Live Case Study

Aggressive Autopilot vs Manual GEO Optimization — RiteKit's 30-Day Story

We turned MentionFox Aggressive Autopilot loose on ritekit.com on 2026-05-09. No human in the loop, no hand-tuned content. Goal: see how 30 days of auto-driven AI visibility work moves the needle vs the day-0 baseline.

measurement in progress

Day 0 · Baseline

Where RiteKit started

Per-LLM win rate, top cited competitors, and sample winning + losing conversations.

Read the baseline →

Day 7

First-week movement

[Day 7 TBD — measurement in progress]

Day 14

Mid-window read

[Day 14 TBD — measurement in progress]

Day 21

Compounding effects

[Day 21 TBD — measurement in progress]

Day 30 · Final

30-day final story

[Day 30 TBD — measurement in progress]

Methodology

Every snapshot uses the same fingerprint so the deltas are comparable.

Subject: ritekit.com — social-media + hashtag tooling, founder @sashagrey-style category
LLM allocations: gemini_flash 30% · gpt4omini 30% · deepseek 25% · mistral 15% · claude 0% (deactivated for the test)
Prompts: 15 monitoring queries spanning the RiteKit category (hashtag generators, scheduling, Twitter optimization, influencer tools)
Daily polling: ~50-call light scan, snapshot recorded in case_study_runs
Milestone reports: deeper ~120-call scan at days 7 / 14 / 21 / 30, with full per-LLM and competitor breakdown
Win definition: RiteKit cited (any case-insensitive substring) in the LLM's response

What "Aggressive Autopilot" actually does

It's the same engine MentionFox runs for itself, just unsupervised. Aggressive mode pushes the LLM allocation toward the cheaper, faster models (gemini_flash + gpt4omini do the bulk of conversations), keeps Claude switched off to control cost, and lets the question bank cycle without manual approval. Every winning conversation feeds the next prompt's context. Every losing one feeds the next round of GEO content. Thirty days, no human babysitter — that's the test.