Part 2 of 3 — How people choose an AI assistant chatbot for a particular need, and the five distinct segments those choices reveal

In part one we met the everyday chatbot user: tech-comfortable, juggling several assistants, typing more than talking. But averages flatten people. The more interesting question is why someone reaches for one chatbot over another — and there, the population splits apart in revealing ways.

We asked our 1,000 respondents what drives their choice of chatbot, across dimensions like raw response quality, avoiding subscription fees, style and tone, image and document handling, personalization, company ethics, and — crucially — whether the assistant seems to understand emotion. Then we ran a cluster analysis on the answers.

As we’ll show, this a space that evolving into a many-to-many relationship: there are many needs out there, and many ways to address those needs. This is not an ecosystem dominated by a single benchmaxxing parameter: rather, people choose the appropriate AI assistant chatbot for each of their needs.

How people choose, in aggregate

Across the whole sample, the loudest signals are practical: most people choose on quality, on cost, and on the convenience of having one tool that does everything. That's unsurprising. But buried in the data is a smaller, sharper pattern — 12% of people say they always choose a chatbot based on how well it understands emotion. That's not a rounding error. It's a meaningful minority for whom the relationship, not just the output, is the point.

12% always choose their chatbot based on emotional understanding — a small but distinct group for whom the feel of the interaction matters most.

Five segments

The cluster analysis produced five groups. They're roughly the same size, with one notably smaller, and each has a clear personality.

Cluster 1 — Practical Capability Seekers (20%)

These users actively switch tools between work and personal life and choose primarily on capability: response quality, avoiding subscription costs, and media features like interpreting and generating images and documents. They are completely indifferent to emotional connection — the lowest score of any group — as well as to company ethics and corporate policy. Savvy, outcome-focused, treating chatbots as functional instruments.

Cluster 2 — Quality-First Loyalists (21%)

The defining trait is loyalty: they score highest of any group on wanting “one chatbot for everything.” Within that single tool they care deeply about quality and about style and tone, and they'd prefer to avoid paying. Context-switching, integrations, and company policy barely register. They've found their assistant and they're sticking with it.

Cluster 3 — Discerning Multi-Factor Choosers (24%)

The largest and most thoughtful segment. They score highly on every single dimension — something no other cluster does. They split tools across contexts and care about quality, media capability, ethics, company policy, personalization, and integrations all at once. These are power users who have mapped out exactly what they want and demand it across multiple providers.

Cluster 4 — Passive One-Tool Users (23%)

The mirror image of Cluster 3. They want one chatbot for everything and they'd rather not pay — and that's essentially the whole story. Response quality, personalization, and emotional connection all score low. These are casual users who default to whatever free tool is in front of them and barely think about the choice.

Cluster 5 — Emotionally-Led Power Users (12%)

The smallest segment and, arguably, the most demanding. Every person here said they always choose based on emotional understanding. But they aren't soft, low-effort users — they also score highest across the board on quality, style and tone, media interpretation, and personalization. They want an assistant that performs brilliantly and genuinely gets them. Their language about their favorite tools is telling: “Bestie,” “Companion,” “the one that knows me best.”

The segments live different lives

These aren't just attitudinal differences — the groups differ demographically too.

• Age and gender. The Emotionally-Led Power Users are by far the youngest segment (median 46 vs. 56–57 for the two oldest groups) and the most female (63%). The Passive One-Tool Users are the oldest.

• Work. Multi-Factor Choosers lead on full-time employment (61%). Loyalists and Passive users have the most retirees. The Emotional segment has the highest student share (13%), consistent with its younger age.

• Caregiving. Multi-Factor Choosers and the Emotional segment are the most caregiving-intensive — both show high rates of parenting teenagers and living with elders. Passive users are the most likely to have no caregiving duties and to live alone.

• Devices. Multi-Factor Choosers are the most gadget-rich households; Passive users score lowest on every device category. The Emotional segment skews more toward Android than iPhone.

What they want next reveals the most

One of the most telling questions was this:

In the future, chatbots will be able to do lots of things better than now. Pick the five from this list that you think will be most valuable to you as chatbots improve and evolve.

When we looked at the whole thousand-user population of this study, we saw some clear patterns. The most popular choices were the roles that chatbots play well now: an assistant, a teacher, and an editor, each picked by 65%+ of respondents. Surprisingly, perhaps, the next most popular was the chef, suggesting recipes and meals, chosen by 40%. The next set of answers  focused on the sort of emotional support we’ve begun to see as a pattern: mentor, coach, listener, crisis triage, companion, encourager, partner, friend. Between those options, 80% of respondents picked at least one of those, and 55% picked more than one. They’re all choices that emphasize the emotional role that AIs can play: not just rational tools to make us cleverer, but tools that make us better social humans.

We then split up these answers by segmentation. Here the divergence is sharpest. Clusters 1, 2, and 3 overwhelmingly want productive roles — assistant, editor, teacher. The Emotionally-Led Power Users break hard the other way, toward relational roles: listener (41%), best friend (27%), companion (26%), even crisis triage (27%) — every one of these their group's highest score by a wide margin. The Passive users, true to form, lead only on “DJ,” suggesting even their aspirations stay low-effort. (Note that emotionally-oriented findings come from the smallest cluster (around 117 people), so treat the fine detail as directional rather than precise.)

Why segmentation matters

The practical lesson is that there is no single “chatbot user” to design for. A feature that delights the Multi-Factor Choosers — say, deep integrations and document handling — is invisible to the Passive crowd and beside the point for the Emotionally-Led. The clearer your picture of which segment you're serving, the sharper the product you can build. We don't have to solve every problem for everyone; we can specialize and still be useful.

In part three, we leave current users behind and look at the larger prize: the people who haven't started using chatbots yet — and a few portraits of who tomorrow's users might actually be.

Next in the series: The people who haven't started — and the people we're building for.