I built a tool to find out if AI recommends your product. It says no about mine.

by

I run Glotier. It answers one question: when a buyer asks ChatGPT or Gemini or Perplexity for the best tool in your category, do you get named, or does your competitor?

Most of what I could find in this space either costs about $2,000 a month or quietly makes the numbers up. So I built the cheap honest version, and I have been building it in public.

Here is the part that usually gets left out of posts like this.

I ran our own product through it. Across 21 buyer questions in our own category, AI names Glotier in zero of them. That number is on our pricing page. I am not going to sell you a visibility tool while pretending we are visible.

What I can show you is that the reading is real. I opened Cloudflare's crawler analytics on our own domain this week and in 24 hours: 3,210 requests from AI crawlers. ChatGPT-User 134, PerplexityBot 109, OAI-SearchBot 48, ClaudeBot 40, GPTBot 2. The distinction matters more than the total. GPTBot is the training crawler and it is at 2. ChatGPT-User is what fires when a person asks ChatGPT something and it goes and reads a page to answer them. So assistants are not archiving us, they are reading us to answer live questions. That is the traffic nobody has a dashboard for yet.

How it actually works, without the buzzwords:

You paste your product link. It reads the product live, works out the questions your buyers would actually type, and freezes a set of them so week to week is comparable rather than vibes. Then it asks those questions on a schedule across three assistants and records the answers verbatim, including who got named instead of you.

Everything after that is built on that measurement rather than on a content calendar:

  • A writer that publishes one article a day, argued from your own measured data, with a gate that blocks it if it invents a statistic. That gate exists because it caught the agent writing "our data across 1,000+ SaaS sites" under my name. It was not true. Nothing with a fabricated number ships.

  • A Reddit agent that finds the threads where your category is actually being discussed and drafts a comment for each. It reads the last ten comments it wrote first, so it does not repeat itself, and there is a similarity check that blocks a draft that is a reworded version of a previous one.

  • An X agent that reads the last posts in the series, picks the angle it has not used lately, and writes the next one.

  • A source agent that tells you where the answers in your category come from. In ours, Reddit is a source for 57% of them. Directory listings barely register. That single finding changed what we spend time on.

Every draft waits for you to approve it. None of them post on their own. I have been on the other side of an automated posting tool and I am not shipping that.

The honest state of it: we have been cited by ChatGPT six times in a fortnight, and every one of those citations points at a game teardown page rather than at anything a buyer would read. So the engine works and the aim is off, and fixing the aim is what I am doing now.

There is a free check if you want to see your own number before deciding anything:

Happy to answer anything, including the parts that do not work yet.

14 views

Add a comment

Replies

Be the first to comment