How will online behaviour change over the next ten years?
The biggest change is likely to be a shift from searching and clicking to asking and delegating. Instead of visiting websites and comparing information ourselves, we will increasingly ask AI assistants for answers and let them carry out tasks for us. But while the technology will change dramatically, the underlying psychology of why we go online, to connect, to belong, to find information and to save effort, will stay much the same.
It is worth starting with a note of caution: predicting technology is notoriously difficult, as I discuss in How far into the future can you see with clarity? But some trends are already clear. Ofcom’s latest Online Nation report found that UK adults now spend an average of four and a half hours online a day, up 10 minutes on the previous year. Generative AI is rapidly becoming mainstream too: ChatGPT had 1.8 billion UK visits in the first eight months of 2025, compared with 368 million in the same period of 2024.
Our online lives are also becoming more concentrated. Ofcom found that two companies, Alphabet and Meta, now account for more than half of the time people in the UK spend online. At the same time, more of our online activity is moving into private spaces such as messaging groups and closed communities, away from public social media feeds.
Trust is likely to be a defining issue. Only a third of UK adults (33%) now say they feel the internet is good for society, down from 40% the year before. As AI-generated content becomes harder to distinguish from the real thing, people are likely to place greater value on verified sources, trusted brands and real human connection. (See “How will we know what is real online when AI can fake almost anything?” below.)
What will not change is human nature. We will still look for shortcuts, follow other people, respond to emotion and seek belonging, as covered across these FAQ pages (see the FAQs on internet psychology and online behaviour). The organisations that thrive will be those that understand people, not just technology. And it is worth occasionally asking, as I did with my students: What will you do when the internet ends?
What skills will people need to thrive in an AI-enabled world?
People will need a combination of AI literacy and distinctly human skills: critical thinking and judgement, creativity, adaptability, curiosity, and the ability to work well with others. As AI takes on more routine tasks, the skills that matter most are those that help us decide what to do, check whether AI is right and bring people with us.
Employers expect substantial change. In the World Economic Forum’s Future of Jobs Report 2025, employers predicted that 39% of workers’ core skills will change by 2030. AI and big data topped the list of fastest-growing skills, followed by networks and cybersecurity and technological literacy. But creative thinking, resilience, flexibility and agility, and curiosity and lifelong learning were also rising in importance.
The first essential skill is using AI well. That means more than writing good prompts. It means knowing what AI is good at, where it makes mistakes and how to check its work. People who accept AI output unquestionably are vulnerable to errors and to the gradual erosion of their thinking (see the FAQs on the psychology of artificial intelligence). I believe we should all be arguing with artificial intelligence.
The second is judgement, which depends on knowledge. It is tempting to think that because AI knows so much, we no longer need to. But you cannot spot an AI error in a field you do not understand, and real expertise remains valuable (see Has the knowledge economy made your expertise obsolete? Or Could an elderly brain help your business future?).
The third group is human skills: creativity, empathy, persuasion, leadership, and collaboration. These are harder for AI to replicate and become more valuable as routine work is automated. Many people underestimate their creative abilities; often what they lack is confidence, as I explore in Do you lack creative intelligence, or do you just need confidence?
Underpinning all of these is the ability to keep learning. The specific tools will change many times over the next decade. People who stay curious, adapt quickly and manage their own attention well will be best placed to thrive. (See also the FAQs on digital productivity and workplace behaviour.)
Will AI assistants replace search engines?
Probably not entirely, at least not in the near future, but the way people search is already changing fast. More people are asking AI assistants questions instead of typing keywords, and search engines themselves now provide AI-generated answers at the top of the page. The result is that people increasingly get answers without visiting websites at all.
Traditional search remains enormous. Ofcom found that Google Search is used by 82% of UK adults and handles 3 billion searches a month. But AI is changing how it works. Around 30% of Google searches now include AI-generated overviews, and 53% of UK adults say they frequently see AI summaries.
This has a big effect on behaviour. A Pew Research Center study tracking the browsing of 900 US adults found that without an AI summary, people clicked on a search result in 15% of visits, but when a summary appeared, this fell to 8%. They clicked on links within the AI summary itself only 1% of the time, and were more likely to end their browsing session after seeing one.
The psychology behind this is simple. People are cognitive misers who prefer the easiest route to an answer (see the FAQs on internet psychology and online behaviour). A conversational answer feels faster and more natural than scanning ten links. AI assistants also cope well with the complex, conversational questions people actually want to ask. Pew found that more than half of searches with 10 or more words produced AI summaries, compared with only 8% of one- or two-word searches.
