There's a mistake almost every company makes when it starts worrying about AI presence: it jumps straight to action. Rewrites pages, hires PR, produces content — all before knowing where it's starting from. Six months later, someone asks "did it improve?" and no one can answer, because no one measured the starting point. The baseline is that starting point. Without it, you can't prove progress or prioritize effort.
What a Baseline Answers
An AI-presence baseline is a snapshot, at a defined moment, of how AI talks about your brand. It answers, in numbers and examples, four questions:
In which questions in your industry do you appear — and in which do you vanish? When you appear, with what prominence (first, middle, end)? What does AI say about you — is it correct, is it positive? And who appears in your place — which competitors dominate where you fail?
That set is the ruler against which everything you do later will be measured. It's not a one-off report to present; it's the baseline that gives meaning to all future measurements.
Step 1: Define the Right Set of Questions
The quality of the baseline depends entirely on the questions you choose. And the temptation is to choose wrong in two ways: questions that flatter you (so the result looks good) or questions too vague (that don't reflect real intent).
The right set is the questions a real customer would ask AI before deciding on a purchase in your market — in their language, not your jargon. Include recommendation questions ("what's the best X for Y"), comparisons ("A or B?"), and problem questions ("how to solve Z"). Cover the real range of intents, not just where you're strong.
Step 2: Choose the Models That Matter to You
You don't need to measure every assistant in the world. You need to measure the ones your audience uses. For most businesses, that means the main conversational models and the most relevant AI search interfaces in your market and language.
The point is to be deliberate: measure where your customers actually ask, not where it's easiest to collect data. A baseline in a model your audience doesn't use is a pretty, irrelevant number.
Step 3: Record With Method, Not Impression
Here's the difference between a useful baseline and a collection of screenshots. For each question, in each model, record in a structured way: did you appear (yes/no), in what position, in what tone, and which other brands showed up. Repeat each question a few times, because the randomness of models makes a single run fragile.
The goal is to move from "I thought we appeared" to "I appear in 12 of 40 questions, almost always at the end, and competitor X appears in 31." The precision of the record is what makes the baseline actionable.
Step 4: Turn the Snapshot Into Priorities
A baseline that becomes just a document is a waste. The value is in reading it as an action map. Where do you vanish completely and should be present? That's the visibility target. Where do you appear, but AI gets it wrong? Accuracy target. Where does a competitor dominate questions that are your core? Competitive target.
The baseline turns the diffuse anxiety of "are we doing well in AI?" into a prioritized list of fronts. That alone justifies the effort of building it.
Why the Baseline Has to Be Repeatable
A detail many get wrong: the baseline is only worth it if you can redo it identically. If the question set, the models, and the method change with each measurement, you're not comparing progress — you're comparing different things. The consistency of the measurement is what lets you say, with confidence, "we improved."
That's why it's worth documenting exactly what you measured and how, so the next measurement is a photograph from the same angle. Comparing baselines identical in method is what reveals your progress curve.
From Manual Effort to Continuous Tracking
You can build a first baseline by hand, and it's worth doing to build awareness. But maintenance — repeating with the same rigor, across multiple models, over time, without bias and at scale — is where manual work collapses. The results change, the volume grows, the consistency is lost.
That's the problem Genoma solves from day one: establishing a rigorous baseline of your AI presence and turning it into continuous tracking, so you never again have to answer "did it improve?" with a shrug. Measure before you move. It's the step that separates strategy from cheerleading.