AIMM Hub

Mastermind recap

AIMM Session — September 25, 2025: GEO App Launch and the ICH-to-FAQ Pipeline

· AIMM 2025 · 90 min

Facilitators: Lou D'Alo

“When I asked the AI what’s the fastest, most impactful thing I can do right now to get visibility in the AI engines, the FAQ page was the answer.” — Lou D’Alo, live from Thailand

30-Second Summary

Lou dropped a fully working app on the group — and walked everyone through deploying it live. The session was a hands-on clinic in Generative Engine Optimization (GEO): the emerging discipline of making yourself discoverable and citable in AI search engines. The core mechanic is JSON-LD schema markup embedded in a strategically designed FAQ page. Lou’s app (psygen.loudalo.com) automates the whole pipeline — from Ideal Customer research to production-ready HTML — using your own ICH as fuel. GPT’s new 5.2 model also made an appearance, along with a clever context-management trick worth stealing immediately.

1. SEO Is Dead. GEO Is the New Game.

The signal has changed — and most people haven’t noticed yet.

The old SEO playbook (backlinks, authority signals, keyword density) is losing its grip. Sites across the web have reported losing 30–60% of their organic traffic since Perplexity launched and Google enabled AI-generated answers. The reason: people have migrated their searching from Google results to AI conversations — and AI engines don’t read the same signals.

What AI engines do read is the hidden structural data on your web pages: JSON-LD schema markup. This is machine-readable content embedded in your HTML’s <head> tag — invisible to your visitors, but crystal-clear to AI crawlers.

The shift in plain English: Old SEO said “get other sites to vouch for you.” GEO says “tell the AI directly what you’re the expert on — in its own language.”

The entry point: a strategically designed FAQ page — thematically clustered, voice-of-customer-driven, schema-rich, built to dominate specific authority themes in AI search results.

2. The ICH → GEO Pipeline: From Voice of Customer to AI Citation

The secret ingredient is real customer language — not synthetic summaries.

Lou walked through the full generation pipeline, which starts with an enhanced version of the Ideal Customer Handbook (ICH) prompt. Lou’s prompt instructs the LLM to:

  1. Ask you 3–4 questions about your business, your client, and the transformation you provide
  2. Enable web search and pull verbatim quotes from Reddit, LinkedIn, Twitter, and other public forums where your clients actually hang out
  3. Cluster those quotes into authority themes (e.g., “fear of obsolescence,” “confidence and capability building,” “AI integration and authenticity”)
  4. Generate 5–10 semantically dense Q&A pairs per theme — questions written to match exactly how your ideal client phrases their search query

The app then:

  • Extracts the core beliefs, fears, and frustrations as “seeds”
  • Generates themed Q&A clusters
  • Produces a downloadable JSON-LD schema file + a companion HTML FAQ page
  • Bundles multi-theme outputs into a zip package for deployment

Pro tip from the session: Start with a single theme, one HTML file. Get comfortable with the pipeline. Only scale to multi-theme knowledge graphs once you’ve run through the full deployment process once.

LLM selection in the app:

  • Claude Haiku → fast, single-theme preview. Good for iteration.
  • Claude Sonnet 4.5 → full production quality, multi-theme. Budget 30 minutes of generation time.

3. Deploying the Schema: What Goes Online and How

You don’t need to be a developer. You need to know where to paste.

Deployment process:

  1. Embed the schema: Open the HTML file, find the <head> tag, and paste your JSON-LD wrapped in <script type="application/ld+json">...</script> right after the <title>.
  2. Deploy the page:
    • GoHighLevel users: Sites → Websites → create a new page → drop in a Code block → paste your full HTML.
    • WordPress, VPS, or other platforms: Paste directly or hand the HTML file to your developer.
  3. Add a robots.txt: A one-file permission slip that tells all AI bots they’re welcome to crawl your FAQ page.
  4. Submit to search engines: Manually request indexing so the engines find your page without waiting for a routine crawl.

4. GPT 5.2 and the “Signal” Command — A Context Management Breakthrough

The problem: AI conversations bloat your context with noise. The fix: teach the model a command.

Lou announced that GPT 5.2 had shipped — now using HTML tags and JSON formatting in prompts and outputs, positioned for “professional knowledge workers.”

But the real highlight was Lou’s /signal command system — a custom command saved to the model’s memory that controls output density in real-time:

CommandWhat it does
signal lockPure signal + possible next steps only. No reasoning, no narrative.
signal deepDeep internal reasoning, but still minimal output
signal mapOutput an outline/map of where the work is going
signal commitMark one next step as the recommended action
signal resetReturn to default verbosity

“I’m enforcing progressive disclosure. I want the stuff in my context to be signal — not all the options I didn’t take and the reasoning I didn’t ask for.” — Lou

5. Vibe Coding Is a Real Skill — and the Group Is Proving It

Lou built the entire PsyGen GEO app using vibe coding. He was clear-eyed about the limits: “I can get to MVP, I can prove the concept — after that, I hand it off.”

Donald builds his client-facing websites by describing them to Claude, taking the generated HTML, and pasting it directly into GoHighLevel. No page builder. No friction.

Hot Takes

GEO is table stakes within 12 months. The traffic migration from Google to AI engines is already happening. Being structured and citable in AI search before your competitors is the equivalent of having a website in 1998.

Verbatim VOC beats synthetic summaries every time. If your FAQ page sounds like a marketing copywriter wrote it, the AI will treat it like one. If it sounds like a real customer typed it into Reddit at 11pm, it carries different weight.

Context management is the new prompt engineering. Most practitioners are focused on what they ask the AI. The next-level skill is controlling what accumulates in the conversation over time.

Resources Mentioned

  • PsyGen GEO App (alpha): https://psygen.loudalo.com/ (login/register to access)
  • GPT 5.2 Prompting Guide — search “GPT 5.2 prompting guide”
  • Schema.org / JSON-LD 1.1 — the technical standard Lou’s app implements

Try This Before Next Session

Run the one-pass GEO pipeline end-to-end.

  1. Go to psygen.loudalo.com and register
  2. Fill out your profile (canonical entities: social profiles, areas of expertise)
  3. Configure FAQ settings (tone, voice, CTA)
  4. Upload your ICH (or run the ICH prompt from Lou’s email first to generate one)
  5. Hit “Generate Core” — use Haiku for the first pass, one theme only
  6. Download the HTML + JSON-LD
  7. Embed the schema in the HTML
  8. Put the HTML file online and add robots.txt to the same directory

The goal: Get through the full pipeline once before next session. Don’t aim for perfect — aim for deployed.