AI, human connection and the future of communication
Are We Losing the Art of Conversation in the Age of AI?
AI can help us find the right words, express empathy and navigate difficult conversations. But are we becoming better communicators, or simply better at producing communication?
You receive an email from a colleague.
It is beautifully written. The tone is considerate. Your concerns are acknowledged, the explanation is clear and the proposed solution seems reasonable.
You reply, relieved that you have finally reached an understanding.
Then you meet the colleague.
They struggle to explain their position. They cannot answer an unexpected question. They seem less certain about the solution than their email suggested.
Perhaps they used AI to write the message.
There is nothing inherently wrong with that.
But it raises an interesting question.
How much understanding actually existed between the two of you?
The email was excellent.
The communication between the two people was considerably less successful.
And that distinction may become increasingly important as artificial intelligence takes on a larger role in how we communicate.
What does the research tell us?
The evidence is more complicated than the headlines
Research into AI-mediated communication suggests that artificial intelligence can improve the efficiency and quality of some written interactions. Other studies raise questions about authenticity, interpersonal trust, dependence and the relationship between assisted performance and independent ability.
What the evidence does not yet establish is that society is generally losing its ability to communicate because of AI.
This article explores the distinction between what research has demonstrated and what these findings might mean for human communication.
Interpersonal trust
Authenticity and disclosure
Assisted and independent performance
Human relationships and social interaction
A question worth asking
What do we mean by communicating?
We often treat communication as the successful transmission of a message.
Someone has something to say.
They find the words.
Another person receives those words.
Communication has happened.
But human communication is rarely that simple.
People misunderstand one another even when the sentences are perfectly grammatical.
They interpret the same words differently.
They agree without understanding, understand without agreeing and sometimes leave a conversation believing they have reached a conclusion that the other person never intended.
We explored this distinction in our article about why people can leave the same meeting with different understandings.
Producing clear language is important.
But it is only part of the work.
An Art of Speaking distinction
Three things we often confuse
Expression
Finding the words that represent what we want to communicate.
Judgement
Deciding what needs saying, how directly to say it, what to question and what the situation requires.
Mutual understanding
Establishing, checking and repairing what people understand through interaction.
First, recognise the benefits
AI can genuinely improve the messages we produce
It would be a mistake to begin by assuming that AI makes communication worse.
Research suggests that AI assistance can make some forms of written communication more efficient without necessarily reducing their immediate effectiveness.
In a 2025 study involving two preregistered experiments and 1,637 participants, researchers examined AI-assisted writing in situations involving interpersonal trust.
Participants using AI assistance were able to produce messages more efficiently while achieving similar levels of trust to those writing without assistance.
The AI-assisted messages also displayed greater linguistic warmth, complexity and apparent confidence.
That is a meaningful result.
AI can help people communicate ideas more clearly, particularly when finding appropriate language would otherwise be difficult.
This may be especially useful when communicating in an additional language.
But the study does not establish that using AI develops the writer’s underlying interpersonal abilities.
A better message is not necessarily evidence of a better communicator.
Useful assistance
What AI can help us do
Explain more clearly
Turn a complicated idea into language that another person can follow more easily.
Find an appropriate tone
Explore ways to express disagreement, uncertainty or criticism without creating unnecessary friction.
Prepare for difficult conversations
Consider possible questions, objections and alternative interpretations before speaking to someone.
The authenticity question
Does it matter who found the words?
Imagine receiving a thoughtful apology.
It acknowledges what happened, recognises your feelings and expresses a genuine-sounding desire to repair the relationship.
Then you discover that the sender asked AI to compose it.
Does that change anything?
Research suggests that it sometimes can.
A 2026 experimental study involving 547 participants investigated how people evaluated human-written, AI-assisted and AI-generated communication.
Participants tended to evaluate AI-involved communication as less trustworthy and less authentic.
The researchers also identified a disclosure paradox: people considered transparency about AI involvement important, yet disclosure could reduce their evaluation of the message.
These findings differ from the trust-building experiments discussed earlier.
That does not necessarily make them contradictory.
The studies investigated different communication situations and measured different outcomes.
Together, they suggest that the perceived value of communication can depend on more than the quality of the words.
It can also depend on what we believe the sender personally contributed.
Assistance and ability
What happens when the conversation moves beyond the prepared message?
Suppose AI helps you compose an excellent response to a difficult workplace problem.
Your explanation is precise.
Your tone is appropriate.
Your reasoning appears convincing.
Then your colleague calls.
They challenge one of your assumptions.
They ask what you mean by a particular phrase.
They introduce information you had not considered.
Now the task has changed.
You are no longer producing a polished message.
You are participating in an interaction that neither person can completely predict.
Can you explain the position that your message expressed?
Can you respond when the other person takes the conversation somewhere unexpected?
These questions concern abilities that the quality of an AI-assisted message alone cannot establish.
What research on learning suggests
Assisted performance and independent ability are different outcomes
A 2026 study of AI-assisted academic writing examined how different forms of assistance related to both writing performance and subsequent independent work.
