Signal

Inspiration

Every creator and small brand asks the same two questions on repeat: "Am I even showing up when people search for this?" and "What do I post today?" Most tools answer one or the other — an SEO checker, or a caption generator. We wanted one tool that connects the two: tell a brand where it's weak, then immediately generate the content to fix it.

What it does

Signal is a three-part content engine:

• Visibility check — paste in a website and it fetches the real live title tag, meta description, and headings off that page, then grounds an analysis of your competitors, your current search/AI-visibility standing, and concrete SEO + GEO (Generative Engine Optimization) actions in that real data — not generic advice.
• Content generator — pulls a real, current trending headline from Google News on your topic, then generates a platform-native post, caption/script, hashtags, and hook options tuned to your brand's voice. It also outputs a reusable prompt template, so the brand owner can keep generating similar posts themselves later.
• Refine mode — paste an existing draft and it rewrites it, explaining exactly what it changed and why (e.g. cutting generic openers like "In today's fast-paced world").
• Creator marketplace — every plan can be matched with a real creator (editor, ghostwriter, SEO writer) who takes the generated draft to finished, publish-ready work.

How we built it

Built with Next.js and TypeScript, deployed on Vercel. Content generation runs on Groq (openai/gpt-oss-120b) for fast inference. The visibility check does a live server-side fetch of the target page's HTML and parses the title/meta/H1s directly — no third-party SEO API needed. Trending angles are pulled live from Google News RSS. Styling and motion with Tailwind and Framer Motion.

Challenges we ran into

Getting structured JSON output reliably out of the model was the trickiest part — occasionally a generation would nest a stray key inside a string value and break parsing. We added validation with an automatic one-shot retry (with an explicit correction prompt) rather than letting a single bad generation crash the request.

Accomplishments that we're proud of

Grounding the visibility check in a real, live fetch of the actual page rather than pure LLM guesswork — so the SEO recommendations reference the brand's actual title tag and meta description, not boilerplate advice.

What we learned

How much of "AI visibility" work is really about giving the model verifiable ground truth to reason from, rather than trusting it to know the current state of a page from memory.

What's next for Signal

Real search-ranking data via a search API, structured-data (JSON-LD) generation for GEO, and scheduling/publishing integrations so generated content can go straight out to the platform.

Built With

• framer-motion
• google-new
• groq
• nextjs
• react
• tailwindcss
• typescript
• vercel