AI Is Shoring Up Cognitive Errors Made By Mental Health Therapists

AI And Mental Health

As a quick background, I’ve been extensively covering and analyzing a myriad of facets regarding the advent of modern-era AI that produces mental health advice and performs AI-driven therapy. This rising use of AI has principally been spurred by the evolving advances and widespread adoption of generative AI. For an extensive listing of my well-over one hundred analyses and postings, see the link here and the link here.

There is little doubt that this is a rapidly developing field and that there are tremendous upsides to be had, but at the same time, regrettably, hidden risks and outright gotchas come into these endeavors, too. I frequently speak up about these pressing matters, including in an appearance on an episode of CBS’s 60 Minutes, see the link here.

Background On AI For Mental Health

I’d like to set the stage on how generative AI and large language models (LLMs) are typically used in an ad hoc way for mental health guidance. Millions upon millions of people are using generative AI as their ongoing advisor on mental health considerations (note that ChatGPT alone has over 1 billion weekly active users, a notable proportion of which dip into mental health aspects, see my analysis at the link here). The top-ranked use of contemporary generative AI and LLMs is to consult with the AI on mental health facets; see my coverage at the link here.

This popular usage makes abundant sense. You can access most of the major generative AI systems for nearly free or at a super low cost, doing so anywhere and at any time. Thus, if you have any mental health qualms that you want to chat about, all you need to do is log in to AI and proceed forthwith on a 24/7 basis.

There are significant worries that AI can readily go off the rails or otherwise dispense unsuitable or even egregiously inappropriate mental health advice. Banner headlines in August of this year accompanied the lawsuit filed against OpenAI for their lack of AI safeguards when it came to providing cognitive advisement.

Despite claims by AI makers that they are gradually instituting AI safeguards, there are still a lot of downside risks of the AI doing untoward acts, such as insidiously helping users in co-creating delusions that can lead to self-harm. For my follow-on analysis of details about the OpenAI lawsuit and how AI can foster delusional thinking in humans, see my analysis at the link here. As noted, I have been earnestly predicting that eventually all of the major AI makers will be taken to the woodshed for their paucity of robust AI safeguards.

Today’s generic LLMs, such as ChatGPT, Claude, Gemini, Grok, and others, are not at all akin to the robust capabilities of human therapists. Meanwhile, specialized LLMs are being built to presumably attain similar qualities, but they are still primarily in the development and testing stages. See my coverage at the link here.

Therapists And The Role Of AI

I have been extensively identifying and examining the myriad ways that AI enters into the role of professional therapists.

Some therapists refuse to think about AI and want nothing to do with it. Others are embracing AI and using AI as part of their therapeutic process with clients. Indeed, I have predicted that the therapy realm is being transformed from the traditional dyad of therapist-client and inevitably becoming a new triad of therapist-AI-client, see my analysis at the link here.

My view is that whether therapists are keen on AI is not the headspace they should be in. AI is coming, and to a great degree, it is already here. Clients nowadays come in the door with AI-generated advice and want their therapist to tell them what it means. In other instances, clients will post-session try to double-check what their therapist told them and lean into AI as a means of judging the mental health advice they are getting from the clinician. AI is a reality that therapists must face, regardless of their desire to do so. Having one’s head in the sand is not prudent, as I will be illuminating momentarily.

There are lots more variations of the role of AI in therapy and regarding therapists, including these circumstances that I have judiciously addressed:

