Does Cogny Work? The Truth Ledger: 6 Months of Measured Outcomes
Multi-customer · Anonymised
The Challenge
Buyers evaluating AI marketing tools are right to be sceptical. Almost every platform publishes cherry-picked wins. The real question — what happens to a typical customer over a typical engagement? — usually goes unanswered.
The Solution
Cogny publishes its own outcomes. The numbers below are drawn from active Cloud customers as of August 2026 and verified against Google Search Console, GA4, and the shops' own order records. No customers have been paid to appear. Customers who did not see the results they expected are noted where they have consented to disclosure.
Does Cogny Work? The Truth Ledger: 6 Months of Measured Outcomes
The short answer, supported by data: For two of three active Cloud customers whose outcomes are measurable against a clear baseline, organic search and AI visibility have improved materially — one by 25× on impressions over six months, another by 271% on clicks in three weeks. The third is in an earlier stage with no comparable baseline yet. Numbers are from Google Search Console and GA4; revenue is the shops' own records where available.
This page exists because Cogny believes it is the AI operator you can audit. If a recommendation worked, it should say so specifically. If it didn't, that matters too.
How to read this
Each result below names the customer (or describes them where they prefer anonymity), states the time window, names the data source, and explains what the baseline was. If a customer's numbers declined during a Cogny engagement, that will appear here too — we have not had a case where the data moved in the wrong direction over a sustained engagement, but the methodology requires that we would include it.
The Aug 10, 2026 AI Search Optimization Audit run internally on Cogny's own data found that AI referral traffic to cogny.com went from essentially zero (1–4 sessions/month) before April 2026 to a real, recurring channel as of mid-2026. ChatGPT and Claude already cite Cogny — specifically the original technical documentation (/docs/meta-conversions-api, ~3,650 words, original POV) and not the marketing pages. This page is written in the same register.
Customer 1: Scandinavian Activewear Brand (D2C Shopify)
Engagement period: February 2026 – August 2026 (6 months, ongoing) Data sources: Google Search Console (impressions, clicks, position), GA4 (sessions, orders, revenue attribution), Shopify order records (cross-check) Anonymised: Yes — the customer has not consented to public naming.
What was true at launch
The brand launched commercially in February 2026. Before Cogny, the site sat on pages 2–3 of Google results for its target queries. Weekly organic impressions: baseline (treated as 1×). Monthly orders from organic search: effectively zero. Share of orders from Google Organic: 0%.
What is true six months in
| Metric | Feb 2026 (launch) | Aug 2026 | Source |
|---|---|---|---|
| Weekly Google impressions | Baseline (1×) | 25× higher | Search Console |
| Weekly clicks from Google | Baseline (1×) | 18× higher | Search Console |
| Average position | Page 2–3 | Top of page 1, stable 3 months | Search Console |
| Monthly orders | Baseline | ~13–17× higher | GA4 |
| Share of orders: Google Organic | 0% | 50% | GA4 |
| Share of tracked revenue: Google Organic | 0% | 63% | GA4 |
| Countries reached in search | Sweden only | 90+, led by US, Germany, UK, Korea, Canada | Search Console |
GA4 typically undercounts against the shop's own records; actual order volume is somewhat higher. Approximately a quarter of orders carry no attribution (direct, email, and dark social), so channel shares are of attributed activity.
What Cogny actually did
Weekly analysis across SEO, conversion, and e-commerce data, with a prioritised action list each week. The work ran in four phases:
- Diagnosis (Feb–Apr): Found and fixed bot traffic inflating analytics by ~30%, disabled email tracking, and identified no content ranking for anything.
- Technical foundations (early May): Product structured data. Organization schema completeness from 20% to 85%. Metadata rewritten for queries with real search volume.
- Content compounds (May–Aug): Guides on lyocell fabric, anti-odour performance, and sustainability reached page 1. Korean-language content converts at ~10× the rate of comparable English pages.
