AEO Analytics Platform
Self-initiatedAn answer-engine optimization platform that measures brand visibility across AI search engines and generates prioritized, evidence-backed fixes.
Working MVP shipped and demoed to the Director of Data and Insights and the Textron AI Council. Handed off the MVP and code for refinement and deployment. Built as the in-house alternative to a ~$100K/month (~$1.2M/year) commercial tool - cost avoidance, not delivered savings.
How It Works
Key Product Decisions
Multi-engine independence
Queries OpenAI and Gemini as separate engines. Results never blended. Each engine's metrics stand alone so engine-specific blind spots stay visible.
Evidence verification
Every brand-mention verdict includes a quoted evidence span. That span is mechanically verified as a real substring of the AI engine's response. If it can't be confirmed, the verdict is discarded. No model involved in this check.
Deterministic site-health scoring
SSR detection, JSON-LD validation, bot-access auditing, content depth, entity coverage. Score is 0-100, fully deterministic. No LLM touches it.
Priorities hard-capped at 4
Recommendations are capped at 4 per run, deduped against completed work, tagged with per-engine directional impact estimates. Never claims causation.
Immutable audit trail
Raw AI engine responses stored before any parsing. Every metric traces to a stored row. Old data never mutated.
Per-engine failure isolation
One engine failing never breaks a run. Failed calls are excluded from the denominator, not counted as zero.
The Story
During my product management internship at Textron (Fortune 500, specialized vehicles), I discovered that the marketing team was struggling with AEO - answer-engine optimization, the AI-search equivalent of SEO. They were evaluating a commercial tool called Profound, quoted at roughly $100K/month.
Nobody told me to build this. I talked to teams across the organization out of curiosity, found this gap, and took ownership. I scoped the problem, designed the architecture, made every product and trust decision, and shipped a working MVP via AI-directed implementation.
I demoed the platform - walking through the logic, the software architecture, and the workflow - to the Director of Data and Insights and to the Textron AI Council. I then handed over the working MVP and its code for the council to take forward.
Video walkthrough coming soon.