Home/Blog/METHODOLOGY
METHODOLOGY

Why age and income tell you nothing about what someone asks an AI

Two groups, the same product category, a nine-point difference in brand presence. A classic age-based persona would never have shown you that gap.

Felix Zeh·August 7, 2026·7 min read·AI-assisted
25%50%75%100%B2C · PRIVATE NEED STATE51.8 %B2B · COMMERCIAL NEED STATE43.0 %
What you'll take from this article
  • Why demographic personas are the wrong unit for AI visibility
  • The three building blocks a need state is made of
  • A measured nine-point gap between two groups in the same category
  • How to build your own need states in half a day
51.8%presence in the private need state (n=83)
43.0%presence in the commercial need state (n=79)
8.8point gap within the same category

Almost every marketing folder has a persona file in it. "Andrew, 42, engineer, married, household net income 75,000 euros, interested in tech and DIY." Fine for campaign planning, maybe. For the question of what Andrew types into a chat window, not so much.

What's wrong with classic personas?

Nobody types their age into ChatGPT. People type their problem. And two people with identical demographics type completely different things depending on the situation they're in at that moment.

The reverse is also true: two demographically very different people type almost identical questions when they share the same problem under the same time pressure. For this question, demographics are simply the wrong axis to sort by.

What is a need state made of?

We work with three building blocks, and all three describe the moment of the question, not the person as a whole.

The need. What problem needs solving, what goal needs reaching? Not "interested in garden equipment", but "wants a battery system that scales across several device classes".

The pressure. Why now and not some other time? Time-related, financial, social or operational. The pressure explains why someone starts researching at all instead of putting it off.

The context. What situation is the question coming out of? On a phone between two appointments, or in the evening at a desk with time to spare. With prior knowledge or without. As the decision-maker, or as someone who still has to convince someone else.

On top of that comes the decision phase, because the same person asks differently during awareness than they do right before a purchase.

What does that look like in a real case?

In the case we analyzed, we studied two need states within the same product category.

Group one, private. People with large properties who've accumulated a mismatched fleet of tools over the years, a different battery for every device. The pressure is seasonal: the winter-pruning window is short, and last year two trees leafed out before the work got finished. The context: time to research, their own budget, deciding alone.

Group two, commercial. Operations managers in landscaping and grounds maintenance, eight to fifteen employees. The pressure comes from the start of the season with a worn-out fleet, an upcoming generational handover, or a big contract just won. The context: the investment has to be justified against order volume and argued for internally.

Both are asking about the same product category. Both groups include people from their mid-thirties to their late fifties. Demographically, you'd barely be able to separate them. By need, pressure and context, they're two completely different worlds, and they phrase things accordingly.

What did the measurement show?

The brand under study showed up in 51.8 percent of answers to the private group, but only 43.0 percent of answers to the commercial group. Almost a nine-point gap on the exact same product category.

And it flipsIn the survey two weeks earlier, the ratio was exactly reversed: 19.4 percent private versus 35.1 percent commercial.

That second finding is the less comfortable one. Which need state a brand serves better isn't apparently a fixed property, it can shift with the model doing the answering. A model switch happened between the two surveys, which is why no trend over time can be drawn from this.

Still, it makes the case for the methodology. An age-based persona wouldn't have had a category in which such a shift could even become visible.

How to build your own need states

  1. Start with triggers, not people

    Collect ten to fifteen concrete situations in which someone starts taking an interest in your category. Sales and customer service can give you this off the top of their heads.

  2. Name the pressure behind each trigger

    Why now? If you can't answer that, it's not a real trigger, just general interest, and you can cross it off.

  3. Group triggers with the same need and pressure

    Fifteen situations usually collapse into three to five groups. You rarely need more, and more rarely gets properly worked anyway.

  4. Write two opening questions per group, in the customer's own words

    The way it would actually be typed, casual language, no brand name. "I've got a different battery for every tool, it's driving me crazy" beats "battery system compatibility garden equipment".

  5. Define the group's goal

    How do you know an AI answer actually helped this person? Without that criterion, you're only measuring mentions, not usefulness.

  6. Add demographics afterward, if at all

    Still useful for media planning. For content and prompts, it's secondary, and shouldn't drive how you form the groups.

Frequently asked questions

Can we keep using our existing personas?

Partly. What you know about triggers, goals and decision paths is valuable and transfers over. What's purely demographic doesn't help with this question. Usually a single workshop is enough to re-sort existing personas along need, pressure and context.

How many need states make sense?

Three to five. In the case we analyzed, two groups were enough to make a clear difference visible. More than five almost never get fully worked in practice, and then you have a nice file instead of an actual effect.

Do I need qualitative interviews for this?

They help a lot, but they're not a must to get started. Sales, customer service and specialty retail know the triggers well. Interviews pay off most when you want to understand the need underneath, meaning why someone actually buys.

What's the difference to Jobs to be Done?

There's a lot of overlap, the core principle is related. We add two dimensions that matter for AI visibility: the pressure that explains the timing, and the decision phase, because the same person asks differently during awareness than right before a purchase.

How this was measured

Need state = need times pressure times context, each tied to the decision phase. Six personas from two need states, each simulated four times independently. Presence values across all answers of the given group and all phases. Survey from July 26, 2026 (gpt-5.6-terra); comparison values from the survey on July 12, 2026 (gpt-4o-mini).

Transparency note · AI-generated content

A note on this article: research, analysis and writing were produced with the support of AI systems (the Ex Tenebris agent team) and reviewed editorially by Felix Zeh. Every figure cited comes from a real LUX/GEO analysis; the analyzed brand is anonymized to protect the client relationship.

FZ
Felix Zeh
Founder, Ex Tenebris · Stuttgart, Germany

Has spent over a decade at the intersection of marketing, data and analytics, and uses LUX to measure how brands show up in the answers of ChatGPT, Claude, Gemini and Perplexity. Questions or pushback on this article? kontakt@extenebris.de

Get in touch

How visible is your brand in the AI answer today?

The methodology behind this article is the same one we use for every assessment we run, tailored to your category, your competitors and your need states.