AI Marketing Strategy Agent
Full-stack SEO audit and paid media strategist that crawls a site, scores performance, and delivers unified organic + paid recommendations
Built an AI-powered marketing consultant that audits a website's SEO health and automatically recommends paid media strategies, keyword opportunities,...
Problem: Many small businesses and startups know they need better digital marketing but lack the expertise or budget for agency audits. SEO tools often surface technical issues in isolation. Advertising platforms optimize campaign settings. Almost none connect both into a unified acquisition strategy. I wanted to build a system that analyzes a website holistically and generates actionable recommendations across organic search and paid media.
Solution: I developed a full-stack marketing agent that crawls a site (up to 50 pages, respecting robots.txt), runs on-page SEO analysis, evaluates Core Web Vitals with Lighthouse on the homepage and top-linked pages, and synthesizes findings into prioritized SEO fixes plus a paid media strategy — keyword themes, channel mix, budget guidance, and landing page opportunities. Users can run the full technical audit without an API key; optional OpenAI integration upgrades recommendations to AI-written, business-specific guidance.
SEO Audit: Meta titles and descriptions, heading hierarchy, canonical tags, Open Graph and Twitter cards, image alt text, structured data, internal linking patterns, HTTP status errors, robots.txt and sitemap checks, and Lighthouse performance/SEO/accessibility scores — all surfaced in a filterable, tabbed dashboard with JSON, Markdown, and CSV export.
Paid Media Strategy: The agent infers business type from site content and proposes keyword targets with intent labels (commercial, transactional, informational), match types, and landing page mappings. It recommends channels (Google Search, Meta, LinkedIn, retargeting, and more) with priority tiers, budget ranges, and tactical next steps — generated by OpenAI when a key is provided, or by a structured rule-based fallback.
AI Recommendations: Rather than simply reporting issues, the application prioritizes recommendations by likely business impact, labels quick wins vs. long-term fixes, and explains why each change matters. Structured prompts feed crawl findings and Lighthouse summaries into GPT-4o-mini for SEO fix instructions; a separate prompt pipeline produces the paid strategy brief.
Tech Stack: Next.js 16 (App Router), TypeScript, React 19, and Tailwind CSS v4 for the UI. Playwright for headless crawling, Cheerio for on-page HTML analysis, Lighthouse + Chrome for Core Web Vitals, OpenAI API for optional AI layers, and Docker on Railway for production deployment (browser automation requires a containerized runtime, not serverless).
My Role: End-to-end ownership: product ideation, UX design (including day/night theme and tabbed results), AI workflow and prompt engineering, crawler and SEO analysis pipeline, Lighthouse integration, paid media strategy framework, background job orchestration with disk persistence, Docker/Railway deployment, and testing.
What I Learned: Building this reinforced how AI can move beyond content generation into strategic decision support. Instead of replacing marketers, AI can rapidly synthesize website performance, search optimization, and advertising best practices into practical recommendations that help businesses make better marketing decisions — especially when grounded in real crawl data rather than generic templates.
Optional OpenAI API key in the UI unlocks AI-written SEO fixes and paid media strategy. The technical crawl, on-page analysis, Lighthouse scores, and rule-based recommendations work without any API key.