Orchestration is the layer that turns a pile of good content into a journey someone can actually travel. Most teams treat it as strategy. In practice it's a file you maintain, and when nobody maintains it, discovery just happens to you.

Third week applying Kellogg's I-MOS framework (Prof. Mohanbir Sawhney) to content operations. We've built Memory and Modular services. This week: Orchestration.

I'm going to show you mine, including the part that was broken.

For fifteen years, discovery mostly meant one thing. Rank on Google, earn the click. Orchestration was keyword mapping and internal links.

That still matters. It's no longer where a growing share of your buyers start.

They start by asking an AI. And an AI answer doesn't hand back ten blue links. It composes one answer, from a small set of sources it decided to trust.

The job moved from ranking to being cited. Being cited is a structural property of the page, not a flourish in the writing.

Three CSV files in a folder under version control. That's the whole system.

1. content-master.csv. One row per content piece: id, title, channel, status, url, slug, publish date, pillar, funnel stage, persona, primary CTA, offer tier, keyword target, last refreshed. Status moves through idea, briefed, drafting, review, scheduled, published, archived. This is the inventory. Without it I can't answer "what do I have" without opening twelve folders.

2. link-inventory.csv. One row per live link: source URL, target URL, anchor text, link type, date added, and the date I last confirmed it still resolved. Internal, external/out, inbound. This is the graph, and it's what lets me ask a question that no dashboard answers. Which of my pieces is under-linked, and what should be pointing at it?

3. link-opportunities.csv. Computed from the other two, never hand-maintained. Where a link should go, the suggested anchor, the rationale, the priority. I regenerate it rather than curate it, because a hand-curated list of things I should link is just a to-do list, and to-do lists rot.

CSV on purpose. They diff cleanly in git, they open in Excel, and no vendor can deprecate them.

The first time I ran the audit against a live post, it found a dead internal link.

An issue from June pointed at claritedge.com/p/make-ai-sound-like-your-brand. That slug never existed. The draft had guessed the URL, flagged its own guess in a note, and shipped anyway. The real post is The Fastest Way to Make AI Sound Like You, at a different slug. The link had been resolving to nothing the whole time.

Nobody emails you about a dead internal link. It doesn't throw an error or show up in a report. It quietly costs you the handoff between two pieces that were supposed to talk to each other, which is the entire point of orchestration.

That's the argument for keeping a record. Judgment finds the opportunity. Only a record catches the decay. It's the same reason I stopped treating the dashboard as the system and built a loop instead.

Here's a number I'd rather state than hide. claritedge.com has a Domain Rating of 1.1 (Ahrefs, checked Jul 30 2026).

That's near-zero backlink authority, and it makes the strategy obvious rather than difficult. I can't win competitive head terms against established content-marketing domains this year. No amount of craft closes that gap.

So I don't target them. I target long-tail, high-specificity questions where the results page is thin and expertise beats authority, and I optimize for being the clearest, best-sourced answer rather than the highest-ranked page. The angle map from a few issues back is where those questions come from: it turns a quarter of real customer questions into a calendar, and those questions are the exact prompts your buyers are typing into a model right now.

That isn't a consolation prize. Muck Rack's May 2026 analysis of AI citations found roughly 84% of the links AI tools cite are earned, non-paid media, with paid and advertorial content at a rounding error. You can't buy your way into the answer. Clear structure, real sourcing, and a named author are what a model can lift and attribute, and none of those are gated by domain authority the way blue-link ranking is.

Open a spreadsheet. One row per published piece. Add a column for every internal link it contains, and the date you last confirmed that link resolved.

You'll find something. The first pass always does.

Then do the second check: search your own category in ChatGPT, Perplexity, or Claude, and see whether you're in the answer.

Reply DISCOVER and I'll send the AEO starter: the three checks (citable, question-complete, sequenced) to run against one piece of your content, plus the CSV schema above so you're not designing the columns yourself.

And tell me what your first link audit turned up. Mine found a link pointing at a page that was never there.

— Jill

P.S. Next week closes the I-MOS arc. Governance: kill-switches, autonomy boundaries, and why human judgment is the one bottleneck worth keeping.