The problem
Wellows ranked in Google but had a small citation footprint inside AI answers: 176 LLM referral sessions a month across ChatGPT, Gemini, Perplexity and the other engines. The gap was not authority in the traditional sense. It was that nothing on the site was structured the way retrieval systems select and quote content.
What I did
- AEO framework. Koray Tuğberk GÜBÜR's semantic logic combined with Google's query fan-out patent, used to model how a question actually expands before an answer is assembled.
- Citation experiments. Grounded versus non-grounded tests, tracking explicit citations and implicit brand mentions separately, because they move independently and mean different things.
- Tool- and community-led distribution. Three free tools built as LLM entry points, plus AEO across Reddit, Quora and G2, with n8n and GA4 wired together for reporting.
The system in one picture
- Distribution channel
- Tool built as an LLM entry point
- Measurement loop
The outcome
LLM sessions rose from 176 a month to roughly nine times that across the eight months this case study covers (GA4). Average position improved from 28.9 to 17.3 and session duration went from 45 seconds to over six minutes. The second number matters more than it looks, because it separates traffic that arrives and leaves from traffic that arrives and reads.
Before and after
| Metric | Before | After |
|---|---|---|
| LLM sessions | 176 / month | ≈9× the baseline |
| Avg. position | 28.9 | 17.3 |
| Session duration | 45 sec | 6+ min |
| Paid spend | n/a | $0 |
Wellows. GA4 for sessions and duration (July 2025 – February 2026); Search Console for position (last 3 months vs the previous 3).
This is stage 4, per-engine measurement, in how I sequence AI search optimization.
Frequently asked questions
What did the AEO work at Wellows involve?
Three parts: an AEO framework combining Koray Tuğberk GÜBÜR's semantic logic with Google's query fan-out patent; grounded versus non-grounded citation experiments; and distribution through three free tools built as LLM entry points plus answers on Reddit, Quora and G2, with n8n and GA4 wired together for reporting.
What changed besides LLM traffic?
Average position improved from 28.9 to 17.3 and session duration went from 45 seconds to over six minutes. Site-wide in Search Console, clicks rose from 2.29K to 7.27K and impressions from 680K to 3.78M, comparing the last three months with the previous three.
Why track citations and brand mentions separately?
Explicit citations and implicit brand mentions move independently and mean different things, so the citation experiments tracked each one on its own instead of blending them into one number.
Was this a client engagement?
No. It was my in-house work at Wellows, with zero paid spend, measured in GA4, Looker Studio and Google Search Console.