I spent a month optimizing two sites for AI citations (GEO) and got zero results — here's the actual diagnosis
I've been running GEO (Generative Engine Optimization) experiments on two sites for about a month. Neither site got cited by ChatGPT or Perplexity once.
Today I finally sat down and diagnosed why.
The answer was embarrassing: Google hadn't indexed a single blog post on either site. 20+ articles each, zero indexed. The GEO chain is: Google indexes → AI crawlers discover → AI cites. I hadn't even cleared the first step.
Root causes:
- No external backlinks at all, so Googlebot had no reason to prioritize crawling
- No IndexNow setup, so Bing (which powers ChatGPT search) was never notified of new posts
- Both sites had FAQPage sections in HTML but no JSON-LD schema markup
- No llms.txt on one of the sites
Fixed today:
- Manually requested indexing via GSC for the 10 best posts across both sites
- Set up Bing Webmaster Tools and verified both domains
- Bing meta tag deployed to both sites
- Confirmed FAQPage JSON-LD was already in the template (missed this one — it was already there)
The GEO playbook I wrote says "expect 4-8 weeks after indexing." The clock hadn't even started.
I wrote up the full methodology at alexsignal.com/blog/how-long-does-it-take-to-get-cited-by-ai-search-engines-after-publishing-2026 — the irony of the domain not being indexed is not lost on me.
Sharing this because I suspect a lot of people doing "GEO" are in the same boat without realizing it.
The useful diagnosis is that GEO work has a dependency chain, and indexing is the first gate. I’d add a small preflight before every content batch: confirm pages are indexed, inspect crawl discovery, and test one query in both Google and Bing before interpreting citation results. Otherwise structured data and llms.txt can become activity without evidence. The 4–8 week clock also makes cohort tracking important: record publish date, index date, first citation date, and query type per page. That separates technical delay from a content or intent problem. Once the pages are indexed, will you compare the ten manually submitted URLs against untouched pages as a control? That would make the diagnosis much stronger.
This hits close to home. I just launched two apps — ***** and ***** realized I'd been so focused on the App Store that I never checked if Google had even indexed my landing pages.
The IndexNow tip is gold. Setting that up today.
One thing I'd add: for app landing pages specifically, I found that having a proper privacy policy and terms page (real URLs, not just in-app) helped with indexing speed — Google seems to treat those as trust signals.
The “the clock hadn’t even started” realization is probably the most useful part here. Easy to spend weeks optimizing the final step of a funnel without checking whether the prerequisite step is actually working.