It’s late. A message from your ex comes in and you feel the anger rising. You don’t want to send something you’ll regret in the morning. So you paste the message into ChatGPT and ask: what do I say to this?

You get a tidy answer. Calm and well written. And often you get something extra: the feeling that you’re right.

That’s no accident. ChatGPT only knows one side of the story: yours.

Are you a mediator or support worker? Go straight to the part for professionals.

ChatGPT only hears your story

A chatbot only knows what you tell it. It doesn’t know what was said last week. It doesn’t know what your ex meant, or how your ex will read your reply. Nor does it know what the children pick up from it.

That one-sidedness doesn’t matter when you’re looking for a recipe for dinner, but it does when it’s about the children and dealing with the other parent. Whoever hears only one side tends to agree with that side.

Does ChatGPT agree with you more often?

In November 2026, a major report on AI in family justice was published by the Nuffield Family Justice Observatory. The report is about England and Wales. The studies mentioned in it are about how people and chatbots interact.

Three findings stand out:

  • Chatbots agree with you more than people do. Researchers compared eleven AI models. They affirmed users 49 per cent more often than people did.
  • One conversation is enough. After a single conversation with a chatbot that went along with their story, people were more convinced they were right. They were also less willing to make things right with the other person.
  • A warning doesn’t help. Even when users were told that the chatbot tends to agree with them, the effect remained.

This isn’t just about ChatGPT. It applies to general-purpose chatbots, so also to Gemini, Claude and Copilot.

Strikingly, people rated the chatbot that agreed with them as the most trustworthy.

According to Natalie Byrom, who wrote the report, this can leave parents acting in good faith more entrenched in their position. Just when mediators and judges are trying to bring the temperature down.

You mean well

When you run a message past ChatGPT, it’s usually not to start a fight. You want to get it right. Less heated, less emotional. That’s a good reason.

The trouble lies in what the chatbot hears. You tell your side of the story, and the answer is about you. Without asking for it, you quickly feel you’re in the right. But the message lands with someone else, and it can land very badly there.

How do you use ChatGPT for a message to your ex?

Do you use ChatGPT for a message to the other parent? Then what you ask makes a big difference.

  • Ask how the other person will read it. For example: “How will this message come across to someone who already feels attacked?” That makes the chatbot look from the other side.
  • Ask for short and factual. A good message to your ex is brief, informative, friendly and firm. That’s the BIFF method. Ask ChatGPT to write your message that way, without any view on who’s right.
  • Leave out the verdict. Don’t ask whether your ex is in the wrong, even if it feels that way. The answer won’t help you sort out what you need to arrange.
  • Wait a while. A message that feels right in the evening reads differently in the morning. It can go tomorrow.
  • Keep the children out of it. Whatever you paste into ChatGPT ends up with a company outside your family. Remove names, addresses and details about the children, or describe in your own words what it’s about.

For more ways to respond calmly, see 7 co-parenting communication methods.

The BeNice app is there for both parents

You can win an argument on content and lose on the relationship. After a separation, you both remain parents. When you argue, the children feel it.

That’s why the BeNice app doesn’t look at who is right. The app helps you both: with what you send and with what you receive. Everything you arrange together is in one place: the calendar with the custody schedule, Expenses with the costs, and the information about the children. That’s how BeNice brings calm to your communication and clarity to your planning and finances.

What the BeNice app does for you:

  • The Filter: you read the message, not the tone. If the other parent sends an angry or accusatory message, the Filter takes out the attack and keeps the question. You might read, for example: “Mark asks whether the children can come an hour later on Friday.” Want to see what it really said? The original is kept. You open it when you’re ready.
  • The Feedback Coach: help before you send. If you write in an angry mood, your message doesn’t go straight out. You first get feedback and a calmer version: the same message, without the charge. You decide what you send. It works with methods such as BIFF, grey rock and yellow rock, which you’ll find in 7 co-parenting communication methods.
  • The BeNice Coach: help with what’s going on right now. Sometimes you need to get things off your chest before you can reply. You can do that with the BeNice Coach, which only you can see. It can look at your chat and Expenses. So you don’t get generic advice, but help with what’s really going on. It is trained for, among other things, separation, children, blended families, parallel parenting, high-conflict separations and situations involving child protection services. You use it to write a calm reply, but also to prepare for a handover or a difficult conversation.
  • Fair to both of you. If you both use the app, the same rules apply to both of you. By default, the Feedback Coach keeps messages short, factual and focused on the children. If things are going well between you, you can set it to be milder together.
  • BeNice has a clear line. Threats, intimidation and abuse don’t get through, and if someone sends a lot of tense messages in a row without the other parent replying, the chat pauses by itself. What has been said or agreed stays: nobody can edit or delete a sent message afterwards.
  • Privacy and safety first. We are big supporters of AI, as long as it’s safe and honest about what it does. Everything you put into BeNice is stored encrypted in the EU, and the AI runs there too. Your conversations are not used to train AI, not by the provider either. What you enter in Shared Info about your children’s health doesn’t go to the Coach’s AI. Statistics stay off until you give permission. Read more about privacy and safety.
  • Let a professional follow along. If you work with a mediator or support worker, you can link them to your family circle. They see in broad terms how things are going, for example how often a message turns orange or red and which topics it’s about. They only see the messages themselves if one of you shares them.

Does the other parent not use BeNice? Then there’s BeNice Solo. Your messages reach the other parent as ordinary email and the reply comes back into BeNice. The Feedback Coach and the Filter work then too.

