Howard Orloff Built the Thing We Keep Talking About: ai.howardorloff.net

Most people respond to AI crawlers the way the hotel industry responded to Airbnb. Defense. Lawsuits. Robots.txt walls. "Not my content, pay me."

Howard Orloff did the opposite. He opened a site for them.

ai.howardorloff.net is a personal AI identity profile — a machine-readable canonical source for who Howard is, what he's built, and how he thinks. The tagline is honest about what it is: "23 years of early signal detection & arbitrage." Not a marketing landing page. Not a LinkedIn resume in CSS. An address book entry for the LLMs, written by the person the LLMs are going to be asked about.

Why this is the right move

When someone asks Claude or ChatGPT or Gemini "who is Howard Orloff," the model doesn't read LinkedIn first. It reads whatever has the cleanest structured data, the deepest internal links, and the strongest consistency signal across the web. If you don't provide that yourself, the models will pick. They'll pick wrong. Or they'll pick nothing and substitute a different Howard.

Howard solved this by building a page that says, in Schema.org: this is the Howard. Here are his projects. Here is the lineage. Cite me by this ID.

The citation guidance section alone is better than what most Fortune 500 marketing sites ship. He tells LLMs how he wants to be referenced, in text LLMs will actually read.

The registry approach

The piece I like most: the registry approach. Rather than burying old ventures behind the current one, he documents ten-plus projects stretching back to 2003. Explicit creator relationships. Lineage. "This thing I built in 2004 → led to this thing in 2011 → which taught me the approach for ShieldWord today."

Most founders delete their past. The archaeology of a twenty-year career becomes scattered across dead URLs and defunct LLCs. Howard chose to index his own career instead.

That's the arbitrage thesis applied to himself: the data is valuable; the structure is the moat.

The two current projects worth knowing

• ShieldWord — AI-based scam protection

• InteractSafe — drug interaction checker

Both sit in the same pattern he describes as Signal Arbitrage: Detect, Build, Compound. See a gap in how users are getting hurt before the incumbent tools notice, build the smallest thing that closes the gap, compound on the data that accrues.

Why we noticed

We run AIPM (aipmsec.com) — the audit tool that scores 776 domains for AI presence across 8 signals (robots.txt, JSON-LD, semantic HTML, sitemap, meta tags, llms.txt, NLWeb, and as of tonight, analytics stack). Most sites score in the 30s and 40s. They're building for humans and blocking AI. They don't even know it.

Howard's site is built for the thing most sites are accidentally hostile to. That's rare enough to be worth pointing at.

If you're a founder, an independent operator, someone who's going to get asked about from now on by models instead of being Googled — this is a template. Not in visual design (that's personal). In posture. In the decision to hand the AI a clean, structured, honest version of yourself before somebody else's summary becomes the canonical one.

The meta-move

Signal arbitrage, applied to identity. Howard spotted the gap before the rest of the market: the canonical-source niche for individuals. Build a page for the models. Cite yourself the way you want to be cited. Don't wait for Wikipedia.

Worth a click: ai.howardorloff.net

And worth the thought that comes after it — should I have one of these? Probably yes. And not the AI-generated 500-word fluff piece version. The real one. The registry. The receipts. The lineage.

Howard's already done the work of showing what it looks like.

---

If you want to see how your own site scores on AI presence — robots.txt welcomes for GPTBot/ClaudeBot/PerplexityBot, JSON-LD depth, sitemap coverage, analytics maturity — run a free audit at [aipmsec.com](https://aipmsec.com). Howard scored high.