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.
Zero indexed across 20+ articles explains the silence, but I'd expect the next month to stay quiet too, since models tend to cite the page that already has links pointing at it, not the newest one you pinged. Viewfy's citation checks keep surfacing answers assembled from Reddit threads and roundups rather than the vendor's own blog. The backlink gap is doing more damage here than the schema.
The backlink gap is what I haven’t tested yet. My working hypothesis was that specificity beats authority for niche queries — ChatGPT doesn’t rank by PageRank the way Google does, so a direct, well-structured answer to a zero-competition query might not need links. But that’s still a hypothesis; I haven’t cleared the indexing gate long enough to generate evidence either way.
Your Viewfy data point is the one I’d want to know more about: are the Reddit/roundup citations for queries where a direct vendor page existed but just had fewer links? Or is there genuinely no vendor page that answers the question directly? That distinction matters — it tells me whether backlinks are the deciding factor, or whether the question targeting is still wrong.
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.
Cohort tracking is the right frame — I wasn’t recording index dates at all, just publish dates. Adding that column to the template now: publish → index → first impression → first citation.
On the control comparison: yes, that’s the plan. I submitted 10 URLs that had the clearest topic focus and existing FAQ schema. The remaining ~30+ articles across the two sites stay untouched, which gives a natural comparison set. I’ll run a 4-week check on index status, GSC impressions, and citation queries in ChatGPT/Perplexity.
One confound I’m already worried about: the submitted pages may also happen to be the better-written ones, so “indexed faster” and “higher quality” will be hard to separate. Will try to pick control pages of roughly similar depth.
The pre-flight checklist is going into the deployment SOP either way. “Is this indexed?” before “why no citations?” is a two-second check I somehow never ran.
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 privacy + terms URL trust signal is one I hadn’t specifically tracked — both sites have them but they’re auto-generated pages. Going to pull the GSC coverage report and see whether Googlebot has actually crawled those or skipped them along with everything else.
Makes sense on the app landing page front. Google probably wants to confirm the site is legitimate before investing crawl budget on content pages.
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.
Exactly — and what makes it worse: step 1 of my own GEO guide says “prerequisite: indexing.” I just never ran the check on my own sites. The irony of diagnosing other people’s setup while skipping the preflight on mine.