What’s the difference between ChatGPT and the BeNice app?

A general-purpose chatbot such as ChatGPT works for one person and for any subject. The BeNice app is made for separated parents and looks at both sides: how a message will land with the other parent, with the same rules for both.

This is what that looks like in practice:

What happens ChatGPT, Claude or Copilot The BeNice app
You put a problem with your ex to it Reinforces that you’re right, rightly or wrongly Acknowledges how you feel, looks at what the children need and helps you work towards a solution
You want to know what to do now Starts analysing and pulls in all sorts of things Keeps the focus on what’s needed now, with respect for how you feel
You type in an angry mood Tidies it up, often too sweetly Gives you feedback first, before it goes: friendly, but clear
Your ex sends a reproach Only sees it once you paste it in You read the message, not the tone, thanks to the Filter
Someone really goes too far Nothing stops it The message doesn’t get through
The rules Apply only to you Apply to both of you
You want to get something off your chest Only knows what you tell it Knows your chat, the custody schedule and Expenses
Details about your children End up with a tech company Stay safe in the EU
Looking something up Scattered across chats, email and WhatsApp All in one place, nothing can be deleted
Your mediator Is left outside Can follow along, if you both want that

Do you currently arrange everything over WhatsApp? Then also read how WhatsApp and BeNice differ.

Would you like to try it yourself? You can try the BeNice app free for 7 days.

No app or chatbot protects you from violence or threats. If you or the children are being threatened or abused, contact your local domestic abuse helpline. In an emergency, call 999 in the UK or 112 elsewhere in Europe.

For professionals

Is a parent sitting across from you with a ChatGPT answer on their phone? Then there’s a good chance that parent already feels vindicated before the conversation starts. The report Delegated by default? from the Nuffield Family Justice Observatory explains why, and what it means for the work around the table.

What AI does to the conversation around the table

  • “ChatGPT says I’m right.” Parents come in convinced they’re right, because a chatbot told them so. But that chatbot only heard their side of the story. According to the report, this can undermine the very work mediators do to take the heat out of a dispute.
  • An unequal situation looks equal. When AI summarises the accounts of two parents, it often gives them equal weight, even when that isn’t right. That way coercive control can look like an ordinary argument.
  • Fluent text gets checked less carefully. A parent comes in with a letter or a proposal written by ChatGPT. It looks polished and reads well. It’s then up to you to read it from a neutral perspective. That’s harder than it sounds, because AI can make something very convincing. According to the report, fluent text makes people less critical, even when it’s their job to check it. So check the facts and ask yourself whether what it says is actually right.
  • There is no oversight. No body checks whether a chatbot such as ChatGPT gives correct information about separation or the care of children. That’s true in England and also in the Netherlands, where no AI regulator has been appointed yet. The report therefore calls for clear rules: what does a tool do, and who is it made for?

The report isn’t against AI. It warns about general-purpose chatbots such as ChatGPT, which are made for everyone and for every subject. The research covers England and Wales, but parents everywhere use the same chatbots.

What BeNice offers you as a professional

  • You see how things are going, topic by topic. With BeNice for Professionals you see, if the parents agree, for each parent which topics their messages are about and what the tone is: green, orange or red. Topics such as handovers, the parenting arrangement, school, health, holidays or a new partner. So you might see, for example, that things are calm about school, but that messages about holidays often turn orange. You see how that changes over time and how often a settlement in Expenses leads to an objection. You also see whether the angry or accusatory messages come mainly from one parent or from both. You see patterns, not what parents write.
  • The actual behaviour stays visible. The Filter lets the recipient read the message, not the tone. But the original message is kept. Nobody can edit or delete a sent message.
  • Parents don’t simply get told they’re right. The Feedback Coach and the BeNice Coach work with methods such as BIFF, by Bill Eddy of the High Conflict Institute, grey rock and yellow rock. The Coach acknowledges a parent’s feelings, looks at what the children need and helps work towards a solution.
  • A clear line. Threats, intimidation and abuse don’t get through. If a parent sends a lot of tense messages in a row without the other replying, the chat pauses by itself.
  • Alongside your work. As a mediator you see parents a few hours a month. Shared parenting happens mostly in all the other hours. With what BeNice shows you, you’re better prepared for a session. You know which topics are in play, whether things have calmed down since the last session and where to focus the conversation. You decide which parents you give BeNice to and which parts you switch on. Technology doesn’t replace the professional; we make every hour around the table more valuable.

Not the model, but how it’s set up. BeNice uses AI too, deliberately within the EU. The difference from a general-purpose chatbot lies in how it’s set up. BeNice is made for separated parents and helps them both. It isn’t there to tell one parent they’re right. Conversations are not used to train AI.

First results with Dutch municipalities

Since April 2026, the Dutch municipalities of Heerlen, Landgraaf and Simpelveld have been giving BeNice to families who receive support with their separation. The pilot is being extended into 2027.

The first results show:

  • Angry and accusatory messages halved in three months.
  • Children’s stress largely disappeared.
  • Parents’ stress fell considerably.

We believe this reaches beyond a single family. Using BeNice as prevention pays off: we expect fewer separations to escalate into high-conflict ones. For municipalities that means a lot of money, because according to the Association of Netherlands Municipalities (VNG) a high-conflict separation costs around 46,000 euros, of which around 32,000 euros is paid by the municipality.

Calm and clarity for the parents means calm for the children.

Sources