Brand Intelligence Toolkit
A structured, evidence-led audit of homepage clarity, positioning, differentiation, trust, and conversion — built for strategic review, not generic AI commentary.
Live: https://brand-analyst.vercel.app/ (redeploy this version to the same project, or update the URL once redeployed)
What changed in this rebuild
- Runs on the Anthropic API (Claude) instead of OpenAI. Same audit logic, different model provider.
- Branding now matches the rest of the tool set — same design tokens (paper/ink/persimmon, Neue Montreal/Switzer/IBM Plex Mono) as the Category Thesis Generator and Technical-to-Buyer Translator, so all three read as one built system rather than three one-off experiments.
- Same three-stage flow as the original: Extract → Evaluate → Prioritize.
What it does
- Extract — fetches the live homepage server-side and pulls the visible text (headings, copy, CTAs), stripped of scripts/styles/markup.
- Evaluate — scores eight fixed strategic categories (clarity, positioning, differentiation, trust, conversion, audience specificity, proof strength, voice consistency), each 1–10 with a rationale grounded in the actual page content.
- Prioritize — returns a ranked action plan of the 3–5 highest-impact fixes.
Stack
- Next.js 14 (App Router)
- Anthropic API (
claude-sonnet-5), called server-side from a route handler — the API key never reaches the browser
- No database — stateless, one request in, one JSON response out
- Page fetching + text extraction happens server-side via a lightweight regex-based HTML strip (no headless browser, so JavaScript-rendered pages with little server-rendered content will extract thin text — the tool says so honestly rather than guessing)
Running locally
npm install
cp .env.example .env.local
# edit .env.local and add your real ANTHROPIC_API_KEY
npm run dev
Visit http://localhost:3000.
Deploying to Vercel
- Push this repo to GitHub (replacing the existing
brand-analyst repo contents, or as a new repo if you'd rather keep history).
- In Vercel, import the repo (or point the existing
brand-analyst project at the updated repo).
- Under Environment Variables, add:
- Remove any old
OPENAI_API_KEY variable if one exists — it's no longer used.
- Deploy.
Notes
- The API key is a server-side env var (not
NEXT_PUBLIC_) — only read inside app/api/audit/route.js, never shipped to the client.
- The model is prompted to say so honestly rather than fabricate an evaluation when a page returns too little extractable text (common on heavily JavaScript-rendered sites).
Built by Ashley Pola — brand and narrative strategist.