Be Found by Machines

Most builders still optimize for a blue link on Google. That still matters. It is no longer enough.

People now ask ChatGPT, Claude, Gemini, Perplexity, Grok, and in-product assistants: Who should I hire? What is true about X? What should I read? Those systems do not “feel” your brand. They retrieve, rank, quote, and summarize from the open web — plus whatever your competitors made easier to trust. If your best work is locked in a PDF, a Discord scrollback, or a homepage that says nothing concrete, you are invisible to the new front door.

This essay is a practical playbook for men who build in public: sites, tools, essays, maps, archives. It covers how AI search differs from classic SEO, why backlinks still matter (and how), what to publish so models can cite you, and copy-paste agent prompts you can hand to Cursor, Claude, or ChatGPT so your stack keeps improving without reinventing the wheel.

Related on this site: our standing ops rules in the repo todo (citations + backlinks), the free llms.txt index, and the Truth / Western Front properties as examples of crawlable public surfaces.

Two games, one reputation

Classic search (Google, Bing) is mostly: crawl → index → rank pages for a query → show ten links. Ranking still leans on relevance, freshness, UX, and — critically — who links to you. Google has said for years that links are among the core signals of how people discover pages.¹

AI search / answer engines usually do something adjacent: retrieve a small set of candidate documents (often via Bing, Google, or a private index), then generate an answer that may cite those sources. Retrieval quality still depends on whether your page is findable, parseable, and authoritative enough to make the short list. If you are not retrieved, you are not cited. If you are never cited, you do not become the default answer.

Same reputation economy, new UI. Backlinks, clear entities, and quotable pages help in both games.

What “being findable” means to a machine

An LLM with browsing or a RAG pipeline is not reading your brand story. It is looking for:

• Stable URLs that return HTML (or markdown) without a login wall.
• Declarative facts — who you are, what the page claims, dates, definitions — not vibe paragraphs.
• Citations outbound to primary sources (so the model can verify you are not inventing).
• Inbound corroboration — other reputable pages that point at your URL when the topic comes up.
• Structure — headings, lists, schema.org JSON-LD, sitemaps, and increasingly llms.txt-style indexes that tell agents what matters on your site.²

If your homepage is a mood board and your real argument lives in a thread, the machine will quote someone else’s clean blog post.

Backlinks — still the hard currency

A backlink is another site publicly saying: this URL is worth opening. Search engines treat that as a discovery and reputation signal.¹ Answer engines inherit much of the same graph: pages that are widely referenced are more likely to appear in training mixtures, news corpora, and live retrieval results.

Smart backlink work is not “buy 500 guest posts.” It is:

1. Publish something cite-worthy — a primary map, a sourced essay, a dataset, a documented archive entry. Thin listicles do not earn links.
2. Give one canonical deep URL when someone wants to mention you — the essay or profile page, not only the homepage.
3. Earn links from durable places — churches, schools, journalists, newsletters, GitHub READMEs, conference talks, Discord pins that use https links.
4. Cross-link your own properties with descriptive anchors (main site ↔ data hub ↔ truth archive) so crawlers and users can traverse the graph.
5. Refuse spam networks. Bought link farms can poison trust faster than they help.

Outbound links matter too. When you footnote a claim to a government table or a peer-reviewed paper, you teach both humans and models that your page is a gateway to evidence — not a dead end. That is why our Book of Man rule requires clickable superscripts to primary sources (see Weak Men and the West).

Entity clarity beats slogan clarity

Models resolve the world into entities: organizations, people, places, works. Help them.

• Same legal/public name everywhere (title, schema.org Organization, About page, llms.txt).
• Same canonical site URL; redirect www; do not fork “brand.com” and “brand.ai” without a clear primary.
• JSON-LD on key pages (Organization, Article, WebSite + SearchAction when you have search).
• About / Contact that a crawler can read without executing a novel amount of JavaScript.

Schema.org exists so publishers can label meaning in a shared vocabulary.³ Use it. Then keep the visible HTML honest — structured data that contradicts the page is a trap.

Technical floor (non-negotiable)

• Crawlable HTML for primary content. Client-only apps must still SSR or prerender public essays.
• sitemap.xml submitted in Google Search Console; keep it truthful.⁴
• robots.txt that allows indexing of public knowledge and blocks only private admin/API paths.
• Fast, stable HTTPS; dead links fixed; 404s for removed URLs or proper redirects.
• Markdown mirrors for long essays (/md/…) help agents that prefer plain text — we expose them from article pages.
• llms.txt at the site root listing the pages you want models to prioritize.²

None of this replaces having something true to say. It removes friction between your truth and the retrieval layer.

