The Most Common AI Mistakes Therapists and Counsellors Make

Tools like Heliia and Nabla generate match scores based on symptom similarity and availability, which can feel objective and safe. You risk pairing clients with therapists whose style or capacity misaligns with what the client actually needs to feel heard.

The fix

Use the algorithm output as a starting list only, then trust your own knowledge of each therapist's presence style and current capacity to stay attuned.

When ChatGPT or Eleos highlights a pattern in your notes (avoidance of emotion, recurring defence mechanism), it is easy to accept it as confirmation of something you half-sensed. You may then close off curious listening and start confirming the pattern instead.

The fix

When an AI note flags a pattern, write down what you actually observed in the session that made you curious, separate from what the AI suggested.

Eleos AI generates notes as you work, which can split your attention between what the client is saying and what the AI is capturing. Your clinical ear becomes trained to notice things that are easy to transcribe rather than things that matter most.

The fix

Keep notes off during the session itself; spend five minutes immediately after writing what you need to remember, before you check what Eleos captured.

The time saved by using Woebot for between-session check-ins or Eleos for documentation can feel like a clinical win, but it can also reduce the contact hours that build relational depth. You may over-automate what requires human presence.

The fix

Count how many client interactions are now AI-mediated rather than therapist-to-client; if it is more than ten percent of total contact, audit whether those interactions still serve the relationship.

Automated between-session tools can miss the vocal tone, silence, or embodied shift that tells you a client is no longer safe. You rely on the alert system and miss the clinical responsibility that lives in direct listening.

The fix

Use AI check-ins as addition to your direct contact schedule, not replacement; if a client misses a session, reach out yourself rather than waiting for the algorithm to flag it.

When you know Eleos will parse your words into structured fields, you start speaking in patterns it understands. Your notes become optimised for the software rather than capturing what was actually present in the room.

The fix

Write one private note in your own language immediately after the session; only then check what Eleos generated and edit it back toward what actually happened.

Eleos can produce session summaries that look complete and clinical, which creates a false sense that documentation is done. You skip reading it because you were there and the AI was listening, so it must be right.

The fix

Read every AI-generated summary before it goes into the client record; mark what is missing or wrong and make the correction yourself.

When Eleos has already captured the session content in structured form, it can feel like you have already processed the work. You tell yourself you will discuss it with supervision, but the case stays in the file instead.

The fix

Bring at least one case per month to supervision that you documented with AI; discuss what the software captured and what it missed about your own clinical process.

You are trained to document carefully, but Eleos and ChatGPT are not therapists bound by the same ethics. If you note a client's self-harm plan or abuse disclosure directly into the AI tool, you have handed a clinical secret to a third party.

The fix

Keep a separate handwritten note for details that could harm the client if exposed; use the AI tool only for process notes and emotional themes that are already semi-public.

The pull to verify that the AI captured a key moment happens mid-sentence. Your attention flickers away from the client's face and voice to the screen, and the client feels it.

The fix

Place the recording device or note-taking tool outside your line of sight; you will hear if something important was missed and can note it after the session ends.

Because Eleos or Nabla can produce a transcript and clinical summary, you may unconsciously reduce the intensity of your own attention. The thought that the AI is listening can ease the weight of holding the client's story in your own mind.

The fix

Before each session, acknowledge to yourself that you are the one who must be present; the AI is insurance, not substitute.

When you paste a client's language into ChatGPT to understand what they meant, you lose the chance to sit with the ambiguity and ask them directly. The interpretation feels quicker but misses the relational work of discovery.

The fix

When you are curious about what a client meant, bring the question back to them in the next session instead of seeking explanation from an AI.

Woebot and other platforms can deliver psychoeducation and mood tracking, but they do so without knowing the client. The efficiency can make you question whether your own attentiveness adds real value, which erodes your confidence in clinical judgment.

The fix

Identify one thing you do in sessions that an AI tool cannot do; name it to yourself as your actual clinical gift, separate from what can be automated.