Growth exposed
a harder problem.

The growth was measured on matched terms.

Monthly non-brand organic clicks grew from 579 to 1,283 and impressions from 29,394 to 147,155. Average position improved from 17.7 to 9.4. The comparison uses May–July 2025 against May–July 2026, excludes brand queries throughout and comes from Google Search Console rather than a third-party traffic estimate.

Across the full 16-month Search Console window, the US blog generated 12,612 clicks and 1,482,003 impressions. Its share of all US non-brand organic clicks increased from roughly 2% to 12.7%. The export is page-filtered to /blog/ and query-filtered to exclude roadsurfer brand terms.

Top-three rankings increased from 22 to 441 while total ranked keywords rose only 13%, from 1,209 to 1,372. The distribution—not just the total—shows that the gain came primarily from re-optimising existing pages rather than publishing more of them.

THE FINDING

Rankings improved while click-through fell.

Average position improved 8.4 places while click-through rate fell 56%. Those measures should not normally move in opposite directions. Five top-ten pages held 447,105 impressions—39% of blog impressions—at less than 0.5% CTR. The clearest case averaged position 7.8, received 238,842 impressions and recorded 0.08% CTR.

The aggregate CTR decline is not clean proof on its own. Fivefold impression growth changed the query mix and introduced more low-CTR discovery queries. The page-level pattern—strong positions, substantial impressions and extremely low CTR—is the more useful signal.

THE RESPONSE

Build for citation, not only ranking.

I developed an answer-engine optimisation programme around extractable answers, clear entity signals, evidence, page structure and citation monitoring. By August 2026, the programme recorded 947 citations: 299 in ChatGPT, 181 in Google AI Overviews, 145 in Gemini, 116 in AI Mode, 83 in Grok, 77 in Perplexity and 46 in Copilot.

The accurate framing is “from no measurable presence,” not “from zero.” Citation tracking matured during the measurement window, so the earlier state was under-observed rather than provably empty.

The governance behind this work is described on the approach page. The proprietary AEO methodology itself is intentionally not published.

THE OPERATING SYSTEM

Sixteen workflows, with people at the gates.

I built 16 AI-assisted workflows across a four-stage pipeline for six market languages. The system covered research, keyword strategy, briefs, drafting, refreshes, QA, CMS formatting, localisation screening and website checks.

Company documentation estimated roughly seven hours saved per article cycle across research, refresh and publication QA. The final decisions remained human: topic selection, QA verdict and publishing. This is the same autonomy model described on the approach page.

GA4 provides a second, differently scoped signal. Organic Search sessions landing on US blog pages increased from 8,392 in calendar 2024 to 24,732 from September 2025 through August 2026—roughly threefold. The windows are not like-for-like, so this supports scale rather than a precise percentage uplift.

WHAT I WOULD DO DIFFERENTLY

Instrument earlier. Escalate sooner.

• Start citation tracking before optimisation, with stable prompts and reporting definitions.
• Advocate earlier for specialist tooling rather than relying on partial manual observation.
• Separate query-mix effects from page-level CTR anomalies in the reporting model from day one.