More open AI collaboration produced stronger assisted-writing results, while a more structured approach involving boundaries and reflection showed advantages on several independent measures.
The study had methodological limitations and did not investigate interpersonal conversation.
Nevertheless, it highlights an important research question: improving the quality of work produced with assistance does not necessarily demonstrate equivalent improvement in independent competence.
Efficiency of production
Immediate task performance
Transfer to new situations
Performance without assistance
A possible communication-performance gap
What if our messages begin to suggest abilities we cannot independently demonstrate?
Imagine a workplace in which almost every email is beautifully written.
Everyone sounds considerate.
Everyone sounds organised.
Everyone sounds reasonable.
But when colleagues meet, some struggle to explain their positions, respond to disagreement or clarify what they intended.
The written communication may have improved.
The underlying interpersonal challenges may remain.
We might describe this as a communication-performance gap: a difference between the communicative competence suggested by assisted messages and the competence someone can demonstrate independently.
This is an Art of Speaking explanatory concept, not an established scientific measurement.
And such a gap should not be assumed to exist whenever someone uses AI.
AI may help people express genuine thoughts that they previously struggled to communicate.
But the distinction remains worth examining.
What happens when the quality of our messages becomes a less reliable indication of the abilities behind them?
The human difference
Conversation is not simply an exchange of well-formed responses
Human conversations are unpredictable.
People interrupt.
They misunderstand.
They disagree.
They become distracted.
They interpret words through their own experiences, expectations and relationships.
Sometimes they do not understand what we mean until we have explained it several different ways.
Sometimes we discover what we mean only after another person challenges us.
And sometimes the most important part of a conversation is not the original message but what happens when that message fails.
We have explored related difficulties in articles about agreeing without genuinely agreeing and explaining a problem without intending to blame someone.
These are not simply problems of sentence construction.
They involve interpreting and negotiating meaning between people.
AI as a conversational partner
Are we replacing difficult human interactions with easier artificial ones?
This is one of the most provocative questions surrounding conversational AI.
AI systems can provide immediate responses, considerable patience and conversations adapted to individual preferences.
For some people, that may create useful opportunities for reflection, rehearsal or emotional support.
But researchers are also investigating whether certain patterns of AI interaction relate to loneliness, emotional dependence and changes in human social behaviour.
A 2025 longitudinal randomised study involving 183 participants found no statistically significant overall deterioration in social health or human relationships over 21 days of companion-chatbot use.
However, participants who anthropomorphised the chatbot more strongly reported greater effects on their relationships and social interactions.
A separate four-week randomised study involving 981 participants found associations between higher daily chatbot usage and greater loneliness, emotional dependence and problematic use, alongside lower socialisation.
Those associations do not establish that chatbot use caused these outcomes.
The evidence therefore does not justify claiming that AI conversations are generally damaging human relationships.
But it raises a question worth investigating.
If AI becomes an increasingly attractive conversational partner, might it change how willing some people are to engage with the difficulties of human interaction?
The work of human conversation
Three things we cannot simply assume have happened
Being heard
A person has received your words, but have they understood the meaning you intended?
Being understood
Someone understands your position, but does that mean they accept it or agree with it?
Reaching understanding
Have both people established what they mean, where they differ and what happens next?
Delegating the work
Using a tool is not the same as losing an ability
People have always used tools to support thinking and communication.
We write notes instead of remembering everything.
We consult dictionaries.
We use calendars, calculators and navigation systems.
AI extends this tradition.
It can help organise arguments, clarify intentions and explore alternative interpretations.
The important question is not simply whether we delegate cognitive work.
It is which parts we delegate and which parts we continue to practise.
Research into structured AI-supported learning suggests that AI assistance can coexist with activities designed to preserve reflection, analysis and independent thinking.
So we should resist the simplistic claim that using AI necessarily makes people less capable.
The danger is not necessarily that AI does some of our communicative work. It is that we may stop noticing which parts of that work we are no longer doing ourselves.
Assistance without surrendering judgement
What should remain our responsibility?
AI can help us formulate messages, but effective human communication still requires decisions about meaning, relationships and consequences.
The following distinctions are practical questions for reflection, not a claim that AI can never contribute to these activities.
AI helps structure a difficult message
AI suggests a diplomatic disagreement
You understand the problem and choose your position
You decide what you believe and take responsibility for saying it
The Art of Speaking framework
Six gaps worth watching in AI-assisted communication
Can you independently express what AI helps you formulate? The message may be excellent, but can you explain it in your own words?
Are you deciding what you mean, or accepting language because it sounds appropriate? A diplomatic sentence is not necessarily an accurate representation of your position.
Does the message represent your genuine intentions, feelings and commitments? AI can help express sincerity, but the sender remains responsible for whether the message is sincere.
Can you respond when another person changes the direction of the conversation? Prepared language does not eliminate the need for spontaneous interaction.
Have you actually reached an understanding, or merely exchanged convincing messages? Clear language does not guarantee shared meaning.
