How a small Norwegian news site got cited by ChatGPT, Grok & Perplexity — in weeks
Droneavisa.no, a Norwegian drone-news site launched in spring 2026 with no backlink profile and no brand recognition, earned citations in ChatGPT, Grok, Perplexity, Google AI Overviews and Copilot within weeks — by getting crawlability, server-rendered structured data and visible trust signals right. No llms.txt. Here's the exact playbook, with numbers.
The receipts — AI citations, measured in Ahrefs (June 4, 2026)
| Answer engine | Citations |
|---|---|
| ChatGPT | 5 |
| Grok | 4 |
| Perplexity | 3 |
| Google AI Overviews | 1 |
| Microsoft Copilot | 1 |
Source: Ahrefs AI citations index, droneavisa.no overview. Site had no llms.txt at time of measurement.
[01]The context: a brand-new site with zero authority
Droneavisa is a small editorial operation: daily Norwegian-language news about drones and aviation technology, one named author, a fresh domain. No PR agency, no link-building budget, no existing audience. If AEO only worked for big brands, this site would be invisible.
It wasn't. Within weeks of getting the technical foundation right, real prompts in five different answer engines were returning Droneavisa as a cited source.
[02]What was actually done
Access — let the AI in. robots.txt explicitly allows GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and friends. Sitemap index plus a Google News sitemap for fresh stories. Active RSS feed. Nothing blocked, nothing hidden.
Parseability — hand AI clean facts. All JSON-LD is server-rendered in the raw HTML (many AI fetchers never execute JavaScript — check view-source, not DevTools). News stories use NewsArticle schema, the rules-guide page carries a 10-question FAQPage, and Speakable points at the visible headline and lead paragraph.
Trust — look like a source worth quoting. One clean Organization entity with sameAs links. A real author with Person schema, visible bylines and timestamps in the HTML body. Source attribution written into the article text ("Ifølge Avinor…") — because that's literally the citation format LLMs reproduce. A deep about-page with editorial principles, a sources policy and a corrections policy.
[03]What was deliberately skipped
- llms.txt — zero citations were lost by not having it. The site earned all of the above without one.
- Fabricated ratings/AggregateRating — manual-action risk, and engines don't reward fake numbers.
- Crawl-budget micro-tuning — irrelevant for small sites; effort went to content and structure instead.
[04]The lesson that surprised us
When we ran third-party AI audits on the site, two out of three falsely reported that schema was missing — because many AI fetchers strip <script> tags and never see JSON-LD at all. The fix: duplicate every signal in visible text. Bylines, dates, source attribution, fact-dense lead paragraphs. Schema for the engines that read it; visible text for the ones that don't.
That finding is now Part 3.5 of the free checklist below — you won't find it in the generic AEO guides.
The AI Visibility Checklist covers every step above — crawlability, schema, the quotable line, and how to measure.
Get the free checklist →
Notes & references
[a]How long does it take to get cited by ChatGPT?
Faster than most expect. Droneavisa recorded citations across five engines within weeks of getting the foundation right. Indexing and trust compound over months, but the first citations can arrive early when the basics are in place.
[b]Do you need llms.txt?
No. This site earned all its citations without one. The measurable levers were crawlability, server-rendered structured data, and visible trust signals.
[c]Does this work outside English?
Yes — Droneavisa publishes in Norwegian. Answer engines cite well-structured sources in any language when the question matches.
← Back to iamcited.com · Case study № 001: built with AI · Get the free AI Visibility Checklist