166 page-one rankings to 2,516. Ten AI Overviews to a thousand.
A premium laundry detergent with scents composed like fine fragrance. Six months of search and AI work took a domain from near-invisible to the brand answer engines reach for in its category.
Four numbers. Six months.
From an Authority Score of 13 and 166 page-one rankings to a brand the answer engines reach for when someone asks about hard water, sensitive skin, or fragrance longevity.
Up 66.8% quarter on quarter. $63,462 became $105,875, with $169K cumulative across the six-month engagement.
Source · GA4 · Revenue attribution · Sep 2026From 166 at baseline. Top-three positions grew from 54 to 304; positions 4–10 went from 112 to 2,212.
Source · Semrush · Sep 2026 vs Mar 2026Up from roughly 10 at baseline. 247 distinct pages cited across ChatGPT, AI Overviews, AI Mode, and Gemini.
Source · Google Search Console · AI features · Sep 2026Doubled from 13. Referring domains grew from 158 to 291; backlinks from 430 to 1,384.
Source · Semrush · Sep 2026 vs Mar 2026Built for both layers of search.
AI weighs a brand's own content and third-party signals differently depending on what is being asked. Bee Online built for both, in different ratios — because an agency that only does one will win half the queries in a category and never understand why.
- 01Content Layer
200+ blogs, 235K+ words — how-to first
How-to formats prioritised across the content mix, because instructional content is the query space answer engines absorb first. Each piece engineered around a specific question the category was already being asked — hard water, sensitive skin, fragrance longevity, enzyme blends — so the content competed at the exact moment Moziwash needed to be cited.
- 02AI Citation Layer
394 engineered FAQ answers + schema
A dedicated schema-rich FAQ module built site-wide, plus FAQPage JSON-LD on every blog — the format that converts into citations rather than just impressions. 394 answers engineered to match the exact phrasing answer engines extract from. The result: 142 owned People Also Ask boxes, up from three a year ago.
- 03AI Readability Layer
llms.txt, entity wiring & extraction-optimised structure
llms.txt deployed so language models reading the site get a clean map of what the brand is and sells. Entity-graph wiring and extraction-optimised page structure so a model doesn't have to infer — it reads. This layer is what turned impressions inside Google's AI features into 635K generative-AI impressions.
- 04Ranking Velocity Layer
Weekly content refresh against live ranking data
Existing pages re-optimised against live ranking data on a weekly cadence rather than a quarterly calendar. This is what moved 112 keywords in positions 4–10 to 2,212 — not new content, but existing pages pulled into positions that convert. Money pages built specifically for highest commercial-intent queries, so the informational layer had somewhere profitable to point.
- 05Authority Layer
DA 40+ placements & press-mention entity linking
Off-site authority built through DA 40+ placements and press-mention entity linking — the third-party consensus layer AI draws on for evaluative queries. Backlinks went from 430 to 1,384 and referring domains from 158 to 291. This is the layer that made the owned-site rankings stick and the citations credible to answer engines.
- Content programme — 200+ how-to blogs, 235K+ words, engineered for the exact questions the category is already being asked.
- FAQ engineering — 394 structured answers with FAQPage JSON-LD, built to convert into citations rather than impressions.
- AI-readable site layer — llms.txt, entity-graph wiring, and extraction-optimised page structure across the full domain.
- Weekly ranking refresh — live-data re-optimisation of existing pages on a weekly cadence, not a quarterly one.
- Commercial intent pages — collection pages built for highest commercial-intent queries, giving the content layer somewhere profitable to point.
- Off-site authority — DA 40+ placements and press-mention entity linking to build the third-party consensus AI draws on for evaluative queries.
Every layer moved.
Revenue, rankings, citations, and authority — all in the same direction. The six-month before and after, in full.
Two query types. Two different win conditions.
Working both layers is what produced these numbers. An agency that only does one will win half the queries in the category and never understand why.
Ten AI Overview placements to a thousand. In a category where the product barely differentiates in a paragraph, being the cited answer is the differentiation.The Bee Online operating principle · Moziwash engagement
Built for ecommerce brands where citation is the differentiation.
What we built for Moziwash isn't a content retainer. It's a dual-layer system — owned content engineered to be extracted, and external authority engineered to be cited — built specifically for categories where every product looks the same in a paragraph of copy.
If your brand competes in a category where the recommendation is the product — where being cited in an AI answer is a more durable moat than a ranking — the same operating model applies.
A simplified, prospect-safe view of a performance engagement. Specific tactics, tools, vendors, and unit-economic detail are withheld by design.
Shared with permission. Not for redistribution. Specific strategies, vendor stack, and unit-economic detail available under NDA.