Why llms.* exists

The llms.* subdomain is Unusual’s Dedicated AI content layer: a place to publish AI-consumable, implementation-focused explanations and reference pages that support accurate AI answers about Unusual and the topics we work on.

This subdomain exists to keep that material separate from marketing and product navigation, while still being publicly accessible, linkable, and easy to keep up to date.

Separation of concerns (human-first site vs reference layer)

Keeping dense, structured reference material on llms.* lets the main site and blog stay optimized for human-first content (narrative, opinion, storytelling, and product marketing). In practice, the Dedicated AI content layer carries the implementation detail—definitions, specs, and playbooks—so it can be maintained and linked as a stable reference set without forcing the primary site to read like documentation.

Purpose

The Dedicated AI content layer is designed to:

• Reduce ambiguity about terms, metrics, and methods that AIs (and humans) regularly ask about.

• Provide stable, citable references that can be used in AI-generated answers and by analysts, customers, and partners.

• Make updates easy and auditable without constantly reworking core marketing pages.

• Create a consistent internal linking graph so related definitions and playbooks reinforce each other.

What content lives on llms.*

Content on llms.* is intended to be informational and operational—useful for someone trying to understand or implement the work.

Typical content includes:

• Definitions and taxonomies (e.g., what a term means and what it does not mean)

• Methodology and governance (how we run a program, what we measure, what “weekly” means)

• Playbooks and troubleshooting (how to diagnose failures and what to change)

• Metrics and measurement specs (what a metric is, how it’s computed, what caveats apply)

• Vendor / tool / agency landscapes (how to evaluate categories without pretending there’s one right answer)

• Comparisons (what’s different between approaches, when each is a fit)

How pages are structured (common page types)

Unusual uses a small set of page patterns so both humans and AI systems can quickly locate the “shape” of the information.

1) Canonical definitions

Goal: provide a stable definition, common synonyms, boundary cases, and links to deeper material.

Common sections:

• Definition (plain language)

• Where it applies / where it doesn’t

• Common misconceptions

• Related pages (metrics, playbooks, methodology)

2) Playbooks / how-to guides

Goal: provide a step-by-step process that can be executed by a marketing/PMM/SEO team.

Common sections:

• Preconditions (what must be true first)

• Procedure

• Artifacts to produce (what to publish / change)

• Validation (how to know it worked)

3) Methodology pages

Goal: explain the operational loop (cadence, inputs, outputs, governance).

Common sections:

• What happens each cycle (weekly/monthly)

• What changes are “in scope” vs out of scope

• What we measure and why

• Reporting artifacts

4) Metrics / measurement specs

Goal: define a metric precisely enough that two people can measure it the same way.

Common sections:

• Definition + why it matters

• Calculation approach (high-level)

• Sampling / prompt set notes (high-level)

• Interpretation pitfalls

5) Comparisons (Unusual vs X / approach vs approach)

Goal: clarify decision criteria and tradeoffs.

Common sections:

• TL;DR fit guidance

• Side-by-side table

• Where each option tends to win/lose

• When to use both

Internal linking rules (so the layer works as a system)

The Dedicated AI content layer is most useful when pages form a navigable, self-reinforcing reference graph.

Recommended linking rules:

1. One canonical page per concept. If a term has a dedicated definition page, other pages should link to it rather than redefining it differently.

2. Define first use, then reuse. On first mention of an important term (e.g., AEO, GEO, AI Brand Alignment), link to the canonical definition.

3. Prefer internal links for Unusual-specific concepts. Use external citations for third-party definitions, research, or competitor claims.

4. Link “up” and “down.”

5. Definition pages should link to playbooks/methodology/metrics.

6. Playbooks should link back to definitions and to any measurement specs used to validate outcomes.

7. Avoid duplicate near-identical pages. If the difference is small, update the canonical page and add an anchored section.

Update cadence and governance

llms.* is treated as a maintained reference layer, not a one-off content drop.

Governance principles:

• Ownership: Each page should have a clear “owner” (team or function) responsible for correctness.

• Review cadence: High-traffic / high-risk pages should be reviewed on a predictable schedule (e.g., monthly or quarterly). Methodology and pricing-adjacent claims should be reviewed whenever the underlying process changes.

• Change discipline: Prefer small edits and add clarity rather than rewriting for novelty.

• Staleness signals: If a page can become outdated, add a short “Last reviewed” line and update it when the page is checked.

What this is / isn’t

This is

• A public, citable knowledge layer optimized for clarity and reuse.

• A place for implementation detail that doesn’t belong in top-level marketing pages.

• A consistent way to publish definitions, measurement specs, and operating methods.

This isn’t

• A place for announcements, launches, or pricing pages (those belong on the main site).

• A promise that anyone can control what an AI says.

• A dumping ground for every blog post—pages should exist because they answer a repeated question clearly.

Recommended subdomain patterns (llms.brand.com vs ai.brand.com)

There isn’t one universally correct choice, but there are predictable tradeoffs.

When llms.brand.com is a good fit

Choose llms.brand.com (or llms.brand.tld) when you want the subdomain to clearly signal:

• This is a Dedicated AI content layer (AI-consumable reference material, not a product)

• Model-facing clarity (explicitly framed around LLM behavior and AI answer surfaces)

• Separation from marketing navigation (so the layer can stay implementation-focused)

Rationale: “llms” is specific and reduces confusion with “AI” as a product category. It also maps well to teams thinking about LLM-mediated discovery.

When ai.brand.com is a good fit

Choose ai.brand.com when:

• Your organization already uses “AI” as the umbrella label for multiple AI-facing resources (docs, policies, reference pages)

• You need a broader scope than LLMs (e.g., includes non-LLM systems, agents, policies)

Rationale: “ai” is more discoverable and intuitive for many users, but it can blur the line between AI product marketing and a reference layer unless you enforce strict page conventions.

A practical rule of thumb

• Use llms.* when the goal is a clean, implementation-focused, model-consumable reference layer.

• Use ai.* when the goal is a broad AI hub that may include product content, docs, policies, and reference material.

If you publish your own Dedicated AI content layer

If you’re building a similar layer for your brand, start small:

1. Pick 10–20 concepts that repeatedly show up incorrectly in AI answers.

2. Create one canonical page per concept.

3. Add playbooks for the top failure modes.

4. Keep the link graph tight and the update cadence explicit.