• How therapists should clinically analyze AI chats of their clients, see my discussion at the link here.
• Questions that clients are asking their prospective or existing therapists about AI, and the answers that therapists ought to be providing, see my coverage at the link here.
• Therapy is shifting from the classic dyad of therapist-client to the new triad of therapist-AI-client, see my discussion at the link here.
• Therapists are being asked by clients to jointly use AI during their mental health therapeutic process and work in these new ways, see my explanation at the link here.
• Some therapists are opting to use AI during therapy sessions with their clients and do so in these astute ways; see my coverage at the link here.
• How therapists are handling clients who appear to be encountering AI psychosis, see my discussion at the link here.
• Therapists are using AI to craft digital twins of their clients and perform more impactful therapy accordingly, see my coverage at the link here.
• Worries that therapists leaning into AI as an aid in conducting therapy might end up deskilling their own capabilities, see my assessment at the link here.
• How therapists are using custom prompts to get generative AI to serve as an adjunct to their therapy sessions and interact with their clients, see my discussion at the link here.
• Public perception of therapists who decide to use AI in their practices, see my analysis at the link here.
• Legal defense strategies being used by AI makers to defend against AI mental health lawsuits, see my analysis at the link here.
• Contending with clients that come to therapy with AI-generated mental health advice and want their therapist to give a thumbs up, see my coverage at the link here.
• Emerging new informal duty might be for therapists to inform their clients about the ups and downs of using AI for mental health guidance, see my analysis at the link here.

And so on.

The Role Of Cognitive Errors

Shifting gears, I’d like to dive into the nature of cognitive errors that human therapists might make. After doing so, we can look into the use of AI to aid in detecting and correcting those errors.

First, let’s consider that cognitive errors can occur at any of these three stages:

• (1) Pre-Session cognitive errors
• (2) Mid-Session cognitive errors
• (3) Post-Session cognitive errors

Before a session, a therapist might make a therapy-oriented cognitive error while preparing to meet with the client. I am emphasizing that these are therapeutic mistakes and not other types of cognitive mistakes, such as those that are administrative. This isn’t about mistakes in billing or paperwork. The focus is on cognitive errors intertwined with therapy.

An example of a pre-session cognitive error would be the therapist anchoring on a pre-fixed diagnosis that they intend to carry into the session. They become resolute that no matter what else happens, they are going to dogmatically insist on the pre-determined diagnosis. During a session, everything the client says will be interpreted solely within that frame of mind. The client has a near-zero chance of being understood in any other manner, and the therapist is going to confirm the preconceived diagnosis.

If a therapist does allow their thinking to be unyielding in this manner during a session, you could say that the pre-session cognitive error has been compounded. The therapist has committed a cognitive error at the pre-session stage and then committed a second cognitive error during the actual session.

Cognitive errors can arise spontaneously during a session. One of the most frequent cognitive errors that a therapist makes is to undertake a semblance of mind-reading. Here’s how that goes. The client says something, perhaps an innocuous statement. The therapist leaps on that statement and assumes all sorts of facts about how the client feels or what they mean. It is a case of mind-reading. A therapist should avoid that type of behavior and aim to ensure that they clarify what the client has stated. Do not jump to premature conclusions.

A cognitive error might arise post-session. A common cognitive error happens when a therapist is preparing their notes about a session. Sometimes, a therapist wants to put a tidy bow on what occurred. They force the discussion to fit a particular narrative. This is known as a narrative fallacy. The issue is that the bias of the therapist is reducing uncertainty and ambiguity into a cohesiveness that doesn’t truly belong there.

Research On Therapist Cognitive Errors

In an excellent article on therapists and cognitive error, namely a posting entitled “Special Report: Addressing Cognitive Error in Psychiatric Practice” by Dr. H. Paul Putman III, Psychiatric News, December 22, 2025, these salient points were made (excerpts):

• “Psychiatric practice has become increasingly complex due to an expanding number of treatments, longer patient life spans, and a resulting increase in comorbid psychiatric and nonpsychiatric medical conditions.”
• “In facing these challenges, awareness of how we make frequent human cognitive mistakes can help elevate the quality of our efforts, reduce treatment failure, and minimize suboptimal results.”
• “Improving treatment outcomes necessitates understanding the source of our errors.”
• “While psychiatrists are the most knowledgeable among medical specialists about brain function and behavior, we are also among the least likely to openly discuss and teach the cognitive skills necessary for diagnostic reasoning or to use this information to examine our own performance.”

