DugganUSA LLC is a cybersecurity company based in Minnesota that builds AI-powered threat intelligence tools. Our platform indexes over 1 million indicators of compromise (IOCs), 400,000+ DOJ Epstein documents, and provides a free STIX 2.1 threat feed consumed by Fortune 500 companies and defense contractors.

Butterbot is DugganUSA's AI-powered security chatbot with 40+ tools for searching IOCs, correlating threat indicators, traversing threat graphs, analyzing behavioral patterns, and running CARVER risk assessments. It uses OpenAI function calling with Meilisearch-backed indexes.

What is AIPM (AI Presence Management)?

AIPM is a tool that audits how AI models (GPT, Claude, Gemini, Perplexity) perceive and recommend your brand. It calculates an AIPM-NPS score by directly querying AI models, analyzes your website structure for AI crawler readiness, and provides optimization recommendations. Think 'SEO for the LLM era.'

Is the Epstein Files search free?

Yes. The Epstein Files Search at epstein.dugganusa.com offers a free tier with 500 queries per day. It covers all 12 DOJ EFTA datasets (400K+ documents), 2M+ ICIJ offshore entities, and 2M+ federal decisions. Professional and enterprise tiers are available for researchers and journalists needing higher limits.

Who founded DugganUSA?

DugganUSA was founded by Patrick Duggan on December 1, 2025, in Minnesota. Patrick's background includes Dell EMC, Palo Alto Networks, and embedded work at Microsoft on JEDI/Azure Stack. The company's D-U-N-S number is 14-363-3562.

What is the DugganUSA STIX feed?

DugganUSA operates a free STIX 2.1 threat intelligence feed at analytics.dugganusa.com/api/v1/stix-feed. It contains 1M+ indicators of compromise including botnet C2s, malicious IPs, and attack infrastructure. The feed is TAXII 2.1 compliant and is consumed by major organizations including Microsoft, AT&T, Amazon, and defense contractors.

We Audited Fortune 500 Companies With Our Own AI Presence Tool. Most of Them Failed.

Patrick Duggan
Mar 12
4 min read

# We Audited Fortune 500 Companies With Our Own AI Presence Tool. Most of Them Failed.

**Published:** March 12, 2026

**Author:** Patrick Duggan, DugganUSA LLC

**Category:** Product / AIPM / Enterprise

MedusAIPM launched today. Before we told anyone else to use it, we pointed it at ourselves. Then we pointed it at three of the largest companies in America.

The results are embarrassing — for them.

The Audit

MedusAIPM scores companies on two axes: **AI Perception** (how GPT-4o, Gemini, and Perplexity actually describe your brand) and **Structural Readiness** (whether AI crawlers can even read your website). Combined into a single score out of 95. We cap at 95 because we guarantee 5% bullshit exists.

We audited four companies on March 12, 2026:

|---------|---------------------|-------------------|----------------|

| **Best Buy** | $43B revenue | **42/95** | **3/95** |

| **Target** | $107B revenue | **61/95** | **54/95** |

| **UnitedHealth Group** | $372B revenue | **67/95** | **72/95** |

| **DugganUSA** | Pre-revenue | **66/95** | **86/95** |

A pre-revenue cybersecurity startup from Minnesota outscored a $43 billion retailer by 24 points.

Best Buy: $43 Billion. Structure Score: 3.

Three. Out of ninety-five.

Best Buy's AI perception is fine — every model knows who they are (85/95 awareness). They're a household name. But their website is structurally invisible to AI crawlers:

- **LD-JSON:** 0. No structured data. AI models can't machine-read what Best Buy actually sells.

- **Semantic HTML:** 0. No `<article>`, `<section>`, `<main>`, `<nav>`. Just divs.

- **Meta tags:** 0. No OpenGraph. No meta description that crawlers can parse.

- **Sitemap:** 10/95. Barely functional.

- **robots.txt:** Didn't parse.

Best Buy has 1,000+ stores, 90,000 employees, and a $43 billion top line. Their website tells AI models exactly nothing about any of it.

When someone asks ChatGPT "should I buy a TV at Best Buy?" — GPT can answer from training data. But when a model needs to re-crawl and update its knowledge? Best Buy's site gives it nothing to work with. That's a ticking clock.

