Abstract
The AEO Teardown Generator is the engine behind iSimplifyMe's AI-visibility audits. It crawls a prospect's public pages, scores the site against iSM's canonical 100-point Answer Engine Optimization rubric in a single AWS Bedrock call, and renders a branded, self-contained HTML teardown that names the structural gap between what a site publishes and what AI answer engines can cite and recommend.
Problem
Answer engines — ChatGPT, Gemini, Google's AI Overview, Perplexity — increasingly sit between a business and its customers. Whether they cite a site depends less on traditional SEO and more on structure: atomic answers, schema, retrievable content, clean semantics.
Most businesses have no idea where they stand, and the manual audit that surfaces it is slow, judgment-heavy work that does not scale across a pipeline of prospects. The teardown turns a senior analyst's read into a repeatable pipeline.
Approach
Crawl
A plain-HTTP crawler reads the homepage and a bounded set of same-domain content pages — no paid keys, no headless browser. It extracts schema types, meta tags, heading structure, FAQ presence, and per-page text excerpts.
Score
A single Bedrock call grades the site across seven weighted sections — substance, atomic-answer structure, schema, RAG-readiness, semantic clarity, internal linking, and metadata — summing to one hundred points. Each section returns a score, concrete findings, and a prioritized fix list.
Render
The scored result fills iSimplifyMe's Clay report template — brand-locked type and color, fully self-contained, no external dependencies — and writes a single HTML file ready to host or send.
Deep mode
An optional deep pass adds a live visibility scan: it queries the major answer engines with buyer-style questions and reports how often the business is actually named, which engines return zero presence, and which competitors surface instead. This separates "an engine could cite you" (structure) from "an engine does not cite you" (live reality).
Status
- Core crawl → score → render pipeline shipped.
- Deep live-visibility mode shipped behind a flag.
- One-command publish to the intel hub on S3 and CloudFront.
- Automated scores validated against the manual AEO scanner within one to two points per section.
- Built to fail closed — aborts on unreachable domains, flags wrong-site mismatches, and does not fabricate.
Links
- Related → Bot Analytics