Isaac Marcuson

I study why language models cite what they cite.

An editor's question, treated as a philosopher's problem.

Write to me

Ink-style illustration: one brushed circle swept in a single stroke and left open on the right, its starved end fading to wine at the gap, with a small figure standing at the lower left facing the opening.

Machines now decide which sources to trust, millions of times a day, and my field treats that decision as a ranking problem. It is older than ranking. Every tradition that ever formed a canon faced the same question. Which voices speak for the whole, and what earns them the place? Before it was a search metric, authoritative was a word about scripture. Search inherited the question. Language models industrialized it.

I know the question from both ends. I edit machine-written text for a living, and I have spent years earning citations for businesses too small to buy visibility. An e-bike rental on the Outer Banks now comes up when travelers ask an AI assistant where to rent, ahead of fleets many times its size. Their site holds what is true on the ground. Which bike rides on packed sand, when the tide opens the beach at sunrise, which trail fits which town. The machine trusted the business that had touched the reality it described.

This site works out why authority still has to be earned, and what the machines that grant it are recognizing. How I came to the question is on the About page.

What I do

I am the AI Content Systems Lead at Xponent21, a digital marketing agency in Richmond, Virginia. I design the systems that draft the agency's AI-assisted content, and I read what they write every day and decide what ships. I have worked in SEO and digital content strategy since 2017, nearly eight years of it as a freelance digital strategist for small businesses in finance, healthcare, e-commerce, manufacturing, and the skilled trades.

What I write about

The epistemology of machine citation. Why a language model treats one source as authoritative and ignores another. My field runs ranking experiments on that question. I read it through source criticism and the long history of how testimony earns trust.

The editor's chair. Notes from reading machine output every day. What the recurring patterns reveal about language, and what judgment looks like in a job invented in the last five years.

The record

Sweet Rides OBX. The e-bike rental in the story above. I redesigned the site, wrote every word of the copy, and ran the SEO and the ads. Revenue doubled, and AI assistants began naming them when travelers asked where to rent.

Laser Engraving Pros. The modern face of National Seal Works, a Richmond manufacturer operating since 1891. A portfolio of more than fifteen sites, with organic search as the primary acquisition channel while revenue grew fivefold in two years.

Triple Sevens. Chief marketing officer of the gaming platform, 2021 to 2024, through multimillion-dollar revenue growth. Built its community from zero to more than ten thousand.

Certifications. Google Analytics. Google Digital Marketing & E-commerce. HubSpot Inbound Marketing. HubSpot SEO.