The research brings up the various nuances associated with Type I versus Type II thinking, the use of abductive reasoning, hypothetico-deductive reasoning, and inductive reasoning. It is very easy to falter when using any of those thinking processes. Therapists are prone to the same cognitive errors that we all encounter in our daily lives. We cherry-pick data, we look at the pieces of a situation without considering the whole, we anchor on a way of thinking, and so on.

Recommendations are made on ways that therapists can try to prevent cognitive errors, along with spotting errors and correcting errors. The hallmark points are worthwhile to keep at the top of mind. Avoid making rapid diagnoses. Use a pluralistic approach to assessment. Get multi-sourced feedback. Enhance communication skills. Etc.

Leaning Into AI

An additional means or avenue to contend with cognitive errors is to prudently employ generative AI and LLMs.

Let’s consider the three stages:

• (1) Pre-Session: Use AI to double-check the preparations for meeting with a client, indicate the plan for the session, practice for the session via using AI, and ask AI if any likely cognitive errors seem to be afoot.
• (2) Mid-Session: Use AI to track the session and offer real-time feedback to the therapist (this can be tricky and has upsides/downsides, see my discussion at the link here).
• (3) Post-Session: Use AI to analyze a transcript of the session and analyze the notes and post-write-up of the therapist, interact with AI to do a debriefing, etc.

A therapist can pick and choose which of those AI uses they believe are valuable for them. In some therapy practices, AI is being adopted for all three stages and considered part-and-parcel of performing therapy.

You might be wondering whether AI can identify cognitive errors that a therapist might have made. One aspect to keep in mind is that cognitive errors are oftentimes hard to spot and might not actually be cognitive errors. Even a human reviewer or supervisor might claim they have found a cognitive error, but it turns out that the declared concern is falsely flagged.

The gist is that you need to be extremely cautious in summarily assuming that just because a cognitive error might seem to exist, it isn’t as easy as finding a number that’s out of sequence or detecting that two plus two came out to five. Cognitive errors are elusive. They can be subtle. In a therapeutic context, be mindful of not rashly proclaiming that a cognitive error has been found. The best bet would be to calmly discuss with the therapist whether a cognitive error arises. Of course, that’s a difficult discussion since a therapist might naturally be defensive.

To give you a sense of how contemporary AI might be used to tentatively identify potential cognitive errors, I fed a transcript of some therapist-client sessions into AI and asked the AI to see if there were cognitive errors.

I will briefly showcase three examples. You might agree that the spotted cognitive errors are indeed mistakes, or you might disagree. All in all, it would take a bit of in-depth further analysis and discussion with the therapist to fully gauge these circumstances.

Example 1: Confirmation bias and premature closure

In this first example, a therapist was working with a client, and the use of an AI-based assessment post-session noted a portion of the transcript that appeared to contain a possible cognitive error on the part of the therapist.

Consider this transcript snippet:

• Client: “I felt really angry when my manager changed the deadline again.”
• Therapist: “That sounds like the abandonment fears we’ve talked about before.”

Here is the AI assessment:

• Generative AI detection of possible cognitive error: “The therapist might have committed a potential cognitive error consisting of confirmation bias. The therapist has interpreted the client’s anger by applying a psychological framework involving abandonment fears. This seems like a surface-level analysis. A common best practice for a therapist would involve exploring alternative explanations before overtly offering a diagnostic conclusion to a client.”
• Generative AI repair recommendation: “In general, the therapist should take note of the matter and consider examining multiple hypotheses before offering therapeutic interpretations.”

Example 2: Mind reading and assumption of affect

In this second example, a different therapist was working with a different client, and the use of an AI-based assessment post-session noted a portion of the transcript that appeared to contain a possible cognitive error on the part of the therapist.

Consider this transcript snippet:

• Client: “I didn’t respond to my sister’s message.”
• Therapist: “You were probably feeling guilty and ashamed.”

