fixes land.
- 1
Technical audit
An agentic audit of the whole site: indexation, canonicals, schema, performance, images, AI-search readiness — and the on-page copy checked for factual errors.
- 2
Remediation spec
Prioritised fixes, each with a way to prove it landed. Implemented by your developer, or by AI under my direction.
- 3
Competitor gap map
Five to twenty brands in your niche: what is working for them, what page types they run, and what is missing on your site.
- 4
90-day plan
Twenty to fifty target keywords with real demand behind them, grouped into clusters and committed to a roadmap.
- 5
Execution
New pages, articles, title and meta rewrites, and on-page changes across the ninety days.
- 6
Re-run
The audit runs again against the baseline, and the loop repeats one cluster at a time so authority compounds.
What the loop runs on
Your Search Console and Analytics for ground truth, Keyword Planner and Trends for demand, and the audit, competitor and keyword systems I designed and direct. If you already run a crawler or a rank tracker, I will work from your exports as well. More data is better, and the verification pass is the same either way.
The systems are fast. I am the part that is slow.
I run audits through systems I designed for the job: collection, competitor mapping, keyword workflow, re-audit against a baseline. That is why the work covers more ground than a solo consultant usually can.
It is also why the verification pass matters. Automated tooling is wrong in patterned, predictable ways. A crawler that finished hundreds of URLs in seconds never executed JavaScript, so every missing meta description it found is an artefact. A thousand broken links all returning the same status from one host is bot blocking, not broken links. I keep a register of fourteen categories of this.
So every finding is re-tested against the live site before the report is released, and labelled Confirmed, Likely or Unverified with a step you can run yourself. On one engagement that pass caught five false findings in a finished report. I withdrew the draft hours before release and reissued it corrected. Every case study states how many claims were dropped.
- Re-tested, not re-read
- Every finding is run again against the live site before release.
- Labelled by confidence
- Confirmed, Likely or Unverified, each with a check you can reproduce.
- Withdrawn claims published
- Each case study states how many first-draft findings were corrected or dropped.
- Credit where it is due
- Fixes are credited to whoever made them. I specify; your team or I implement.
Want this loop run on your site?
It starts the same way every time: three to five verified findings, free, and no call to sit through.