I built Citegram because I couldn’t answer a simple question: when a buyer asks ChatGPT which tool to use, does it name us? Not in theory, not once, but across the assistants people actually use, week after week, against the competitors we lose deals to. Nothing I tried gave me a straight answer, so I started building one.
This is the story of that question, what I learned while trying to answer it, and the choices behind the product.
The moment the shortlist moved
The trigger was mundane. I asked ChatGPT the kind of question a buyer asks before they ever visit a website: which tool should a small B2B team use for a given job. The answer was good. It was confident, well structured, and it ended with three names and a short reason for each.
What struck me wasn’t the answer. It was the format. A results page gives you ten links and lets you decide. An AI answer makes the decision for you, or at least narrows it to a handful. If you’re one of the three names, you’re in the conversation. If you’re not, you were never considered, and you will never know it happened.
That’s the part that kept bothering me. In search, you can see your rankings. You can see the clicks you didn’t get. In AI answers, a lost shortlist leaves no trace in your analytics. The buyer just shows up with a decision already half made, or doesn’t show up at all.
Why “just check it” doesn’t work
My first instinct was the one everyone has: ask the assistants myself and write down what they say. It works for about a day.
- Answers change. Ask the same question twice and the wording, the order and sometimes the names change. One check tells you almost nothing.
- Assistants disagree. ChatGPT, Perplexity, Gemini, Claude and Google’s AI Overviews don’t read the same sources or weigh them the same way. Being recommended in one says little about the others.
- The question matters more than you think. “Best CRM” and “best CRM for a 20-person startup with a small sales team” are different prompts with different answers. Buyers ask the specific version.
- A spreadsheet doesn’t scale. Twenty prompts across five assistants, run every week, is hundreds of answers to copy, read and compare. Nobody keeps that up.
The useful question isn’t “are we mentioned?” It’s “how often, where, against whom, and because of which sources?” That needs repeated, structured measurement. In other words, it needs a tool.
Why not the API
The obvious way to build that tool is to call each assistant’s API and parse the answers. I looked at it seriously and decided against it, for one reason: the API doesn’t answer the way the app does.
When a buyer uses ChatGPT or Perplexity, the app can search the web, read pages and cite them. The answer they see depends on that search. An API call made without the same setup is a different experience, with different inputs. Measuring it would tell you how a model behaves in a lab, not what your buyer reads. We cover the mechanics in our guides on how AI assistants choose their sources and on query fan-out.
So Citegram does it the other way around. It’s a Chrome extension that asks your questions in the real apps, in background tabs, with the accounts already signed in to your browser. No API key, no setup, and the answers are the ones a person actually gets, including the sources each assistant cites and, for ChatGPT, the searches it ran to build the answer.
That choice has a trade-off, and I’d rather say it plainly: answers come from your own accounts, so they reflect a real but personal session, and they count toward your own assistant usage. For the question I was trying to answer, real answers beat clean but artificial ones.
What Citegram measures today
Here is what the product does, without the marketing layer:
- You add the prompts your buyers ask and the competitors you care about.
- Citegram asks them in ChatGPT, Gemini, Perplexity, Claude, DeepSeek, Grok and Google AI Overviews.
- For each answer it records whether your brand is mentioned, its position, and the links the assistant cited. For ChatGPT it also records the web searches it ran. For Google it records whether an AI Overview appeared and which sources it used.
- The dashboard turns that into numbers: share of voice per brand and per assistant, the most cited domains, and trends across runs.
The first results come in about ten minutes. That speed matters to me, because the value of this data is in comparing runs over time, and people only come back to tools that answer quickly.
What it doesn’t do
I’d rather lose a sale than oversell, so:
- It doesn’t make AI recommend you. It shows you where you stand and which sources shape the answer. The work of earning those mentions, on your site and on the review sites, comparisons and forums assistants read, is still yours. Our guide to getting cited by AI is where I’d start.
- It doesn’t cover every assistant. Google AI Mode and Microsoft Copilot aren’t tracked today.
- It’s not an enterprise suite. If you need managed data collection at scale, SSO and a dedicated account team, there are good tools for that, and we compare them honestly in our roundup of AI visibility tools.
Why it’s priced the way it is
Most tools in this space are built for large brands, and priced like it. I wanted the question “are we in the answer?” to be something a founder or a two-person marketing team could answer the same afternoon, without a sales call.
So the free plan tracks one prompt across all seven assistants, forever. Pro is $10 per seat per month with unlimited prompts and competitors. Running the questions from your own browser is what makes that price possible: there’s no fleet of servers asking questions on your behalf.
What’s next
AI answers are becoming the first page buyers read, and the way they’re built will keep changing. My job is to keep Citegram honest about what it measures, fast to answer, and cheap enough that measuring AI visibility becomes a habit, not a project.
If you’re working on this problem too, or you think I’ve got something wrong, I’d like to hear it. You can reach me on LinkedIn. And if you want to see where your brand stands, the free plan is the fastest way to find out. See pricing.
ChatGPT
