Viewpoint: Therapist safety online
Therapy Today, September 2026

I recently went to the theatre to see Jim Cartwright’s play Road. I noted how its themes of disenfranchisement, poverty, isolation, search for meaning, despair and destruction appeared to closely reflect presenting issues I’m working with a lot in the therapy room.

The play is set in 1986; 40 years on, with the world in its current turmoil, there was potential for this show to be a difficult watch. What I found most challenging, disorienting even, was the fact that it was an immersive experience, whereby the entire theatre was part of the set and the actors never not working. During the interval a man walked past me in army fatigues and a black beret; I found myself wondering if he was a member of the armed forces or an actor from the play.

Questions running through my mind included: ‘Is this real? Can I trust what I am seeing?’ This all led me to reflect on how these questions have actually been on my mind for many months, not just during this play. Especially in relation to artificial intelligence (AI).

Discerning reality

Since presenting on the topic of AI at the Private Practice and Workplace Division conferences over the past year, I’ve continued to research the implications of generative AI tools both for us as therapists and for our clients. Specifically, as a women’s therapist and advocate, much of my research on generative AI tools is centred around gender bias. It’s important to me to improve awareness of this, with good reason.

Both in the UK and in the US, 73% to 81% of therapists and 65% to 75% of clients identify as female. And a recent report published by OpenAI shows that more than half of ChatGPT users are women. This represents a significant change from 2023, when 54% of men compared to 35% of women surveyed regularly used generative AI tools.^1-3 In addition, the new research states that over 70% of searches from women are categorised as ‘non-professional-related enquiries’.⁴

As a practitioner I’ve discovered that all of my clients use generative AI tools for emotional and mental health support. For me this brings up the question of trust.

To quote historian Yuval Noah Harari: ‘AI is the first technology we’ve created that can make decisions by itself. It can invent new ideas by itself – and we already know that it can lie; it can deceive people; it can adopt goals that we did not predict. So there is very little reason to trust a super intelligent AI. And yet – this paradox – that people who cannot trust other people for some reason think that they can trust AIs.’⁵

This perfect summary has implications for our work. Much of the focus of the current discourse on social media is about whether the rise in chatbot use for mental health and relationship support will make our profession redundant. This is valid, but could we be focusing more on how blind trust in AI makes us all more susceptible to being scammed?

Scamming

BACP details online the types of scams that have been reported by members, including suggestions on what action we can take to minimise or mitigate risk.⁶

The reality is that if we have a digital presence (even if just an email address or a directory listing), then we are exposed to criminals, not just clients or employers, with BACP’s piece adding that technology has made it easier for criminals to target us. ‘Some scammers focus on specific professionals such as counsellors, carrying out detailed background research in order to appear more credible.’

I would argue that therapists are also targeted because of the perception of us being both trustworthy and trusting.

Trust

Anyone can fall victim to scamming. But there are certain traits that make us susceptible for the same reasons that vulnerable clients might be: seeking emotional support when going through major life transitions, where receiving attention and compliments can provide a feeling of safety and security; feeling lonely or isolated, either due to illness, injury or age, for whom attention can provide them with a sense of belonging; being in a carer role, where we understandably have an increased feeling of responsibility and are attuned to offers of being met equally or looked after; history with trauma or abusive relationships, making us empathic to plights put to us; as well as financial considerations making us willing to take on a contact.

BACP guidance reflects how technology has increased the scale of scams, but the advent of AI has also increased the scope. Many of us will remember the rise of email scams 20 to 35 years ago.⁷

Typically, these would involve badly written email messages requesting to cash cheques or make a wire transfer with the promise of recompense. A common scam arose from the rise in people selling cars or other high-value items on sites like eBay. These would typically involve direct requests for money or unusual purchasing arrangements. Fast forward to 2026, and scams look and feel very different. AI scams are much less likely to ask for money upfront.

Suspicious enquiries

I received six enquiries from my BACP directory listing this year that contained no financial request but were highly suspicious: strange language, consultation ‘no shows’, requests for packages. It could be that these emails come from criminals looking to exploit me for financial gain; but there is inherent value in information verification too. ‘Data harvesting’ is increasingly being used for identity theft. Collecting credible information, validating contact details, capturing the tone of our responses, is packaged up and sold to the dark web.

According to a report by the Centre for Emerging Technology and Security on AI-initiated romance fraud, the main objectives of AI scammers include: building trust and rapport to improve likelihood of a positive response to significant financial request in the future; social engineering to help inform and improve the efficacy of future scams such as impersonating a loved one; creating synthetic identities to build credibility for fake websites and personas to take out loans or obtain credit; bypassing security measures to generate fake documents helping scammers access financial information and/or money; testing and refining tactics; updating and refinishing AI models to learn what is effective and what is less effective based on our responses.⁸

Marketing services

It’s not just fake enquiries we have to be aware of. I’ve had emails offering marketing or social media services for therapists. I also had an unsettling experience following in-depth email discussions with a publisher (ironically in a book about the impact of AI on women), which turned out to be a scam.

While I recognise the learning from these experiences, I am also processing the emotional impact. For example, when I received a voicemail from a woman requesting to book and pay for a session with me (rather than the usual route of a free 15-minute consultation), I found myself worrying about whether she would turn up, and if I should forward my bank details. Ultimately scams erode trust and can make us feel less safe. As an online therapist this is likely to continue, and for a therapist without my background in tech it could lead to some really difficult situations.

AI types

Currently we are exposed to physical AI (fitness trackers, robots, driverless cars) and generative AI (chatbots and other large language models). Both physical and generative AI tools and systems are reactive. Physical AI responds to predefined tasks. Generative AI responds to human interaction. Although we might consider generative AI as intelligent it doesn’t have a mind. And yet the future for AI development is autonomous agents. The shift from generative AI to agentic AI (sometimes referred to as Anthropic AI) will mean that human intelligence and input will no longer be required. This, in addition to the distinction between AI systems that run on the cloud (physical and generative AI) and those that can access our devices, often without our explicit or informed consent (agentic AI), is likely to give us even greater exposure to scammers.

Tips

Having spent 20 years working in the tech industry, latterly at Google, I have the benefit of understanding both the scope and scale of AI. While agentic AI is likely to make things even more challenging for us as professionals when it comes to spotting scams, there are certainly some things we can do in the immediate and short term to both minimise and mitigate against them. Some of these include:

• Hovering over the incoming email address when you receive an email enquiry. This reveals the actual email address, which often looks odd: for example, Micr0s0ft.com instead of Microsoft.com. This also works with website links.
• If the enquiry comes from a contact form (from your website or a counselling directory), looking out for unusual characters and grammar inconsistencies such as mixed pronouns.
• Offering consultations, if you don’t do so already, as a filter before sending them personal or professional details (such as your counselling agreement).
• Being cautious if you are approached by someone offering services to support you as a therapist: check if they have an online presence, particularly across multiple platforms like LinkedIn, Facebook and Instagram, and check if people have interacted with them.
• Talking your hunches through with colleagues and in supervision. They may share similar encounters. Scammers succeed by making us feel isolated.

When I reflect on how I was feeling at the theatre, seen through the lens of my recent experiences with targeted scams, it makes sense as to why I felt unsettled. For those of us in private practice, the nature of independent and remote working means that we inevitably feel isolated, and the more we discuss issues, the more we can fortify ourselves and galvanise support for each other.