I'm Nik and I've worked 8 years working in Marketing at High Growth Startups. I’ve spent years fixing landing pages for my campaigns, and I realized most bad conversion rates come from structural errors, not just bad copy. (Still copy remains the major part of conversions)
I built Landkit Audit to act as a Linter for conversion.
How it works:
Ingestion: It fetches your site and converts the DOM into a high-fidelity semantic map. This strips away the design noise to reveal the actual information hierarchy your user is navigating.
The Linting Rules: It runs this structure against 50+ heuristics derived from direct response frameworks (Cialdini, Fogg, LIFT).
Heuristic Analysis: It flags specific 'violations' like:
Visual Hierarchy: H1s that fail to pass the 5-second test (inferred).
Friction: Forms that ask for high-commitment data before establishing value.
Anxiety: Missing trust signals near payment triggers.
The Stack:
Frontend: React + Framer Motion (for the Terminal visualization).
Backend: Supabase Edge Functions.
Analysis: Gemini Pro 2.5 (great at logical reasoning/inference).
The hardest part was tuning the prompt to be critical rather than nice. Most LLMs default to being supportive; I had to prompt engineer a persona that acts more like a brutal senior editor.
Try it here: https://landkit.pro/audit
It’s free to run the audit. I’d love to know if the Linting errors it finds on your projects feel accurate.
nikhonit•1h ago
Since we are spinning up a headless browser to map the DOM before feeding it to the inference engine, it's not instantaneous like a regex check