Search engines have twenty years of tooling. AI answer engines have almost none. İzMercek is the operations layer for the second case: measure how legible a brand is to AI systems, compile its canonical knowledge into the formats those systems actually read, and publish the result to the brand's own domain.
The loop
- Measure — a deterministic, rule-based visibility score. No model in the scoring path, so the same site always produces the same number.
- Compile — a canonical knowledge model becomes
llms.txt,llms-full.txt, schema.org JSON-LD and an AI manifest. Versioned, with a diff between versions. - Publish — the artifacts are pushed to the brand's own hosting, so they live at the brand's address rather than on someone else's platform.
- Verify — check what AI systems actually say about the brand, and compare it against what the brand published.
Why it exists
A tool that only prints a score is a tool you use once. The useful part is the work after the score: filling in the knowledge, generating the files, getting them onto the site, and checking later whether anything drifted. That whole loop is the product.
How the studio uses it
This site is İzMercek's own output. The page you are reading, the llms.txt, the structured data and the sitemap were produced and measured with it. When it was first pointed at peepsicklabs.com the score was 76/100; the rebuild that followed took it to 100/100 — and the run surfaced several real bugs in İzMercek itself, which is rather the point of using your own tools.
Status
In development. Not yet publicly available — there is no sign-up and no hosted instance to visit. It is listed here because it is real, it is running, and because the studio publishes what it is working on rather than only what is finished.