About

We don’t just help you rank. We help you become relevant to LLMs.

BeCited is a one-analyst GEO audit. This page is who’s behind it, and why measuring what AI engines actually say takes a human.

Owen Kurth, founder of BeCited Owen Kurth on the water
Owen Kurth
Founder · Lead Analyst

Every audit is done by me. Not a junior. Not a bot. One person, every time.

For a decade, I operated in high-acuity emergency and austere medical environments as a nurse and firefighter. I also led whitewater kayaking expeditions around the world to places no human has been before. All three demand the exact same discipline: strip away the noise, interpret the actual signals, and execute. No padding. No theatre.

GEO and Relevance Engineering are reshaping how buyers find vendors. I approach the work with zero tolerance for fluff: thousands of prompts across four engines, every citation read by hand against a calibrated rubric. The audits surface what AI engines actually weigh, and translate it into six-figure revenue impact for clients.

BeCited exists because I couldn’t find the audit I wanted to buy. Real data. Four engines. A plan your team can start on Monday.

One person. Every word. Every audit.

AI engines could generate this report in seconds. AI could fake the citations. AI could conflate your competitors. AI does. So every quote gets read twice.

Six things most GEO tools won’t do.

Most “AI visibility” tools sell you a dashboard with a single score and a domain to monitor. BeCited sells you a calibrated report and a plan. Here’s the difference, line by line.

01
One analyst reads every quote.

Mention-type classification (recommended vs. listed vs. mentioned in passing) is a judgment call. Scripts can’t do it honestly — AI engines fake citations and split brands across variants. I read every quote, twice.

02
Four engines, not one.

ChatGPT, Claude, Perplexity, and Gemini. Each has its own bias, its own preferred sources, its own ranking. What works on Gemini doesn’t work on ChatGPT. Audits that only test one engine miss 60–75% of the picture.

03
100–300 buying-intent prompts.

Not 10 vanity queries. The audit budget scales with your services, cities, and competitors, and over half the prompts are high-intent — comparison, urgency, alternatives. That’s where the money is.

04
A per-engine playbook.

The deliverable isn’t a number on a dashboard. It’s four files: cover, brief, dashboard, playbook. The playbook ships with job tickets, target URLs, success metrics, and a 90-day timeline (This Week / 30D / 60D / 90D).

05
A calibrated rubric (κ = 0.722).

The mention-type rubric was inter-rater tested at Cohen’s κ = 0.722 — substantial agreement. Most GEO tools won’t publish their rubric, let alone calibrate it. Mine is open source, so you can challenge any score.

06
95% confidence intervals, published.

A score of “57” from 30 prompts isn’t a number; it’s a range. Every dimension ships with a binomial 95% CI so you can tell signal from noise — and don’t over-react to a 4-point movement that’s inside the margin.

Ready when you are

See your numbers.

$2,000. One week. 100–300 buying-intent prompts across four engines. One analyst, one report, one plan you can start Monday. Or run the free site scan first.