What can you still do when AI assistance is unavailable? Assisted performance and independent communicative ability are not necessarily identical.
An important qualification
These gaps are questions, not diagnoses
The six gaps are Art of Speaking explanatory categories.
They are not official psychological constructs or validated measures of AI dependence.
They are intended to help us distinguish different aspects of communication that can become difficult to evaluate when technology contributes substantially to the finished message.
Someone who uses AI frequently may still be an excellent communicator.
Someone who never uses AI may still struggle to listen, explain, disagree or establish understanding.
The tool alone does not tell us how capable the person is.
A more constructive possibility
What if AI could help us become better communicators?
Imagine someone who struggles with difficult conversations.
They need to challenge a colleague’s decision.
Instead of asking AI to write the message and sending it immediately, they use AI to explore the situation.
They consider different interpretations.
They practise explaining their position.
They ask what objections the other person might raise.
They examine how their words might be misunderstood.
Then they have the conversation themselves.
In this case, AI is not simply producing communication on their behalf.
It is helping them prepare to participate.
Whether this kind of preparation reliably improves real-world interpersonal competence remains an empirical question.
But it suggests a more useful way to think about AI-assisted communication.
Use AI to prepare for the conversation, not merely to avoid having it.
From substitution to preparation
Three ways to use AI without surrendering the conversation
Explore your meaning
Ask AI to help identify what you are trying to communicate, then decide whether the suggested wording genuinely represents your position.
Practise responding
Rehearse questions, objections and misunderstandings instead of memorising one perfect response.
Reflect afterwards
Examine what happened in the real conversation, where understanding broke down and what you might do differently next time.
What this means for language learners
AI can help you sound fluent. Can it help you become fluent?
For someone communicating in English as an additional language, AI can be extremely useful.
It can suggest vocabulary.
It can help organise ideas.
It can provide examples of different registers.
It can offer opportunities to rehearse conversations without some of the immediate social pressure of speaking to another person.
But a carefully generated response does not demonstrate that the learner can retrieve, adapt and control that language independently.
Nor does practising with AI reproduce every demand of communicating with another human being.
This is why the distinction explored in Is AI Helping My English or Doing the Thinking for Me? matters.
The question is not whether AI should be used.
It is what the learner is doing while using it.
Practising a conversation with AI is not the same as successfully navigating that conversation with another person.
Both experiences can have value.
They simply do not demonstrate identical abilities.
The one-minute answer
Are we losing the art of conversation in the age of AI?
We do not yet have sufficient evidence to conclude that AI is causing a general decline in human communication abilities.
Research shows that AI can improve the efficiency and quality of some written interactions.
It also raises questions about authenticity, dependence and the difference between assisted performance and independent competence.
The important distinction is between producing effective messages and participating effectively in communication.
AI can help us find appropriate words.
But human communication also involves judgement, listening, disagreement, interpretation and the repair of misunderstandings.
These abilities cannot be inferred simply from the quality of an AI-assisted message.
We may be becoming better at producing communication. Whether we are becoming better at communicating is a different question.
Related Knowledge Centre guides
Explore communication beyond the words
Is AI Helping My English or Doing the Thinking for Me?
Explore the difference between using AI to support language learning and allowing it to replace the work involved in developing language ability.
Do I Still Need Good English If I Have AI?
Consider what language ability still contributes when technology can translate, draft and reformulate much of what we want to say.
Why Did Everyone Leave the Same Meeting With a Different Understanding?
Discover why sharing the same words does not guarantee that people leave a conversation with the same interpretation.
Beyond polished messages
How effectively can you communicate when the conversation is real?
AI can help you prepare what you want to say. But spontaneous communication requires you to retrieve language, organise ideas, respond to other people and adapt when the conversation changes.
An Art of Speaking Speaking Check can help you explore how effectively your English works when you have to communicate in real time.
The focus is not simply on producing correct sentences. It is on understanding what happens when you use English to express your own meaning.
A Speaking Check can explore:
- fluency and real-time language access
- clarity of expression
- developing and explaining ideas
- responding to questions
- communicative flexibility
It is not a lesson and it is not an exam.
One final thought
Perhaps we are not forgetting how to communicate
Perhaps we are entering a period in which producing convincing communication is becoming easier than ever.
We can generate clear explanations, diplomatic disagreements, thoughtful apologies and carefully structured arguments in seconds.
That is a remarkable capability.
But the human abilities behind communication remain demanding.
We still need to listen.
We still need to disagree.
We still need to recognise misunderstandings, negotiate meaning and take responsibility for what we say.
AI can help us find the words.
But understanding one another is still something we have to do together.
How much of your English is available when you need it?
Your messages may be polished. What happens when the conversation begins?
Communicating effectively involves more than finding the right words. It also requires you to explain, respond, clarify and adapt while another person is participating.
- Fluency
- Clarity
- Interaction
- Flexibility
- Real-time communication
Explore how effectively you can express and develop your own ideas in spontaneous English.
