GEO Score Watch — Week 0: Our Entire Portfolio Is Invisible to AI

We build a tool that measures whether AI assistants recommend your brand when buyers ask. We're using it on ourselves, in public, every week, until the numbers move.

This is Week 0 — the baseline. It's bad. Here's all of it.

The scoreboard

Four products. Four zeros. When a buyer asks ChatGPT for the best tool in any category we compete in, not one of our names comes up — the assistant recommends someone else, every time.

What the tool actually does (the method, so you can trust the number)

No vanity metrics. For each product, the tool asks ChatGPT — with web search on — the real buying-intent questions a customer would ask, then measures three things: are you cited, how prominently, and who the assistant names instead. A 100 means you're the answer. A 0 means the conversation happens and you're not in it.

The engine is parity-checked: run it on an established product like Linear and it scores 100. So our four zeros aren't a broken gauge. They're real.

The uncomfortable thing the data showed us

Here's what stung, and what turned out to be the most useful finding of the week.

We logged the exact questions the tool generated for each product. They were nearly identical — and all generic:

• "What are the top AI platforms for developing custom machine learning models?"
• "Which AI tool is best for natural language processing tasks?"
• "What AI platform is best for data analysis?"

The tool generated those same vague questions for Forge (an AI delivery platform), for Dx3 (an intelligence platform), for Faultline (output-integrity). It couldn't tell our products apart. From our brand names and homepages alone, the model read all four as one blurry thing: "generic AI platform."

That reframes the whole problem. We're not losing our category. We're being measured in a category we're not in — because our own pages never told the model clearly what we are. You can't win "best AI delivery platform" if the assistant doesn't believe you're an AI delivery platform. It thinks you're a machine-learning toolkit, and ranks you against Hugging Face and TechRadar listicles.

What we're changing this week

The cheapest lever turns out to be the most precise one. The only thing the tool reads before it decides our category is four signals on each page: the title tag, the meta description, the first sentence of the headline, and the structured data. Fix those, and we change the question the model asks before we ever touch the score.

So Week 0's work is category clarity, per product:

• Forge → "an AI delivery and agent-orchestration platform" (not a task manager).
• Dx3 → "a universal intelligence platform — vector, graph, and memory" (not a CRM add-on).
• Faultline → "AI output-integrity and compliance monitoring" (not a generic dev toolkit).
• NextGen AI → the studio behind the three, pointing at each product by name.

The leading indicator we're watching (it's not the score)

Honest expectation: the score won't jump next week. AI visibility moves over weeks — it rides content getting indexed and trusted sources citing you.

So the first sign it's working isn't a higher number. It's the questions changing. When next week's scan stops asking Dx3 about "small-business CRM" and starts asking about "intelligence and RAG platforms," we'll know the model finally understands what Dx3 is. That shift comes before the score moves. We'll report it here either way — including the weeks nothing changes.

Run the same scan

The tool that handed us four zeros is the one you can point at your own brand. Enter a domain and a category; see your score, the competitors AI names instead, and where you go invisible.

It might tell you you're the answer. It might tell you you're a zero, like we were. Either way you'll know your number — which is more than most brands competing in AI search can say right now.

Run the same free scan on your brand →

Free, no signup — takes under a minute.

GEO Score Watch publishes every week. Week 0 baseline: 0, 0, 0, 0. We'll show you the climb — and the plateaus — in public. Next update: Week 1.