5 Lessons from Hal9's 1,600% AI Traffic Growth

When Hal9 — an AI development platform that helps companies build and deploy AI-powered solutions — first came to Optimly, they had a familiar problem. Their product was strong. Their customers loved them. But when prospects asked ChatGPT, Claude, or Gemini for "AI development platforms" or "tools for building AI applications," Hal9 wasn't in the conversation.

This is the new invisibility problem for B2B companies. You can have the best product in your category and still not exist in the places where your buyers are increasingly doing their research.

Over 60 days, Hal9's Brand Authority Index went from 2.0 to 8.1 out of 10. AI-referred traffic grew 1,600%. And AI channels became their #1 source of qualified leads — converting at 4x the rate of traditional Google traffic.

Here are the five lessons behind that result, applicable to any B2B company trying to grow from AI referrals.

Lesson 1: The problem isn't that AI can't find you — it's that AI doesn't know you.

Most companies approach AI visibility like they approach SEO: create content, optimize for keywords, hope to get cited. This is what we call the "Eyes" approach — you're trying to get AI to find your content through its retrieval system (RAG/web search).

But AI models have two types of memory. There's Retrieved Information — what the model pulls from the web in real time. And there's Foundational Knowledge — what the model has internalized from training data. The first is temporary and viewed with skepticism. The second is the basis for trust and recommendations.

Hal9's content was findable. But the model didn't know Hal9 at a foundational level. So when a buyer asked "what are the best AI development platforms?", the model reached for the brands it trusted — not the ones it found.

The lesson: Before investing in GEO or AI-optimized content, audit what AI actually believes about your company. Not what it can find — what it knows. If the model's internal understanding is wrong or absent, no amount of content optimization will make it recommend you.

Lesson 2: Wrong categorization costs more than invisibility.

Hal9's initial audit revealed something worse than being invisible — they were being miscategorized. One model described them as "a data visualization tool." Another placed them in the "open-source AI framework" category. A third confused them with a completely different company.

This matters because AI models are risk-averse when making recommendations. They default to the brands they can describe with confidence. When a model is uncertain about what a company does, it doesn't recommend it — even if the company technically appears in a search result.

Our data across hundreds of B2B brand audits shows that roughly 60% of AI misrepresentation is categorization errors — not missing information, but wrong information. The model "knows" you, but what it knows is incorrect.

The lesson: Run a categorization audit across ChatGPT, Claude, Gemini, and Perplexity before doing anything else. Ask each model: "What does [your company] do? What category are they in? Who are their competitors?" If the answers don't match your actual positioning, that's your first problem to solve — and it's a different problem than content optimization.

Lesson 3: You need a single source of truth before you can fix anything.

One of Hal9's early mistakes (common among companies we work with) was trying to fix AI misrepresentation by publishing more content. Blog posts, guest articles, LinkedIn updates — all saying slightly different things about who they are and what they do.

This made the problem worse. Each piece of content with slightly different positioning gave the model conflicting signals. Is Hal9 an "AI development platform" or an "AI-powered solutions company" or an "autonomous AI organization"?

The fix was counterintuitive: before creating any new content, Hal9 needed to define — in precise, structured terms — exactly what their brand ground truth was. What they are. What they aren't. What category they belong in. Who they serve. How they're different.

This is what we call the Brand Vault: a canonical system of record that defines how AI should understand your brand. Every piece of subsequent content is derived from this single source, creating consistent signals that reinforce a single, accurate entity representation.

The lesson: Don't try to fix AI misrepresentation by producing more content. You'll amplify the inconsistency. Start by writing a structured brand definition — category, ICP, differentiators, competitive positioning — and make sure every page on your site, every piece of content, and every third-party mention aligns with it.

Lesson 4: Find and fix the sources that are poisoning the model.

Once Hal9 had their ground truth defined, the next step wasn't creating new content. It was identifying which existing sources were teaching AI the wrong things.

AI models learn from specific sources during training. When a model incorrectly categorizes your brand, there's usually a traceable reason — an outdated Crunchbase profile, a competitor comparison page that frames you wrong, an old TechCrunch article from before your pivot, a G2 listing with the wrong category tags.

For Hal9, we identified several poisoned sources: a Product Hunt listing from an early version of their product, a couple of blog posts from tech publications that described them using outdated language, and several directory sites that categorized them incorrectly.

The fix wasn't just publishing correct information alongside the incorrect stuff. It was going to those specific sources and updating, correcting, or replacing them. This is causal source remediation — tracing the misrepresentation back to its source and fixing it there.

The lesson: AI misrepresentation has specific, identifiable causes. Don't just bury bad information under new content — find the sources that are teaching AI the wrong things and fix them directly. This is faster, more targeted, and produces more durable results than a general content strategy.

Lesson 5: AI-referred leads are fundamentally different (and better).

The most surprising result from Hal9's transformation wasn't the traffic growth — it was the lead quality. AI-referred leads converted at 4x the rate of leads from traditional Google search.

This makes sense when you think about it. When a buyer finds you through Google, they're comparison shopping. They clicked one of ten blue links. They might be checking you out alongside five competitors.

When a buyer finds you through an AI recommendation, the dynamic is different. The AI has essentially pre-qualified and endorsed you. The buyer arrives with higher trust, better understanding of what you do, and clearer intent. It's the difference between "I found you in a list" and "an advisor recommended you."

This is why the Brand Authority Index metric that matters most isn't just Answer Presence (whether you show up) — it's Message Pull-Through (whether AI accurately describes your value proposition when it mentions you). Getting mentioned matters. Getting described accurately when you're mentioned is what drives pipeline quality.

The lesson: Don't just measure whether AI mentions you. Measure what AI says about you when it does. The accuracy of AI's description of your brand directly correlates with the quality of leads that come through AI channels.

What to do this week

These five lessons compound. You can't fix poisoned sources (Lesson 4) if you don't have a ground truth to measure against (Lesson 3). You can't define ground truth if you don't know what's wrong (Lesson 2). And none of it matters if you're only thinking about the retrieval layer (Lesson 1).

Here's the starting sequence for any B2B company:

Step 1: Audit what AI believes about you — not what it can find. Run your brand through ChatGPT, Claude, Gemini, and Perplexity with buyer-intent questions (not branded queries). Our Brand Strategist tool does this in two minutes for free.

Step 2: Identify categorization errors. Is AI putting you in the wrong bucket? Describing you to the wrong audience? Confusing you with a competitor?

Step 3: Define your brand ground truth in structured, machine-readable format. Category, ICP, value prop, differentiators, competitive positioning. One document, no ambiguity.

Step 4: Trace misrepresentation back to specific sources and fix them at the origin.

Step 5: Monitor. AI models update, new content shifts the model's beliefs, competitors enter your space. This isn't a one-time fix — it's ongoing infrastructure.