Project - Empire Landscaping — AI Blog CMS

Overview

After launching Empire Landscaping's marketing site, I built a second layer on top of it: a private admin CMS with a built-in AI writing assistant that helps the team publish blog content that actually ranks. It's a real example of the SEO service I offer — baked into the product, not bolted on after the fact.

The CMS lives at a protected /admin route on their live Next.js site. From there, the team can create a new post, edit existing ones in a Tiptap rich-text editor, preview before publishing, and run everything through an AI Writing Assistant that knows Empire's voice, Empire's services, and what Google actually rewards in 2026.

What it does

Rich Tiptap editor with AI-streaming edits

The editor is a full Tiptap setup — headings, lists, links, tables, inline formatting, images — with a custom extension that lets the AI assistant highlight and rewrite specific passages in place. Edits stream in token-by-token via server-sent events, so the team sees the assistant working in real time instead of waiting on a big blocking request.

AI Writing Assistant (Gemini 3 Flash via OpenRouter)

A side-panel assistant uses tool calling to take real actions on the post — not just chat. It can:

• Suggest or rewrite titles, meta descriptions, intros, and body sections.
• Insert schema (BlogPosting, HowTo, FAQPage) when the content structure actually warrants it.
• Update the featured image, alt text, and OG metadata.
• Generate FAQs for genuinely Q&A-structured posts.
• Score the current post against the 5-category rubric and return a numeric breakdown plus a band label.

The assistant is grounded in a markdown knowledge base living inside the repo — SEO principles distilled from Google's Sept 2025 Search Quality guidelines, local-SEO rules specific to the Okanagan, and a 6-pillar blog-writing framework. Swapping models is a one-line change in the route thanks to the OpenRouter integration.

5-category / 100-point SEO scoring rubric

Every post can be scored on demand. The rubric covers:

• Content Quality (30) — depth, originality, readability, first-hand experience signals.
• SEO Optimization (25) — title, meta description, keywords, internal linking, schema.
• E-E-A-T Signals (20) — experience, expertise, authoritativeness, trust.
• Technical Elements (15) — heading hierarchy, paragraph length, image alt text, OG tags.
• AI Citation Readiness (10) — answer-first formatting, passage-level quotability for AI Overviews, ChatGPT, and Perplexity.

The result comes back as a total score, a band (Weak / Okay / Strong / Excellent), and a per-category breakdown with specific fixes — so the team doesn't just get a grade, they get a to-do list.

Cookie-gated admin, protected API

The whole admin surface is gated behind a simple admin_auth cookie check on both the pages and the /api/admin/* routes, so the Writing Assistant and its tools can't be called from outside. Safe to ship on the live site, zero risk of random visitors driving the AI.

Why this matters

This project is a real-world example of the SEO service I offer clients — not theoretical advice, not a PDF audit, but SEO enforced inside the tool the content team actually uses. Empire can publish landscaping articles that are consistently well-structured, locally relevant, and optimized for both Google and AI answer engines, without needing an SEO specialist in the room for every post.

It's the kind of tech solution I build when "just hire a writer" or "buy a SaaS" isn't enough — a custom internal tool that turns a one-time service into an ongoing advantage.

Tech stack

Site link