Rebuilding idea-lab.ai From an F to a B — Live, in Public

Baseline: 18.2 / 100

When we ran the first scan against idea-lab.ai, the AI Visibility Score came back at 18.2 — squarely in the F band. The site had a static index.html head, no llms.txt, weak Schema.org coverage, and zero authored content that an LLM could attribute to a person.

We could have quietly fixed it. Instead, we decided to document the rebuild in public — every commit, every score delta, every dead end. This is the Experiment.

The first three changes

1. Authored a real Author entity. A Person node in JSON-LD linked to LinkedIn and the About page.
2. Shipped llms.txt and llms-full.txt. A 40-line manifest and a 12k-token concatenated knowledge base.
3. Replaced a sitewide canonical with per-route canonicals using react-helmet-async.

After those three changes alone, the score moved from 18.2 to 27.6. The remaining 60+ points are the harder work — and we'll document each one here as it ships.