In January 2026, we registered a new domain. No backlinks, no brand, no indexed pages. Five weeks later, ChatGPT was citing it as a primary source on Indonesian property law — ahead of sites with 550+ citations and years of authority. This is the full story.
The question that started it
We didn't build BaliPropertyRules.com to rank on Google or drive traffic. We built it to answer one question: can a brand-new site, with no existing authority, earn AI citations purely through content quality?
The AI citation space was full of theory. Everyone had opinions about what makes AI engines cite a source. Nobody had run a controlled experiment from zero. So we did.
We chose Bali property law for three reasons. First, the niche is full of misinformation — dozens of English-language sites recycling each other's wrong answers about foreign ownership rules. Second, the primary sources (Indonesian government regulations) exist in Bahasa Indonesia, meaning almost nobody in the English-speaking market was citing them directly. Third, it's a high-stakes topic where accuracy genuinely matters — people making $500K+ property decisions based on whatever AI tells them.
The conditions were deliberately unfavorable. A new domain on shared WordPress hosting. No social media presence. No backlinks. No author credentials in the niche. If the methodology worked here, it would work anywhere.
What we actually did
We didn't start by writing content. We started by understanding the landscape — what AI engines were already telling people about Bali property, and where those answers fell short.
What we found was telling. Most English-language sources cited each other, not the actual Indonesian regulations — a game of telephone, with each site slightly garbling what the previous one said. A major new government regulation (PP 28/2025) had just been issued, and nobody had written a comprehensive English-language breakdown. And when AI engines gave wrong answers about Bali property law, it was because no trustworthy source existed for them to cite.
That third point was the insight. AI engines don't make up answers from nothing — they synthesize from available sources. If the best available source is mediocre, the answer is mediocre. Give them a genuinely better source, and they'll use it.
AI engines don't decide to ignore you. They cite the best available source. The question isn't "how do we trick AI into citing us?" — it's "are we actually the best source?"
Starting with depth, not breadth
We didn't try to cover everything at once. We chose one topic where existing coverage was weakest — PP 28/2025, the new regulation on foreign property rights — and built the definitive English-language explainer.
This article went to the primary source: the official Indonesian government gazette, in Bahasa Indonesia. It addressed the contradictions between what different sources claimed and explained why they disagreed. It was thorough enough that a reader wouldn't need to check anywhere else.
That single article earned 31 citations across AI engines within the first week.
It worked because of something we now see with every client: AI engines trust sources that do the hard work other sources skip. In this case, the hard work was reading Indonesian regulations in their original language. For other niches, the "hard work" looks different — but the principle is the same.
Building from one article to fourteen
One article isn't a strategy. After the first piece landed, we published three more in the first week — visa and property rights, inheritance law, nominee ownership risks. Each one cited primary sources. Each one was designed to reinforce the others.
The result: a body of content where AI engines could verify claims across multiple articles, building trust in the domain as a whole — not just individual pages.
By week two, we had eight articles covering the practical buying process — due diligence, PT PMA company formation, leasehold structures, transaction costs. Ten prompts were now returning BaliPropertyRules.com as a cited source.
By week three, accelerating fast. By the end of the month, 16 published articles, 280 domain citations, and a 16% domain coverage rate — meaning 16% of all tracked prompts in our niche cited our domain, ranking us #4 of 6 established competitors. Against sites that had been building for years.
What surprised us
Two things we didn't expect.
First, speed. We anticipated months of gradual growth. Instead, citations appeared within days of publishing. AI engines can discover and evaluate new content far faster than traditional search engines index it. Being first with accurate coverage of a new regulation or market shift is enormously valuable.
Second, traffic. Within the first week of installing analytics, we saw 5 visitors arriving directly from chatgpt.com. Not from Google, not from social media — from ChatGPT's web browsing feature. People asked ChatGPT a question, it cited our article, and they clicked through. From cities like Aspen, Riyadh, New York, Paris, and Singapore — exactly the international buyer audience a Bali property resource wants to reach.
The numbers in context
280 citations sounds small next to competitors with 550+. But those competitors have years of content and backlink authority. We reached 51% of the top competitor's citation count in four weeks, with a domain that was five weeks old — ranking #4 of 6 in the niche.
More importantly: when we were cited, we were almost always the first source mentioned. Average brand position of 1.02. AI engines weren't just citing us — they were leading with us.
Why this worked — and what it means
The experiment confirmed something that sounds obvious but has real implications: AI engines cite the source that most deserves to be cited.
Not the oldest source. Not the source with the most backlinks. Not the source with the best SEO. The source with the most trustworthy, primary-sourced, well-structured information. Domain age was irrelevant. Brand recognition was irrelevant. What mattered was whether the content earned the citation on merit.
This is fundamentally different from traditional search. Google's algorithm weighs hundreds of signals — domain authority, backlink profile, site speed, keyword density. Most of these are proxies for quality, not quality itself. AI citation engines cut through the proxies. They evaluate the content directly.
That's why we built Cited around this methodology. Not because it's clever — because it's the only approach that holds up when AI engines get better at evaluating sources. Every technical trick eventually gets patched. Content quality is the one signal that compounds.
What we'd do differently
Honesty requires saying what we missed.
We underestimated how differently each AI engine evaluates sources. What earns citations on one platform doesn't automatically transfer to another. Our Perplexity citation rate initially sat at 0% for weeks — it took dedicated freshness and structural work to crack that platform, but we've since earned citations there as well.
We also should have invested in freshness signals earlier. AI engines notice when content is actively maintained versus published and forgotten.
And we should have installed analytics earlier. Five weeks of citation growth without traffic data means we can't fully quantify the click-through impact of those early citations. We know the citations were happening — we just can't prove the traffic they drove before we started measuring.
What we took from this
The experiment confirmed something important: AI citation is earned on merit, and it can be engineered systematically. Not through tricks or technical exploits — through a process that consistently produces content AI engines want to cite.
We've since turned that process into a service. The niche changes, the competitive landscape changes, the specific challenges change. The system we built to solve them doesn't.
If you want to see where your business stands, we'll run the same citation audit on your niche — free.
