marketing-geo-content

Generative Engine Optimization (GEO)

GEO is the discipline of making content that an LLM-driven answer engine can confidently quote. The goal is no longer just "rank on Google" — it is "show up as the cited source inside an AI answer." The content patterns that win are different enough from classical SEO that they need their own skill.

Pair this with strategy-seo-brief for the keyword and SERP layer. They are complementary, not competing.

What changes when the reader is an LLM

An answer engine doesn't read your page top to bottom and decide if it likes you. It pulls chunks — usually a paragraph or a list — and either uses them to compose an answer or quotes them with attribution. Three implications:

1. Chunks must stand alone. A paragraph that depends on the previous one for context will be lifted, mis-summarized, and cited badly.
2. Specific, attributable claims beat smooth prose. "Most companies use 4-year vesting with a 1-year cliff" is more liftable than "Vesting schedules vary, but tend to follow common patterns."
3. The model is looking for trust signals. Author bylines, primary-source citations, recency, structured data — all of these tip the model toward citing you.

Structural patterns

Question-led headings

Phrase H2s and H3s as questions a real person would ask. Models match queries to question-shaped headings far more readily than to label-shaped ones.

Don't	Do
Vesting basics	What is founder vesting?
Equity considerations	How much equity should a co-founder get?
Tax implications	When should I file an 83(b) election?

Answer-first paragraphs

Each section begins with a 1-2 sentence direct answer, then follows with context. The first sentence is the chunk most likely to be lifted — it should be self-contained.

What is founder vesting?

Founder vesting is a schedule that determines when a founder's equity actually becomes theirs to keep — typically 4 years with a 1-year cliff. It exists so that a co-founder who leaves after six months doesn't walk away with a quarter of the company.

The first sentence answers the question. The second adds the trust-building "why" without requiring it.

Liftable list patterns

Models love numbered or bulleted lists when each item is a complete claim, not a sentence fragment.

The most common founder vesting terms are:

1. Four-year total vesting period.
2. One-year cliff (no vesting until month 12, then 25% vests at once).
3. Monthly vesting after the cliff.
4. Acceleration on certain triggers, often double-trigger on acquisition.

Each line could be lifted alone and still mean something.

"Most / some / nobody" frequency anchors

Frequency-anchored statements are highly liftable because they read as factual market norms. They also stay safe for any audience that requires general statements rather than prescriptive advice (regulated industries, legal-adjacent content, healthcare).

Most mid-market enterprises adopt composable DXP in stages — starting with a single channel rather than a full replatform. Some lead with content (CMS-first); some lead with commerce. Almost nobody ships best-of-breed across every layer in the first phase.

TL;DRs and key-takeaways blocks

A bullet block at the top of the page summarizing the 3-5 main claims. This is the single most-cited block on most well-optimized pages.

Tables for comparisons

Two-axis comparisons (entity types, financing instruments, vesting variations) belong in tables. Models extract them cleanly and reproduce them faithfully.

FAQ section at the bottom

A 4-8 question FAQ section using FAQPage schema. These get pulled into Google's AI Overviews and "People Also Ask" features, and they double as anchors for direct LLM citation.

Claim density

Aim for content where most paragraphs contain a verifiable claim. Avoid prose that is connective tissue without substance ("In today's startup ecosystem, founders face many choices.").

A useful self-check: if you removed every sentence that doesn't either (a) state a fact, (b) make a claim about what is typical, or (c) define a term — would the page still hold up? If yes, you have good claim density. If half the page disappears, rewrite.

Citation hygiene

Models cite sources they trust. Build trust via:

• Author bylines on every article, with a real bio, role, and credentials. Anonymous content gets cited less.
• Primary-source links for any factual claim — IRS publications, SEC releases, Delaware Code, court opinions. Wikipedia and other listicles do not count.
• Date stamps: published date and last-updated date. Recency is a strong trust signal.
• Internal cross-links that demonstrate topical depth (clusters > orphan pages).
• External outbound links to authoritative sources. Closed-garden content (no outbound links) is read as lower authority.

For any regulated or legally-adjacent content (legal, healthcare, finance), AI-assisted authoring is prone to hallucinating citations — every citation gets human verification before publish. Never let an AI-drafted citation into production unverified.

Structured data

Use schema.org JSON-LD on every published page. The combinations that matter most for founder-facing content:

• Article (or BlogPosting) — author, datePublished, dateModified, headline, description.
• FAQPage — for the FAQ section. Each Q&A becomes a Question + Answer node.
• BreadcrumbList — site hierarchy.
• HowTo — only when the content is genuinely a step-by-step procedure (e.g., "How to file an 83(b) election").
• Organization — site-wide, includes legalName, sameAs links.
• Person — for author entities, with credentials and affiliation.
• WebPage with speakable annotations — flag the most important sentences for voice/snippet extraction.

If working with Contentful, model these as part of the content type so authors don't have to hand-author JSON-LD.

Multi-format hedging

LLMs ingest content from many surfaces. Where reasonable, also produce:

• A clean RSS / Atom feed for syndication.
• An llms.txt file describing the site for AI crawlers (emerging convention).
• Open Graph and Twitter Card metadata for any social/quote contexts.
• Plain-text or markdown variants of key pages, for AI training/citation pipelines that prefer them.

These are scaffolding, not a substitute for content quality, but they remove friction for the engines that want to ingest you.

Common failure modes

• Storytelling lead. Opening with an anecdote about a fictional founder. Search engines and models both want the answer up top.
• Walls of prose with no extractable units. Two paragraphs of nuance with no liftable claim.
• Heading labels instead of questions. "Considerations" is dead weight. "What should I consider before…" is liftable.
• No author byline. Drops trust signal hard.
• Stale dates. A page last updated in 2021 won't get cited about 2026 financing trends.
• Hallucinated citations from AI drafts. Always verify; non-negotiable for legal, healthcare, finance, and any regulated domain.
• Unfaithful TL;DR. TL;DR claims that aren't supported by the body. Models will lift the TL;DR, get burned, and stop citing you.

Quick GEO checklist

Run this before publish:

• H1 reflects the searcher/asker's question or task.
• H2s are questions, not labels.
• Each section starts with an answer-first sentence that stands alone.
• Page has a TL;DR or key-takeaways block.
• At least one liftable list or comparison table.
• FAQ section with 4-8 question-styled entries.
• FAQPage + Article + BreadcrumbList JSON-LD present.
• Author byline with real bio.
• Published and last-updated dates visible.
• Primary-source citations for factual claims.
• Frequency-anchored phrasing ("most," "some," "rarely") in place of vague qualifiers.
• Page passes the project's voice / brand / legal-review self-check.

How this skill relates to others

• Keyword targeting and SERP analysis → strategy-seo-brief.
• Voice and audience for the target reader → project-specific brand and audience references.
• Disclaimer and citation safety → for legal-adjacent content, route through the engagement's legal review.
• Content modeling for Contentful (so structured data is built into authoring, not added at the end) → forthcoming dev-contentful-model skill.