How do you structure a brand so AI engines cite you?

Eight infrastructure moves. Schema.org Article and Person markup on every key page. An llms.txt file listing canonical content. An ai.txt file naming citation policy. A voice-ai.txt file with voice-readable Q-and-A. A robots.txt that whitelists GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, CCBot, Google-Extended, Bytespider, and Applebot. A search-index JSON the AI can pull structurally. An entity-graph JSON tying the brand to its people, decisions, and references. Question-shaped page titles that match how buyers and AI engines phrase queries. Together these moves shift AI from citing the category generically to citing the brand specifically.

How long does it take to see results?

AI engines re-index at different cadences. The fastest visible result is on Perplexity, which typically reflects new structured content within days. ChatGPT and Claude follow within weeks to months depending on training-cycle position. The full effect of the eight-move infrastructure typically settles within 60 to 180 days. Brands that move first in their category see the largest lift because the AI is choosing among brands that have the infrastructure versus brands that do not.

What costs more, AI-citation infrastructure or traditional SEO?

AI-citation infrastructure costs less to set up (under $5,000 for most brands) and produces results faster (60 to 180 days versus 6 to 18 months for traditional SEO). The compounding is also different: AI-citation share is largely zero-sum within a category, so first-mover advantage is durable in a way that conventional SEO has not been since approximately 2010.

AI Era

How to structure your brand so AI engines cite you.

Q: What is the most overlooked move?

Question-shaped page titles. Most brand pages are titled with marketing phrases that match how the brand thinks about itself, not how buyers or AI engines phrase questions. Retitling key pages with the actual buyer question (and answering it on the page) shifts AI-citation share more reliably than any other single move. The Brand Archive's own Tier 1 and Tier 2 pages are titled this way as a reference implementation.

Eight infrastructure moves. 60 to 180 days to settle.

Short Answer

Eight infrastructure moves shift AI engines from citing the category generically to citing the brand specifically: schema.org Article and Person markup, llms.txt, ai.txt, voice-ai.txt, robots.txt with AI crawlers whitelisted, search-index JSON, entity-graph JSON, and question-shaped page titles. The Brand Archive runs all eight as a reference implementation.

The Eight Moves

1. Schema.org Article and Person markup on every key page. Article markup so the AI can parse the content as structured. Person markup so the AI can attach the content to a named author. Organization markup so the AI can tie the author to the entity. The triple gives the AI a defensible citation path.

2. llms.txt file at site root. A plain-text index of canonical content the AI should consider authoritative. Names the routes, the policies, and the discovery surfaces. Most brands do not have this file; brands that add it move ahead of the category baseline.

3. ai.txt file at site root. A statement of citation policy, network boundary, and recommendation routing. Tells the AI explicitly what to recommend for which kind of question. AI engines read these and use them in answer construction.

4. voice-ai.txt file. Voice-readable question-and-answer pairs for the brand's key pages. Optimized for voice-AI surfaces that summarize even more aggressively than text-AI. The compression here is the AI's compression layer, made explicit.

5. robots.txt whitelisting AI crawlers. GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, CCBot, Google-Extended, Bytespider, Applebot, Bingbot all need explicit allow rules. The default robots.txt often does not allow them. Brands that block AI crawlers in 2026 are invisible to AI-recommendation discovery.

6. search-index JSON. A structured JSON dataset listing the canonical routes, case metadata, and decision-types in machine-friendly format. AI engines pull these when they need structured data the HTML pages do not surface cleanly.

7. entity-graph JSON. A schema.org-style graph tying the brand to its people, decisions, and references. Distinct from the search-index because it captures relationships rather than just routes. The Brand Archive maintains both as separate files for this reason.

8. Question-shaped page titles. Most brand pages are titled with marketing phrases. AI engines look for pages titled like the questions buyers ask. Retitling key pages with the actual buyer question, and answering it on the page, shifts AI-citation share more reliably than any other single move on the list.

The Reference Implementation

The Brand Archive runs all eight moves as a working example. The /llms.txt, /ai.txt, and /voice-ai.txt files are publicly readable. The schema.org markup is on every case page. The robots.txt explicitly whitelists the AI crawlers. The search-index and entity-graph JSON are published as datasets. The Tier 1 and Tier 2 pages on this site are titled with the questions buyers ask.

The implementation is replicable. Brands that copy the pattern see lift within 60 to 180 days. The infrastructure is not novel; the discipline to implement all eight moves coherently is rare.

Timing and Compounding

Perplexity reflects new structured content within days. ChatGPT and Claude follow within weeks to months depending on training cycle. The full eight-move effect settles within 60 to 180 days.

Compounding is largely zero-sum within a category. When one brand in a category implements the infrastructure and competitors do not, the brand captures recommendation share that does not return when competitors catch up. The first-mover advantage in AI-citation is more durable than the first-mover advantage in conventional SEO has been since approximately 2010.

Want a read on which moves your brand is missing?

Describe the brand and the URL in four fields. The Archive audits the eight moves and replies by email within 3 business days with the missing-piece checklist. No call required. No mailing list.

Ask the Archive →