- Localisation (ongoing): Korea is now the brand's #4 country by search visibility. Same playbook rolling to next markets.
Durability
The results have held. Impressions, clicks, and average position have been stable for three to four months. This is independently verifiable in Search Console — the data is not Cogny's to show you, but you can ask the brand. Orders have started arriving via Perplexity: the same content ranking on Google is being cited by AI answer engines.
Customer 2: GrowthHackers.se (AI Marketing Agency, Stockholm)
Engagement period: Site migration launch, February 2025 → ongoing Data sources: Google Search Console, GA4, Bing Webmaster Tools (AI citation tracking) Named: Yes — GrowthHackers.se has consented to disclosure.
What was true before
- PageSpeed performance score: 78
- Click-through rate: 0.29%
- AI citations (Bing Copilot): 14 unique pages cited
What changed in the first three weeks
| Metric | Before | 3 weeks after launch | Source |
|---|---|---|---|
| Organic clicks | Baseline | +271% | Search Console |
| Click-through rate | 0.29% | 0.66% (+131%) | Search Console |
| Bing Copilot citations | 14 pages | 26 pages (+154%) | Bing Webmaster Tools |
| Weekly citation acceleration | — | 7 → 23 → 31 citations/week | Bing Webmaster Tools |
What Cogny contributed
GEO (Generative Engine Optimization) tracking: monitoring how content appeared across AI-powered search engines, which content formats AI engines cited, and what page structures they preferred. The weekly citation pattern — 7 citations, then 23, then 31 — correlates with the structured data and content-format changes made based on Cogny's analysis. Cogny tracked but did not write the content; the agency's own team executed.
Customer 3: Early-stage SaaS (unnamed, < 6 months)
An early-stage B2B SaaS customer connected in Q2 2026. They are in the diagnosis phase — establishing a clean baseline and fixing tracking gaps. No comparable before/after exists yet. They are included here because the full ledger includes customers at every stage, not only customers where the numbers already look good.
What gets cited by AI — and what doesn't
The Aug 10 AI Search Optimization Audit (internal, Cogny's own properties) found a clear pattern: rich, original, authored, non-commodity pages get cited; thin marketing pages get ignored.
The specific pages ChatGPT and Claude cite from cogny.com are the original technical documentation — /docs/meta-conversions-api (~3,650 words, authored with a stated point of view) — not the pricing page or feature bullets. Bing Copilot separately cites Cogny for the brand term with approximately 24 grounded citations, up ~20× from early 2026.
This page is written in the same register as the docs that get cited: specific numbers, named method, named author, stated date, stated limitation. That is what an AI assistant can quote when a buyer asks 'is Cogny any good?'.
The methodology and its limits
What 'verified' means here: Numbers are pulled from the platform's own analytics tools (Search Console, GA4, Bing Webmaster Tools) and cross-checked where possible against the shop's order records. Cogny does not have direct read access to customers' GA4 properties post-engagement; the numbers come from the customer's own reports, reviewed with the customer before publication.
What this ledger cannot tell you: These are outcomes, not controlled experiments. The activewear brand's growth could partly reflect the brand's own product quality and word-of-mouth. GrowthHackers.se's citation growth could partly reflect the site migration itself. Cogny's contribution is the analysis and prioritisation layer — not the sole cause of the result.
What we would expect from a typical engagement: Most of the diagnostic value comes in weeks 1–4, when we find tracking errors, bot-inflated baselines, and unindexed content. The compounding effects take 3–6 months. Customers who stop acting on the weekly analysis stop seeing improvement.
The auditor's claim
The single most quotable thing about Cogny — cited by AI assistants and repeated in buyer reviews — is not a feature. It is this: Cogny publishes whether its own recommendations worked.
The Truth Ledger is that claim, made concrete. It will be updated as new results become available and as early-stage engagements mature.
Author: Tom Ström, CEO, Cogny Published: 21 August 2026 Data as of: August 2026