Most of us are a few taps away from the largest collection of human knowledge that has ever existed. Before the kettle has boiled, you can find out how many people live in St Lucia, what a word means in Kwéyòl or what was agreed at last night's council meeting. Questions that once needed a library, a phone call to someone who knew or a long argument in the pub now take seconds.

And yet it has rarely felt harder to agree on what is actually true. Two people watch the same video and describe two different events. A photograph arrives and the first question is no longer what it shows, but whether it happened at all. Public arguments used to begin from common ground and split over what to do next. Now the split often comes first, over what happened in the first place.

The easy explanation is that the internet has made people stupid, gullible or paranoid. The evidence points somewhere more interesting. The internet has made us far better at finding things out. What it has changed is how we decide what to believe, and that was never really a question of information.

The conspiracy boom that wasn't

Ask most people whether conspiracy theories have spread since the internet arrived and they will say yes without thinking. Flat earth, chemtrails, QAnon and the claim that 5G masts spread Covid all feel like products of the online age. In 2020, people in Britain set phone masts on fire because of that last one.

In 2022, a team led by Joseph Uscinski at the University of Miami tested that assumption instead of repeating it. They tracked American belief in 46 conspiracy theories, some across half a century, along with six theories in six European countries and general conspiratorial thinking in the US between 2012 and 2021. They found no systematic evidence of an increase. More of the individual theories lost believers over time than gained them.

It is one study, not the final word, and it does not make conspiracy theories harmless. But it should make us suspicious of the obvious story. Plenty of people believed the moon landings were staged, or that Kennedy was not killed by one man acting alone, when the only screen in the house was a television.

What changed is that we can see them. A belief that once lived at a kitchen table, in a pub or in a newsletter photocopied for a few hundred subscribers now has a public account, an audience and a comment section. The internet may not have created many more believers. It gave every believer a microphone and gave everyone else a window into rooms we never used to see. A country can look as though it is losing its mind when what it is mostly losing is its privacy.

If belief in strange ideas has not risen, something else must have fallen.

What fell was the referee

It was trust. In 2026 the Reuters Institute found trust in news had fallen to 37% globally, its lowest since measurement began in 2015, with significant drops in 29 of the 48 markets surveyed.

Strange ideas have always existed. What a society also used to have was a referee: a handful of institutions most people would let settle an argument. The broadcaster, the paper of record, the doctor, the court, the official inquiry. You did not have to like them. You only had to accept that when they said something had happened, it probably had.

For much of the twentieth century that arrangement was helped along by scarcity. Britain had a few television channels and a small number of national newspapers. On Christmas Day 1986, around 30 million people watched the same episode of EastEnders. A country that watched the same programmes and heard the same bulletins had, at the very least, the same events to argue about.

But a shared account is not the same as a true one, and the referees got some of the biggest calls wrong. After Hillsborough in 1989, the police and a national newspaper blamed Liverpool fans for the disaster. It took 27 years, and families who refused to accept that version, before an inquest jury found in 2016 that those who died had been unlawfully killed and that the fans' behaviour had not contributed.

The Post Office prosecuted more than 900 sub-postmasters on the evidence of its Horizon computer system, insisting the software was reliable while the people using it said the money had never gone missing. Many were convicted and some were sent to prison. It took two decades, and a television drama, for the country to properly take notice.

In 2018 it emerged that the Home Office had been treating members of the Windrush generation as illegal immigrants. Many had come from the Caribbean as children and had lived here legally for decades. People lost jobs, homes and NHS care, some were detained, and at least 83 were wrongly removed from the country. The landing cards that could have shown when many of them arrived had been destroyed in 2010. An independent review later found the whole thing had been foreseeable and avoidable.

None of these was a conspiracy theory. Each was an institution defending its version of events against the people who had lived through them. The old model of shared truth relied on most people never seeing inside the building. The internet let us in, and it made every failure searchable and permanent.

So some of today's distrust is not irrational. It was earned, and it was never earned evenly. Communities that have been disbelieved by the institutions meant to protect them are often the first to be accused of falling for misinformation, when what they are drawing on is memory.

Knowing is not the same as believing

Distrust explains why the old referees lost their authority. It does not explain why nobody has replaced them.

Most of what any of us knows, we know second-hand. You have probably never personally confirmed that the Earth orbits the Sun, run a clinical trial or counted the votes in an election. You accept those things because you trust a chain of people who did. Knowledge is information. Belief is a decision about whose word to take.

That decision has never been neutral. We tend to believe people who look and sound like us, and that is true of everyone. In a 2019 trial in Oakland, California, Black men randomly assigned a Black doctor were 18 percentage points more likely to take up preventive care, which the researchers put down to trust and communication. The real damage comes when one group holds the jobs that decide who is believed. A 2016 study of white medical students and residents found that about half held at least one false belief about biological differences between Black and white people, such as the idea that Black people have thicker skin, and those who did rated Black patients' pain as lower. The philosopher Miranda Fricker calls this testimonial injustice: giving someone less credibility because of who they are. Long before the internet, some people's word was worth less.

