Verke Editorial

How to tell if AI coaching is working: behavior change signals over feeling signals

Verke Editorial · May 21, 2026

The simplest answer to how to tell if AI coaching is working is to watch your behavior, not your feelings. Things you'd been avoiding start happening. Calls get made. Conversations get had. The thing you dreaded for weeks gets done. The feeling-side — calmer, less self-critical, less rumination — arrives later, sometimes much later, and it arrives quietly. Most people who decide AI coaching "isn't working" are reading the feeling-side too early and missing the behavior-side that's already shifted.

The rest of this article gives concrete signals to actually watch for, the meta-signals that tell you the work is underway even when the obvious signs are quiet, and an honest answer to the "what if nothing is shifting" question — including when to switch coaches, when to step up to human support, and when to let the work breathe through a slow stretch. Change is uneven. Some weeks feel like nothing is happening when underneath something is.

Feelings shift slowly. Behavior shifts faster. That order is counterintuitive — most people assume the feeling-shift arrives first and the behavior-shift follows, like the internal state has to update before the external action changes. It usually goes the other way. You start saying the thing you'd been editing, going to the place you'd been avoiding, ending the call you'd been dragging out — and then, somewhere in the next few weeks, the feeling underneath those actions starts to soften. The behavior is the leading indicator; the feeling is the trailing one.

This delay creates a common trap: the "I should feel different by now" thought, usually around week two or three, that talks people into quitting just before the feeling-side catches up. The mental model of "coaching works when I feel better" is the wrong measurement frame. A more useful frame: coaching is working when the ratio between "things you'd been putting off" and "things you're actually doing" starts shifting — even if you still feel anxious or sad or unsure about the doing.

These are the signals worth weighting most. None of them requires a feeling-shift to count. Each is a discrete, observable change in something you do — which is exactly what makes them reliable:

• Phone calls you'd been avoiding got made — the doctor, the parent, the bill, the difficult colleague.
• Conversations you'd been dreading got had — the boundary you set, the question you asked, the thing you'd been editing.
• The appointment got booked — therapist, dentist, GP, the thing you'd been "meaning to" for months.
• The thing you'd been putting off got started — not finished, just started. Started is the hard part.
• You went to the place you'd been avoiding — the gym, the office, the family event, the social thing.
• You said no when you'd previously have said yes — to the favor, the meeting, the obligation that didn't fit.
• You asked for what you actually wanted — at work, in a relationship, with a friend, instead of hinting and hoping.

Feeling-shifts are real signals — they're just slower and less reliable on their own. Pair them with a behavior-change signal for confidence. These usually arrive in week three or four, sometimes later, and they tend to arrive quietly rather than as a clear before-and-after:

• Less anxious about the thing — usually after the behavior-side has already shifted, not before.
• Less rumination — fewer hours per week spent re-running conversations or future-tripping.
• Easier sleep — falling asleep faster, fewer 3am wake-ups, less middle-of-the-night spiraling.
• More energy — not necessarily physical energy; sometimes just less of the constant background drain.
• Less self-criticism — fewer of the small "I'm so stupid / why did I do that" moments through the day.

A third category — easy to miss, but often the earliest sign something real is happening. These are signals about how your relationship to the work is changing, separate from either behavior-shifts or feeling-shifts:

• Sessions feel more efficient — not necessarily longer, but you arrive faster at what actually matters and waste less time circling around the surface.
• You bring new topics rather than re-litigating old ones — the thing you talked about three weeks ago has settled, and you're working on something else now.
• You notice between sessions — "oh, this is what Judith was talking about" lands in the middle of a Tuesday meeting, on the train home, at the kitchen sink. The work is starting to live outside the chat.
• The vocabulary travels — phrases from session start showing up in your internal narration, the way useful frames do once you've absorbed them.
• You stop checking whether it's working — at some point the meta-question quiets and the work itself becomes the thing.

Different patterns produce different signals. If you came in for overthinking, the signal is fewer hours per week inside the loop and faster recovery when you do drop in — not the loop disappearing entirely. For more on the underlying mechanism, see how to stop overthinking.

If you came in for rumination, the signal is the loops closing on their own faster — what used to take three days to run its course now takes an afternoon. See how to stop ruminating for the underlying frame.

If you came in for the kind of stuckness that benefits from PDT — recurring patterns, self-sabotage, the why-does-this-keep-happening question — the signals are slower and more diffuse. The behavior-shift in PDT-leaning work tends to show up as the pattern catching itself earlier than it used to: you notice you're doing the thing while you're doing it, instead of three weeks later. For more on the modality, see what PDT does. Most readers will also want Inside Verke for what the experience-side of useful coaching actually feels like.

Four weeks is a reasonable review window. By then, if no behavior has shifted and no meta-signals have surfaced — sessions still feel scattered, you're still re-litigating the same opening question every time, the vocabulary isn't traveling — that's data worth treating seriously rather than waiting it out for another month. Three honest moves to consider:

First, name it directly to the coach: "I've been here four weeks and nothing has shifted. What are we missing?" Coaches are explicitly built to receive that line without defending the work. Often the conversation-about-the-stuckness is what unlocks the shift, because it surfaces what you've been editing out of the previous sessions.

Second, switch coaches. The CBT register isn't for everyone, the PDT register isn't for everyone, the ACT register isn't for everyone. If you've been with Judith and the tactical structure isn't landing, try Anna for a slower reflective register; if you've been with Anna and the depth-work isn't finding traction, try Judith for something more concrete. Switching takes ten seconds.

Third, consider stepping up. Coaching is for the part of the work that responds to thinking-out-loud. If what you're carrying needs more — long-running depression, active trauma processing, severe anxiety, substance dependence — the right move is a human clinician, not pushing harder on a coaching tool. AI coaching can sit alongside that work; see how to use AI coaching between therapist sessions for the hybrid framing.

AI coaching is coaching, not clinical care. If you're experiencing severe depression that won't lift, panic attacks interrupting daily life, thoughts of self-harm, active trauma processing, or substance dependence, a licensed clinician is the right next step rather than pushing harder on a coaching tool. You can find low-cost options at opencounseling.com or international helplines via findahelpline.com. Coaches surface these resources directly when the conversation flags severity, and the AI is explicit about not being a crisis line.

Judith's CBT register is built around behavior-based tracking. The session shape — clear question, small experiment, debrief — produces exactly the kind of observable behavior-change signals this article is built around. She'll help you turn vague stuckness into a specific small thing to try, then sit with what actually happened when you tried it. The work compounds across sessions because Judith remembers the experiment you ran last week and what came back from it. For more on the underlying method, see Cognitive Behavioral Therapy (CBT).

Track what's shifting with Judith — no signup, no payment