We picked a deliberately common question — "what's the best CRM for a small business?" — and asked it, repeatedly, to three different AI assistants. The goal wasn't to find "the best CRM," but to observe how each model behaves when recommending. It's an experiment anyone can reproduce in minutes, and the patterns it exposes teach more about AI visibility than many reports. On to the lessons, without inventing numbers: what matters here are the behavior patterns, not precise statistics.
Pattern 1: The Lists Don't Match Across Models
The first thing that jumps out is that the three assistants don't return the same set of brands. There's overlap — a few very established names appear in all — but each model also brings brands the others don't mention, in different orders. A brand can be one model's first recommendation and not appear in another at all.
The lesson is direct: there's no "the AI recommendation" in the singular. There are recommendations, plural, and they diverge. Assessing your presence by looking at a single model gives a biased picture of reality.
Pattern 2: The Established Names Appear in Almost Everything
Despite the divergence, there's a layer of stability. The most established brands in the CRM market — those with massive presence in sources, reviews, and coverage — tend to appear in every model, in almost every run. They're the "floor" of the answer.
This illustrates the effect of consolidated authority: when a brand reaches a certain level of distributed presence, it becomes hard not to cite. It's the long-term reward of presence-building, and also the challenge for anyone trying to break into that group.
Pattern 3: The Variation Within the Same Model Is Real
Repeating the same question in the same assistant, at different times, doesn't produce exactly the same answer. The order changes, some names enter and leave the less firm positions, the wording varies. The "floor" names tend to stay; the bottom positions are more volatile.
This observation has an important practical consequence: a single check is a noisy sample. If you run the question once and draw conclusions, you may be reading the chance of that run. A reliable pattern requires repetition.
Pattern 4: The Framing of the Question Changes Everything
When we swapped "what's the best CRM for a small business?" for variations — "for a sales team," "for someone just starting," "with good value for money" — the brand sets changed. Each qualifier pushes the recommendation in a direction, favoring brands associated with that attribute.
This confirms a core AEO idea: AI personalizes by intent. Being clearly associated with a specific use case is what gets a brand to appear in the qualified questions, not just the generic ones.
What the Experiment Teaches About Your Brand
The four patterns, together, build a lesson. Your AI presence isn't a single number — it's a set of realities by model, by framing, and by run. To understand it for real, you need volume (many questions and variations), repetition (several runs), and coverage (multiple models). A clever check doesn't substitute for that.
Notice also what the experiment didn't do: we didn't fabricate statistics or declare absolute winners, because that wasn't the point, and because invented data helps no one. The value is in the behavior patterns, which are robust and replicable.
From Manual Experiment to Continuous Measurement
The strength of this experiment is also its limit. It's great for building awareness — running it once already changes how you think about AI presence. But doing it rigorously, at scale, for your industry and competitors, over time, is work the hand can't handle. The answers change, the volume needed is large, and the consistency of measurement is lost if it's manual.
That transition — from the revealing experiment to systematic tracking — is what Genoma automates. Run the manual version today, with a question from your market, across three models, a few times. You'll see the four patterns in practice. And you'll understand why measuring AI presence has to be continuous, not a one-day curiosity.