What the internet changed was the range of people we can choose to believe. It multiplied information without giving anyone more hours to check it, so we lean even harder on whom we trust. For most of history that choice was limited by geography and a few outlets. Now it is almost limitless. For nearly any position you want to hold, there is an expert, a study, a clip of a doctor or a thread full of links ready to back it up.

That is why "do your own research" so rarely ends with somebody changing their mind. In a system where every conclusion has a supporting source, research tends to lead back to where you started. More information was supposed to settle arguments. Mostly it keeps both sides supplied.

The internet did not make us worse at knowing things. It made agreeing on them optional.

Falsehood has a head start

The places where we now find information are not neutral about which version wins. In 2018, three MIT researchers studied around 126,000 stories shared on Twitter by about 3 million people between 2006 and 2017. They found that false news travelled further, faster and more widely than the truth in every category, and that false stories were 70% more likely to be retweeted. Taking the bots out made no difference. People were doing it.

The researchers' explanation was novelty. False stories were more surprising and more likely to provoke disgust. The truth tends to be duller, and "the figures were broadly correct" has never gone viral.

This matters more now than in 2018, because the feed has become the front page. The 2026 Reuters report found social media and video networks are now the most widely used source of news worldwide, used weekly by 54% of people, ahead of TV at 52% and news websites at 51%. Most people now meet the news in the place where false stories travel best.

Trust is moving too, from institutions to individuals. News creators and influencers are now part of how people make sense of events. Someone who no longer trusts the broadcaster may still trust a person talking to them directly from their kitchen. That is not automatically a bad thing, but it changes what credibility rests on. Increasingly, we judge a source by how they make us feel rather than by how they came to know what they are telling us.

When everything could be fake

Anyone who takes pictures knows a photograph has always lied a little. You choose the frame, the moment and what to leave out. But for most of its history a photograph was at least evidence that something had been in front of the lens. Generative AI has made that link optional. An image can be flawless and have no event behind it at all.

The obvious fear is that people will believe fake images and videos. The less obvious one may be worse. In 2018 two law professors, Bobby Chesney and Danielle Citron, described what they called the liar's dividend. Once the public knows convincing fakes are possible, anybody caught on camera can claim the real footage is fake. The risk is not only that we believe things that never happened, but that we stop believing things that did.

That cost would not fall evenly either. For years, Black people's accounts of police violence were easily dismissed, until phone cameras made them hard to deny. When George Floyd died in Minneapolis in 2020, the first police statement described a "medical incident". A teenager, Darnella Frazier, had filmed what actually happened, and her video became central to the trial of the officer who knelt on his neck. She later received a Pulitzer special citation. The phone gave the people least likely to be believed something close to proof. The liar's dividend threatens to take it from them first.

The end point is not a public that believes everything. It is a public that believes nothing it does not already want to believe. Cynicism can pass for intelligence, but it works much like gullibility: both let you skip the evidence and go straight to the conclusion you prefer.

Nor is there an obvious new referee waiting. More people are asking AI chatbots about the news, but the same Reuters survey found only 20% trust the answers.

It is not happening everywhere

If the internet alone had destroyed shared truth, trust would be falling at the same rate wherever people are online. It is not. In the same survey, 68% of people in Kenya and Nigeria said they trust most news most of the time, and Finland led Europe at 63%. These are connected countries, with smartphones and social media like anywhere else.

The Reuters Institute notes that falling trust in news is hard to separate from falling trust in government, driven by political instability and divisive elections. That is the more uncomfortable conclusion. The internet amplifies whatever is already happening between a public and the people meant to serve it. Where institutions have kept faith with people, trust can survive the feed. Where they have not, the feed makes the damage impossible to hide.

Trust used to be given. Now it has to be earned.

We are not going back to a few channels and a country watching the same Christmas episode, and we should not want to. That version of shared truth left sub-postmasters ruined, Hillsborough families called liars and people who had lived here for decades told they did not belong.

What the internet has ended is trust by default. For most of the last century, belief was handed automatically to whoever held the microphone. Now everyone has one, and belief has to be earned in public, source by source and correction by correction. Institutions that show their workings, admit mistakes quickly and let people check for themselves will be believed more than those that simply ask to be.

For the rest of us, the most useful question may no longer be "is this true?" but "how would I know?" Who is telling me, how did they find out, and what would change their mind? It is slower and less satisfying than certainty. It is also how knowing turns back into believing.

The internet did not take shared truth away from us. It showed us how much of it we had been taking on trust.