Midnight GEO Pro Pack · v1.2
Your site might be invisible to AI answers. This pack finds out — and shows you what to fix.
ChatGPT, Perplexity, Gemini, and AI Mode now answer questions your customers used to Google. If their crawlers can't fetch your pages, or your passages aren't quotable, you don't exist in the answer — no matter how you rank. This is the GEO/AEO playbook I run against my own 18-domain portfolio: 15 Claude Code skills, the probe harness, my actual audit, and 12 months of updates. $99 one-time.
one-time · Stripe Checkout · license key + one-command install (npx geo-pro), delivered instantly
why this exists
Blue-link rank stopped predicting citations.
After Gemini 3 became the default for AI Overviews and AI Mode in January 2026, citation-vs-top-10 overlap collapsed from ~76% to ~38%. About 44% of citations come from the first 30% of the page body — the engines lift the passage that answers, not the page that ranks. And step zero: if the bot is blocked at the WAF or in robots.txt, none of it matters. Both figures verified April 2026; the bot-reach preflight checks it in minutes.
how it works
Probe → fix → re-probe. That's the whole loop.
Run the probe on your domain
One command asks ChatGPT, Perplexity, Gemini, and AI Mode about your topics and records who cites you, who cites your competitors, and what passages got lifted. The bot-reach preflight runs first: if the bots can't fetch you, everything else is moot.
Read the fix list, apply it
The skills score your pages for extractability — answer-first passages, quotable structure, dated claims — and the evidence chapter shows the exact fixes I applied across 18 real domains. You change what the engines can actually see and lift.
Re-probe and keep the delta
Run the probe again and compare. Citation movement is the only metric that matters in GEO — impressions stopped predicting it. The pack keeps your runs dated so you can prove what changed and when.
what's in the pack
Six things, unboxed. The skills, the named staff, the proof, the code, the updates, the support.
15 Claude Code skills — the actual operating layer
Drop them into Claude Code and run them against your own domains. One skill probes ChatGPT, Perplexity, Gemini, Claude, and Google AI Mode with your URLs and tells you: cited, not cited, and why. Another scores your money pages for extractable passages — can the engine lift a quotable answer off the page, or does it skip you? A third is bot-reach preflight: is GPTBot/ClaudeBot/PerplexityBot even allowed to fetch you, or are you 403-blocked without knowing it. Plus Preferred Sources triage, indexer, entity, and topical-map skills.
Installs as a named team — eight specialists + a chief, not a folder of files
Same 16 skills, assigned once each — the chief also carries agent-flywheel, the operating discipline for running the staff — as Hermes Bot Mode profiles and Grok Bot static cards: midnight-probe, midnight-audit, midnight-passages, midnight-entity, midnight-map, midnight-acquire, midnight-report, midnight-support, plus midnight-chief to route. You still run the files. No cloud staff, no citation guarantee, no extra service beyond the email terms.
The worked 18-domain audit — read my real homework
A buyer-facing chapter with the actual GEO crawl scores, bot-block evidence, and GSC snapshots from my 18-domain portfolio. Property names left in on purpose — you see exactly what an operator's audit looks like, including the sites that got blocked and why. Copy the method, not just the conclusions.
ChatGPT probe harness — runnable code, not theory
The web.run extractor I use to inspect what ChatGPT actually fetches and cites, plus a one-page protocol (temporary chats, memory off, reproducible runs). If you ever wondered why ChatGPT cites your competitor and not you, this is the tool that answers it.
12 months of updates — the pack moves when the engines move
Answer engines change their crawl and citation behavior constantly. Every pack update for 12 months from your Stripe receipt: new zip versions, same seat, same license, delivered by email. You'll re-probe against new engine behavior without paying again.
Email support — one clarification round, direct
support@midnightdev.dev — questions on applying the skills to your properties, one clarification round per buyer after you've actually run a skill. Not a retainer, not a Slack, not a course community.
who it's for
You own or operate sites you want cited in AI answers — your own portfolio, client sites, an agency practice. You use Claude Code (or any agent that can run files and shell commands). You want the operating layer an actual operator uses, including the parts that didn't work. If you want a service where someone does it for you, this isn't that — the consulting page is.
what this is not
- Not a course. No videos, Discord, or office hours.
- Not a dashboard. You get files and a license grant, not a login.
- Not a subscription. $99 one-time. Updates for 12 months, then you keep what you have.
- Not a citation guarantee. Playbooks, not a ranking or traffic promise.
- Not done-for-you. I don't run the playbooks on your domains — you do, with these files.
support terms
After Stripe sends the receipt, email support@midnightdev.dev with the receipt number. The zip and license are delivered by email, same day whenever possible. Teams that need more than one seat, email first — do not buy N times and hope.
Included: email questions on applying the skills, one clarification round per buyer, pack-file updates for 12 months. Not included: site audits, calls, screenshares, Slack, custom skill writing, or anyone running the playbook against your domain for you.
← back to the workfolioThe pack, one more time.
15 Claude Code skills that probe four answer engines and score your pages — also installable as eight specialists + a chief. The worked 18-domain audit. The runnable ChatGPT probe harness. 12 months of updates. Email support. One seat, $99, no subscription. If the probe shows you're already cited everywhere, you learned that for the cost of a dinner.
one-time · Stripe Checkout · license key + one-command install (npx geo-pro), delivered instantly
Next step
Want to talk it through?
Book a 30-minute call. I will tell you plainly whether I can help and what I would test first.