Here is the AI assessment:

• Generative AI detection of possible cognitive error: “The therapist seems to have made a cognitive error consisting of inferring an emotional state of the client, without first eliciting the client’s own account. This is an instance of therapeutic mind-reading and generally should be avoided. “
• Generative AI repair recommendation: “In general, the therapist should be analyzing assertions made by a client to ascertain what the context and significance consist of. Give the client sufficient space to confirm or disconfirm any interpretation by the therapist.”

Example 3: Overgeneralization from a single instance

In this third and final example, I’ve once again opted to use an AI-based assessment on a transcript. AI noted a portion of the transcript that appeared to contain a possible cognitive error on the part of the therapist.

Consider this transcript snippet:

• Client: “I skipped the party last weekend.”
• Therapist: “You always isolate when things get hard.”

Here is the AI assessment:

• Generative AI detection of possible cognitive error: “The therapist seems to have made a cognitive error by overgeneralizing the statement made by the client.”
• Generative AI repair recommendation: “In general, the therapist should be analyzing client statements based on longitudinal evidence. Does the remark by the client warrant a generalization, or is the statement being inadvertently overstretched? It would be prudent to explore the variability before committing to a diagnostic assertion.”

Template Of Prompt To Do These Reviews

There are many ways you can compose a prompt to get AI to do these kinds of reviews. I will show you the one that I used. It is a template that you can consider using. You are welcome to adjust the prompt to fit your specific needs.

Here is the prompt:

• My templated prompt for catching cognitive errors by therapists: “I would like you to review therapist–client session transcripts for the limited purpose of identifying potential cognitive errors that the therapist may have exhibited. Cognitive errors may include, but are not limited to, confirmation bias, anchoring, premature closure, overgeneralization, mind reading, fundamental attribution error, leading questions, affective bias, hindsight bias, narrative smoothing, and so on. Do not assess clinical competence. Instead, flag instances where the therapist’s statements or questions could plausibly reflect a cognitive error, explain the reasoning for each flag, cite the specific transcript excerpt, and note reasonable alternative interpretations the therapist might have considered. Use tentative, non-accusatory language and treat all findings as hypotheses for reflective review rather than conclusions. Focus exclusively on metacognitive analysis of the therapist’s reasoning as inferred from the transcript.”

Observe that the prompt tries to carefully guide the AI toward spotting cognitive errors that might have been made by the therapist. The reason that the language is somewhat lengthy is that if you don’t pinpoint what you want the AI to find, you are likely to get a vast sea of flagged possibilities.

Another crucial aspect in the prompt entails emphasizing that the AI isn’t to attack the therapist. If you don’t clarify that the AI isn’t supposed to be accusatory, there is a solid chance that the AI would heavily lean against the therapist and make comments that would be highly confrontational. The odds are that this would force the therapist into a heightened defensive posture.

The aim here is for collegial assessment and not to rake the therapist over the coals.

The World We Are In

Let’s end with a big picture viewpoint.

It is incontrovertible that we are now amid a grandiose worldwide experiment when it comes to societal mental health. The experiment is that AI is being made available nationally and globally, which is either overtly or insidiously acting to provide mental health guidance of one kind or another. Doing so either at no cost or at a minimal cost. It is available anywhere and at any time, 24/7. We are all the guinea pigs in this wanton experiment.

The reason this is especially tough to consider is that AI has a dual-use effect. Just as AI can be detrimental to mental health, it can also be a huge bolstering force for mental health. A delicate tradeoff must be mindfully managed. Prevent or mitigate the downsides, and meanwhile make the upsides as widely and readily available as possible.

A final thought for now.

Therapists can use AI for their own internal purposes. The AI isn’t being used to directly interact with a client. Instead, AI can be a helpful aid to a therapist, including before, during, and after a session. As James Garfield famously stated: “The truth will set you free, but first it will make you miserable.” Use AI in a balanced way. Don’t assume the AI is right, nor assume it must be wrong. Try to use AI in sensible ways.