For users, the risk is that convenience reduces scrutiny. An AI summary feels authoritative, but it can be wrong, and fewer people check the original sources. For businesses, the challenge is that being found no longer guarantees a visit. Companies need to be the source that AI tools trust and cite, by publishing clear, accurate, well-structured content and building a reputation people seek by name. I explore some of these changes in Is now the right time to take a look at alternatives to Google? (See also “Why do people click on some links and not others?” in the FAQs on online trust, credibility, and consumer behaviour.)
Will AI agents start shopping and making decisions for us?
They already are, although on a small scale so far. AI agents can research products, compare prices and, increasingly, complete purchases on our behalf. Over the next few years, many routine purchases are likely to be delegated to AI, but how far this goes will depend on whether people trust AI enough to hand over their money and their choices.
Growth is rapid. Adobe Analytics, which tracks more than a trillion visits to US retail sites, found that traffic from AI services grew 393% in the first quarter of 2026 compared with a year earlier, and that shoppers arriving this way were spending more and staying longer. By March 2026, AI-referred traffic was converting 42% better than other traffic. But this is from a small base. EMARKETER forecasts that AI platforms will account for just 1.5% of US retail ecommerce sales in 2026. Consultancies expect much more; McKinsey and ICSC project that US agentic commerce could reach $1 trillion by 2030.
Psychologically, the appeal is easy to understand. Delegation reduces effort, and AI agents could solve problems such as choice overload by narrowing hundreds of options to a few suitable ones (see “What is choice overload?” in the FAQs on online trust, credibility, and consumer behaviour). People are most likely to delegate routine, low-risk purchases first, such as household essentials, while keeping control over purchases that express identity or involve emotion, such as clothes, gifts and holidays.
There are risks. People may trust AI agents too readily, an example of automation bias. Agents may be influenced by commercial arrangements that shoppers cannot see. And when an AI makes the choice, people may feel less ownership of, and satisfaction with, what they buy.
For businesses, this could be one of the most significant changes since online shopping began. If an AI agent, rather than a person, is choosing what to buy, many familiar persuasion techniques, such as attractive design and emotional imagery, matter less, while accurate product information, real reviews, clear pricing and reliability matter more. Some businesses may need to rethink their approach entirely, as I discuss in Are you prepared to pivot your business dramatically?
How will we know what is real online when AI can fake almost anything?
Increasingly, we will not be able to tell just by looking. AI can now create images, voices, and videos that most people cannot distinguish from the real thing. In future, knowing what is real will depend less on our judgement and more on where content comes from: trusted sources, verification technology and healthy scepticism.
The research is sobering. In a study by Sophie Nightingale of Lancaster University and Hany Farid of the University of California, Berkeley, 315 participants tried to tell real faces from AI-generated ones and achieved an accuracy of 48%, close to chance. Even after training and feedback, a second group only reached 59%. Most strikingly, a third group rated the AI-generated faces as 7.7% more trustworthy than real faces. The researchers suggested this may be because synthetic faces resemble average faces, which people tend to find more trustworthy. AI has improved considerably since that study.
There is a second, less obvious danger. As the researchers warned, in a world where any image or video can be faked, the authenticity of any inconvenient recording can be called into question. Legal scholars Robert Chesney and Danielle Citron call this the “liar’s dividend”: people caught on camera can simply claim that genuine evidence is fake.
Technology is starting to respond. Rather than trying to detect fakes, the emerging approach is to prove where content came from. The Content Credentials standard, developed by the Coalition for Content Provenance and Authenticity (C2PA), attaches a tamper-evident record to images and videos showing how they were made. In May 2026, OpenAI joined the C2PA steering committee and committed to adding Google DeepMind’s SynthID watermark to its AI-generated images, while Google announced that C2PA and SynthID detection would come to Google Search and Chrome. But coverage is still patchy, and many platforms strip this information when content is uploaded.
In practice, the most important skills will be psychological ones. Slow down before sharing anything that triggers a strong emotional reaction, since that is exactly what manipulators aim for. Check whether reputable sources are reporting the same thing. Be most cautious when content confirms what you already believe. And recognise that trust will increasingly be placed in sources rather than content. I explore the question of trusting AI in Could you learn to trust artificial intelligence? (See also “What makes people trust a website within the first few seconds?” in the FAQs on online trust, credibility, and consumer behaviour.)