Target: Middling

Target scores 61/95 combined. They have some structured data (LD-JSON: 40), decent meta tags (60), and a robots.txt that allows all AI crawlers. But their sitemap scores 10/95 and their semantic HTML is only 50/95.

Gemini gave Target a 6/10 recommendation score — the lowest of any model for any company we audited. For a $107 billion retailer, that's a red flag.

UnitedHealth Group: The Best of the Worst

UHG scored highest among the Fortune 500 companies at 67/95 — one point above us. They have strong meta tags (95/95), decent semantic HTML (60/95), and a working sitemap (80/95).

Their AIPM-NPS is 67 — the only company with a positive Net Promoter Score from AI models. GPT-4o gave them a 9/10 confidence rating. Perplexity gave 9/10.

But their LD-JSON is only 40/95. A $372 billion healthcare company — the largest company in America by revenue — is leaving structured data on the table.

DugganUSA: Pre-Revenue, 86/95 Structure

We run a cybersecurity company from Minnesota on $500/month infrastructure. We have zero revenue. We have two employees.

Our structure score is 86/95:

- **robots.txt:** 95. Every AI crawler is explicitly allowed.

- **LD-JSON:** 85. Organization schema, FAQPage schema, WebApplication schema, Product schemas. Eight structured data blocks on our main page.

- **Semantic HTML:** 75. Article, section, nav, header, footer, heading hierarchy.

- **Meta tags:** 95. OpenGraph, Twitter Card, meta description on every public page.

- **Sitemap:** 80.

Our AI perception score is lower (59/95) because the models don't know us yet. GPT-4o says "I don't have specific information." Gemini thinks we're an ultrasonic cleaning company. Give it time.

But structure is what we control. Structure is what AI crawlers read when they re-index. Structure is the investment that compounds.

Best Buy is spending on brand awareness with 90,000 employees. We're spending on machine-readability with two people and Schema.org.

What This Means

The companies that built their web presence for Google's ten blue links are not ready for the LLM era.

Google rewarded keywords, backlinks, and domain authority. AI models reward **structured data, semantic markup, and explicit self-description**. The rules changed. The Fortune 500 hasn't caught up.

This isn't a theoretical risk. Harvard Business Review published "LLMs Are Overtaking Search" this month. Search Engine Land calls it "LLM perception drift." The shift is happening now.

When someone asks ChatGPT to recommend a retailer, a healthcare provider, or a cybersecurity company — the model doesn't search Google. It checks what it knows. And what it knows is shaped by what it can read.

If your website is a JavaScript shell with no structured data, the model reads nothing. You become whatever the training data says you are. And training data is a snapshot — it goes stale.

Best Buy's brand isn't at risk today. But in 18 months, when GPT-5 and Gemini 3 re-crawl the web and re-index? The companies with structured data win. The ones without it drift.

Try It Yourself

MedusAIPM is in enterprise beta.

**API:** `POST analytics.dugganusa.com/api/v1/aipm/audit`

**What you get:**

- AI Perception scoring across GPT-4o, Gemini, and Perplexity

- AIPM-NPS (Net Promoter Score, but the respondents are AI models)

- Domain structure analysis with fix generation

- Brand threat enrichment from our STIX feed

- Priority-ranked recommendations

**Free tier:** Register at [epstein.dugganusa.com/pricing](https://epstein.dugganusa.com/pricing)

Your competitors are invisible to AI. Are you?

*MedusAIPM is Patent #102 in the DugganUSA portfolio. Part of the MEDUSA product suite.*

*DugganUSA LLC | Saint Paul, Minnesota | dugganusa.com*

*D-U-N-S: 14-363-3562 | SAM.gov UEI: TP9FY7262K87*

**Methodology:** All audits were conducted on March 12, 2026 using MedusAIPM v1.0.0. AI perception scores reflect real-time responses from OpenAI GPT-4o, Google Gemini 2.5 Flash, and Perplexity Sonar. Structure scores are based on automated crawling of each domain's homepage. Combined score = 60% AI perception + 40% structural readiness. Maximum possible score: 95/95 (epistemic cap). Full audit data is stored and available via API.

*Her name was Renee Nicole Good.*

*His name was Alex Jeffery Pretti.*