mrcheckpoint

The Problem

DUI checkpoint notices live in local news posts, police feeds, and PDFs: scattered, irregular, and useless if they arrive late. The 2026 rebuild had to do five jobs at once: ingest and verify locations in real time, publish a bilingual nationwide directory, alert people when it matters, monetize the traffic two different ways without breaking trust, and move off WordPress without losing years of search rankings.

What I Built

• Bilingual nationwide checkpoint directory (English/Spanish): state, county, city, and individual checkpoint pages across the 39 states plus D.C. where checkpoints are legally publishable, with hreflang tags so Google indexes both languages correctly.
• Near Me geolocation, a public tip submission form, and footage/photo upload for tips, stored in private cloud storage.
• Attorney directory as a full paid product: three tiers (Bronze/Silver/Gold), Stripe Checkout billing, and a self-serve onboarding flow where an attorney pays, gets a magic-link resume page, fills out their profile, and e-signs a generated advertising agreement.
• AI website scanner for attorney onboarding: give it a law firm's URL and it crawls up to 8 pages and uses Gemini to auto-fill their bio, practice areas, and contact info instead of making them type it all in.
• Draft/preview system so a new attorney listing stays hidden from the public site until I approve it, with a private preview link so they can see their profile before it goes live.
• Geo-targeted and featured placement (Gold tier can appear directly on individual checkpoint pages), monthly deliverables tracking per sponsor, and tokenized performance report links attorneys can open without logging in.
• Billing lifecycle handling end to end: Stripe webhooks pause a listing on a failed payment and reactivate it automatically on recovery.
• Telegram alert network with one-tap linking, paid Basic/Premium subscriber tiers through Stripe, and Substack-paid subscribers automatically granted the same alert access through a cross-platform sync.
• AI scraping pipeline monitoring 35+ sources: HTML pages, RSS feeds, Playwright for JS-rendered pages, PDF parsing for police departments that only publish notices as PDFs, and authenticated Facebook page monitoring.
• Gemini extracts structured checkpoint data (date, time, location, confidence score) from that source text. High-confidence entries auto-publish, everything else lands in a human review queue in Garden.
• Scheduled jobs handle the boring parts on their own: checkpoints auto-flip from active to completed the day after they pass, and every new publish pings Bing/IndexNow and refreshes the sitemap so it gets discovered fast.
• Full SEO and AI-discovery layer: a ~700-URL dynamic sitemap, a robots.txt that explicitly allows GPTBot, ClaudeBot, PerplexityBot, and Google-Extended by name, Schema.org structured data on every page type, and a WordPress-to-Cloud-Run migration with a ~980-URL redirect map so the legacy rankings and inbound links survived the cutover intact, including keeping the money page at /dui-checkpoints/.
• Garden, the internal ops dashboard: checkpoint review queue, attorney management, a deliverables board, subscriber management, tip/footage review, scraper controls, analytics, and Telegram tools, behind Google OAuth with an email allowlist, IP lockout after repeated failed logins, and Cloudflare Turnstile.

How It Works

Scrapers and tip intake land in a geographic queue. Gemini turns messy local source text into structured checkpoint records; high-confidence items can auto-publish while the rest go to human review in Garden. Verified entries power the public directory, fan out through regional Telegram alerts, and feed geo-routed attorney placements. Attorneys onboard themselves through Stripe Checkout and a magic-link profile flow, with an AI scanner pre-filling their bio from their own website. On cutover, the redirect map sent legacy WordPress URLs into the new product; after every publish, sitemap refresh and IndexNow push the change to search engines fast.

Results

2026 rebuild is live on Cloud Run: bilingual nationwide directory across 39 states, a two-sided monetization model (consumer alerts plus a self-serve attorney ad product with real onboarding and billing automation), an AI ingest, review, and publish pipeline running on 35+ sources, and a completed WordPress cutover that preserved ~980 redirects and the site's SEO history. This is the same platform I've kept running for 13 years, rebuilt from the ground up by me alone this year.

At a Glance

• Continuous delivery13+ years
• Audience reach600K+
• States covered39 + D.C.
• Data sources monitored35+
• Redirects preserved~980 URLs