The rise of technology and AI has affected our everyday life. Over half (57% of the UK have admitted to using AI, and 73% have chatted to AI bots. From travel itineraries and work presentations to even dieting and meal plans, the general public has found AI can be useful for things that feel laborious or hard to get done. One surprising use for the AI is how it’s ‘advising’ many people in relationship turmoil.
Google searches for ‘ChatGPT Relationship Advice’ are up by an outstanding 1037%, and ‘Best AI relationship advice’ is up by 1215% since last year, with more people using the AI model to seek counsel for their relationship troubles.
National Legal Service assessed 440 comments across several Reddit threads that discussed using AI to seek relationship advice. A common theme of what people use AI for in their relationships showed the following:
- Interpreting a partner’s behaviour: detailing a partner’s behaviour to better understand their actions and seeking justification.
- Rehearsal and preparation: Using it as a sounding board to better prepare for conversations with their partner.
- Deciding whether to end a relationship: Detailing their side of events, thoughts and opinions and emotions and asking for AI to judge based on the case put to it.
- Deciding whether to divorce: Weighing up the pros and cons, the legality, finances and logistics of divorce.
- Reading the other party’s chat history: Reading a partner’s AI chat history and using AI to interpret it.
The Risks:
AI models such as ChatGPT, Claude, and Google’s own AI Mode take information from the internet and information the user inputs. This means the model can build up an understanding of only what the user adds and cannot assess the full situation.
Marius Badenhorst, AI specialist warns: ‘LLMs (Large Language Models) are designed to be enthusiastic and to be agreeable. This will mean that users can confuse the nature of these models’ responses as sentient responses.’
In a recent Stanford research paper which analysed the bonds between users and AI chatbots and how they can cause ‘dangerous feedback loops’. Researchers analysed 391,562 messages from 19 users who reported alarming experiences and identified 28 types of chatbot-user behaviours. They found that delusional thinking appeared in 15.5% of user messages and chatbots portrayed themselves as sentient in 21.2% of their messages.
Convincing but inaccurate details: AI has the ability to hallucinate. Serious subjects such as legal precedents, specific legislation, or financial equations that appear legitimate but can be entirely fabricated. In high-stakes scenarios like finalising settlements or signing binding contracts, acting on such misinformation can result in significant financial or legal repercussions.
Absence of legal privilege: Unlike confidential discussions with a qualified solicitor, interactions with AI bots are typically not protected. Depending on the service provider, your data may be stored or analysed, meaning sensitive disclosures could potentially be surfaced during legal discovery processes.
Lack of nuanced situational judgment: Family law outcomes depend on intricate factors—such as complex asset portfolios, pension valuations, domestic conduct, and the specific needs of children—which AI cannot fully grasp. Consequently, general legal concepts are often presented as if they were bespoke professional guidance.
Overlooking critical safety indicators: Legal professionals are equipped to identify red flags like coercive control or financial exploitation. An AI model, however, simply processes the prompt provided, potentially failing to recognise urgent safeguarding requirements regarding children or personal safety.
Misplaced certainty in decision-making: Perhaps the most significant risk is that AI encourages users to forgo expert counsel. This leads individuals to enter mediation without adequate preparation or to agree to unfavourable terms they would have otherwise been cautioned against.
The perspective swap test:
The “perspective swap test” reveals AI bias by asking the model to analyse a situation from the partner’s point of view. When users swap roles and ask the AI to re-judge the same scenario, the model often changes its conclusion.
This demonstrates AI’s bias to its user and places false confidence on their side.
Advice:
Family and divorce specialist from leading law firm, National Legal Service’, Charles Metherell warns that the overuse of AI for relationship advice can lead to overconfidence in people seeking to divorce or what they are entitled to without seeking qualified advice.
“We have seen an increasing number of people come to us after confiding in and consulting with an AI model such as ChatGPT. People are either looking to put the wheels in motion for a divorce or looking to negotiate what they have been told they are entitled to by AI models. The cases vary, but a crucial element has been missed, and that is to seek professional counsel at the first instance.
Although sometimes helpful, AI can mislead and not fully understand the situation without bias. We need to remember that AI, although it has the ability to pull in UK law into its ‘thinking ’, does not have the expertise or experience to guide people through what can be a stressful time of their life.
We understand that people may be in a vulnerable state when they first seek help or advice from AI, but it is also important they get the correct information and aid at the same time.”
While AI tools can offer a starting point for navigating relationship challenges, they cannot replicate the legal experience and nuanced understanding provided by a qualified solicitor. For those requiring support with family or divorce matters, consulting a specialist at the can provide the professional guidance necessary to navigate these complex situations.
