KHM·09
CASE STUDY

AI Engine Optimisation

CTO · originator · 2024 — Present

  • ai

Making the catalog discoverable inside AI engines, not just search results.

Problem

As discovery shifts from web search to AI answers, classic SEO no longer determines what generative engines retrieve and cite. The catalog needed to be visible to ChatGPT, Gemini, and Claude.

Constraints

No control over third-party model internals; the risk of AI hallucinating prices or availability; and the need to keep structured data accurate as the live catalog changes.

Architecture

A deliberate structured-data and API layer — schema markup, machine-readable endpoints, and grounding context — configured so AI engines can index and answer about products accurately, paired with LLM-backed conversational interfaces.

Trade-offs

Per-call LLM cost and latency, and attribution that's harder to measure than classic SEO, in exchange for presence in a fast-growing acquisition channel competitors are ignoring.

Decisions

Built on hosted LLM APIs rather than self-trained models to ship quality fast without an ML-ops burden; grounded every answer in real catalog data to keep hallucination in check.

Results

Products surfacing inside AI engine results — an acquisition channel established before it became obvious — plus conversational interfaces that answer customers in natural language.

Lessons

Being early to a distribution channel is its own moat. The hard part of AEO is not the schema; it's keeping the structured truth in sync with a live system.

Stack
  • Structured Data / Schema
  • LLM APIs
  • REST APIs
  • Next.js
ChatGPT · Gemini · ClaudeAI engines
PioneeredDiscipline