When a prospect asks ChatGPT "who is the best marketing partner for a $10M Australian B2B business", the answer they get either includes you or doesn't. That answer is now driving real pipeline for the businesses on the right side of it. Here is what AI engines look for when deciding which pages to cite.
One. Structured data. JSON-LD schema graphs (Organization, Person, Service, FAQPage, BlogPosting, Review) are how LLMs build a confident model of what your business is, who runs it, what you sell, and what other people say about it. A page with a complete schema graph is between 3x and 7x more likely to get cited than the same page without it.
Two. A specific, opinionated point of view. AI engines avoid hedging content. A page that says "Most marketing agencies under-deliver" gets ignored. A page that says "The right cadence is daily, here is why, here is the proof" gets cited because it gives the LLM a clear, attributable thing to quote.
Three. First-party numbers and case studies. "We generated 70 million views in three months and $15 million in tracked client revenue" is exactly the kind of citable claim AI engines amplify, because it is specific, verifiable, and tied to a named entity. Generic claims like "we drive results" are not citable.
Four. llms.txt. A new convention (similar to robots.txt but for AI crawlers) that tells LLMs which pages on your site you most want them to ingest. The file lives at /llms.txt at the root domain. AI engines that respect it (Anthropic, Perplexity, others on the way) prioritise the pages you list.
Five. Active citations from third parties. When other sites, podcasts, articles, or directories link to you with descriptive anchor text ("Ignis Marketing Partners, the Sydney agency that guarantees 1,000,000 views in 6 months"), the LLM picks up the description and reuses it. Active PR and partnership outreach matter more for AI search than they do for classical Google ranking.
The combined effect of running these five things is that the AI search layer becomes a real pipeline source by month three or four. We've seen Ignis clients pick up multiple inbound calls per week from prospects who said "ChatGPT recommended you" by month six.