Case study · Building in public

66 days, 6 AI citations, and what actually moved

By Oğuzhan Tüsen @oguzhanglotier·Published 6 June 2026·Figures last checked 28 July 2026

We build Glotier to get products named in AI answers, so we run the playbook on ourselves in public and publish the real figures. Just over two months in, the honest scoreboard is small but moving: two different assistants have now cited us, ChatGPT five times and Perplexity once, and every one of those citations sent a real visit. Google has been generating an AI Overview for us since week two, and a three-day SEO pass moved traffic more than anything else we tried. Here is what did it and what did not.

66

days since the domain went live

2

assistants that have cited us

+64%

visits in the 3 days after an SEO pass

Six citations from two assistants, and why the second one matters more

In 66 days, ChatGPT has named Glotier in an answer five times without anyone prompting it, and on 28 July 2026 Perplexity did it too. Both show the same way in Cloudflare Web Analytics with bots excluded: a referral visit from chatgpt.com and one from www.perplexity.ai. That is the strictest evidence a citation leaves behind, because a person can only click a link an assistant actually put in front of them. Here is the screen recording of that referrer showing up, so you do not have to take our word for it.

The second assistant is the more interesting number even though it is smaller. One assistant citing you five times can be a quirk of how that one model retrieves. Two independent systems, reading the web differently and on different days, cannot be. It is the difference between a fluke and a pattern, and it is the first evidence we have that the pattern exists.

That is the whole claim, and we are not going to dress it up as “ChatGPT recommends Glotier”. There is a real order underneath it: being cited, an assistant reading and pulling from your page, comes before being named in a “best tool” shortlist, and the first is how the second eventually happens. We are measurably past that first line. Six ChatGPT citations in 66 days, from a standing start, is the step most products never see happen and almost none of their makers can even tell you about.

We publish it because six is a number you can compare yourself against today. Most products in this position have never been cited once, and almost none of their makers know either way. Going from unmeasured to measured, and from unread to read, is the actual first step, and it is the step nobody sells you.

Every one of those citations sent someone to the wrong page

This is the part that changed how we spend our time, and it only appeared when we cross-checked where the referred visits actually landed. Over the last seven days, every visit an AI assistant sent us landed on a mobile game teardown. Not one reached the pricing page, the product, or anything a buyer would want. Search engines did the same thing at scale: of 97 visits from search, 86 landed on game or app pages and 3 on a page that sells anything.

The cause is not mysterious. Around nine in ten pages on this site are app and game teardowns, each built from a few hundred scraped reviews. That is genuinely original data, which is exactly why assistants read those pages and quote them. It also teaches every one of them that this domain is a mobile game review site, so the traffic they send is people looking for mobile games.

The machine works. We pointed it at the wrong category. If you take one thing from this page, take that: original data earns citations, and citations for the wrong subject earn you nothing.

Three days of SEO work moved more than two months of waiting

The most useful thing we learned this month had nothing to do with AI. After a focused pass on titles, internal links and page structure, the following three days ran 133 visits and 466 page views, up 64% and 85% on the previous period with bots excluded. Around 35 of those visits came from outside Turkey, which for a product sold in English is the segment that actually matters.

That is the part worth copying. The engines that answer “what should I use?” read the same web that Google crawls. Work that makes a page legible to a crawler makes it quotable to a model, and it pays out in ordinary search traffic while you wait for the citation.

What we actually did

The AI Overview, and the honest version of “we rank #1”

Two weeks after the domain went live, Google had indexed it, was ranking it, and had started generating an AI Overview summarising what we do. Ranking #1 for your own brand name is not the flex it sounds like, since any indexed site wins its own unique name. The two things that are signals: how fast a brand-new domain got crawled and trusted, and that Google chose to write a summary of us at all. That only happens when your content is clear and structured enough for a model to understand and repeat.

The real test is competitive, non-brand terms, “AI visibility tool” and “ChatGPT brand monitoring”, where we are still climbing and mostly still absent. If you want the strategy rather than the scoreboard, start with Generative Engine Optimization, or the practical version in how to get mentioned in AI search.