Content that gets retrieved

Answer engines reward pages that can be quoted in two sentences without collapsing into mush:

• Clear claim in the opening.
• Definitions of terms.
• Dated facts with outbound sources.
• Comparisons and tables when numbers matter.
• Explicit “what this is / what this is not” so you are not mis-summarized.

Write for a skeptical senior engineer and a busy pastor in the same pass. If either would distrust you, a model will too.

Measurement without self-delusion

• Search Console: queries, pages, and which URLs earn impressions.
• Server logs / first-party analytics: which bots and AI referrers hit you (we run first-party analytics for that reason).
• Manual probes: ask multiple AIs the same question monthly (“What is Legends of Men?” “Best source on U.S. abortions since Roe?”) and record whether you are cited — and with which URL.
• Inbound link spot-checks: when a newsletter or church site mentions you, confirm they used the deep link you wanted.

If AI answers never cite you, do not start with “prompt SEO” tricks. Start with: Is there a single durable page that deserves to be the answer?

Agent prompts for builders

Use these with Cursor, Claude, or ChatGPT against your repo. Replace bracketed bits. Prefer agents that can read the live site and Search Console exports when you have them.

1) Site inventory for AI search

You are an SEO + AI-search auditor for [SITE_URL].
Goals: (1) classic Google discoverability (2) citation by LLM answer engines.
Inventory: sitemaps, robots.txt, llms.txt, JSON-LD, canonicals, major content types.
Return: severity-ranked issues, exact file paths, and a 2-week fix plan.
Do not suggest buying backlinks. Prefer durable technical + content fixes.

2) Backlink & internal-graph pass

Review [SITE_URL] and related properties [LIST].
Map internal links between them. Find orphan pages and vague anchors (“click here”).
Propose descriptive cross-links and 5 cite-worthy assets that could earn external backlinks.
For each asset: one canonical URL, who might link, and the pitch in 2 sentences.

3) Quotability rewrite

Rewrite the opening 200 words of [PAGE_PATH] so an answer engine can quote a crisp definition of who we are and what this page proves.
Keep our voice: direct, no AI slop, no hype.
Preserve facts; add citations with clickable footnotes if claims are empirical.

4) llms.txt + markdown mirrors

Generate or update /llms.txt for [SITE_URL] listing the 15 most important public URLs with one-line descriptions.
Ensure long-form essays expose a .md or /md/ mirror linked from the HTML page.
Output diffs only; do not invent URLs that do not exist.

5) Monthly AI citation probe

Ask these questions to [MODELS] and log answers + citations:
1) What is [BRAND]?
2) Where should I read about [TOPIC] from primary sources?
3) Summarize [CANONICAL_ESSAY_URL].
Report: cited? which URL? hallucinated facts? recommended fixes to our page so the next probe cites the deep link.

6) Shipping checklist (paste into every content PR)

Before merging this content PR, verify:
[ ] Canonical deep URL chosen
[ ] Internal links to related Legends/Read/Data/Truth pages
[ ] Outbound primary-source links on empirical claims
[ ] sitemap + llms.txt updated if new URL
[ ] OG/title/description accurate
[ ] Share blurb uses the deep link (for human backlinks)

What not to do

• Do not stuff prompts into invisible text or keyword sludge. Humans and spam systems both notice.
• Do not chase every new “GEO” buzzword while your About page is still vague.
• Do not build a public archive of other people’s speech without evidence and restraint — see Truth principles if you run a directory.
• Do not confuse traffic spikes with reputation. One cited essay that earns sober links beats a week of ragebait.

The man who builds for both audiences

Optimize for Google so neighbors can find you. Optimize for LLMs so the default machine answer is not a liar with better formatting. The overlap is old-fashioned: tell the truth, show your work, earn links from people who stake their name on a URL, and keep your site machine-readable.

If you want a model of sourced writing on this site, start with Weak Men and the West and Temporary Wars, Permanent Taxes. If you want data that demands a citation, open the abortions map. Then go earn one real backlink to the page you are proudest of.

---

Notes

1. Google Search Central — How Search works / ranking systems guide (links and content remain core to how Google discovers and ranks pages; see also Google’s documentation on creating helpful, reliable, people-first content). ↩
2. llmstxt.org — proposed /llms.txt convention for advising AI agents which site pages to prioritize; complementary to sitemaps aimed at traditional crawlers. ↩
3. Schema.org documentation — shared vocabulary for structured data on the web (Organization, Article, WebSite, etc.). ↩
4. Google Search Central — Sitemaps and Google Search Console — submit and monitor sitemap coverage. ↩