We have spent centuries learning to question authority.
Religious leaders. Governments. Institutions. Experts. The media.
We have learned that authority can be wrong, biased or self-interested. We have built checks and balances around it. We have asked who made the decision, what evidence supports it and whose interests are being served. But now a new form of authority is emerging. And we rarely question it.
Algorithms.
The authority we rarely question
Not because machines are suddenly becoming gods. The danger is much more ordinary. They are becoming so useful, so convenient and so embedded in everyday life that we increasingly accept their answers without asking how they arrived at them.
An algorithm tells us what to watch. A search engine determines what we see first. A recommendation system suggests what to buy. A social platform decides which stories enter our attention.
And now an AI assistant increasingly tells us what happened, what something means, who is credible and, sometimes, what we should believe. The answer arrives before we have even thought to question the question. When the algorithm starts feeding the belief. There is something else happening that may be even more consequential.
Algorithms don’t just give us information. They learn what information we respond to and give us more of it. We click on something because it interests us. The system notices. It serves us more of the same. We engage again. The system becomes more confident about what holds our attention.
What we choose shapes what we are shown. What we are shown shapes what we choose next. And gradually, the information environment around us can become narrower, more predictable and more reinforcing.
This is how a question can become a belief system.
Consider vaccination misinformation. Someone might begin by watching a single video because they are curious or uncertain. The platform identifies the interest and serves up more related content. They watch another. Then another.
Soon, they are not simply being exposed to one argument. They are being immersed in an information environment in which that argument is repeatedly reinforced.
The more they engage, the more the system learns. The more the system learns, the more it serves.
When repetition begins to feel like evidence
The more familiar a claim becomes, the more credible it can begin to feel. This isn’t confined to vaccination. The same dynamic can operate around politics, conspiracy theories, financial advice, climate change, celebrity culture and even the reputation of a company.
The problem is not simply misinformation. It is the feedback loop.
What we choose → what the algorithm learns → what it shows us → what we engage with → what it shows us next.
Eventually, the system may not need to persuade us. It simply needs to keep showing us what we already believe. That is where the comparison with religion becomes interesting.
Algorithms as invisible authority
Religion has traditionally provided more than belief. It has provided authority, interpretation, guidance and a framework for making sense of the world.
Algorithms can increasingly influence all four. The difference is that we can see the priest. We can read the doctrine. We can challenge the institution.
The algorithm is largely invisible. We often don’t know what information an algorithm has consumed, what assumptions are embedded within it, what it has excluded, or why it has produced one answer rather than another.
And yet we increasingly trust the output. That creates a very different kind of risk.
When false information looks authoritative
In 2023, ChatGPT falsely accused US radio host Mark Walters of embezzlement when generating a summary of a lawsuit that did not contain the allegation. Walters subsequently sued OpenAI.
The unsettling part was not simply that the system was wrong. It was that the answer looked authoritative. During conflicts including the 2025 Israel-Iran crisis, manipulated and AI-generated images and videos also circulated as authentic footage, demonstrating how quickly synthetic information can become part of the public record.
Again, the problem isn’t simply that false information exists. It always has. The problem is that technology can now produce, package and distribute it at a scale and speed that makes verification increasingly difficult.
And there is an important distinction here. Algorithms don’t necessarily have to convince us of something that isn’t true. They can simply make alternative explanations harder to encounter. That may be more powerful.
Because a person who hears one argument and rejects it has exercised judgement. A person who hears the same argument a hundred times from what appear to be different sources may eventually experience repetition as evidence.
When the crisis never really ends
There is another consequence that matters enormously to organisations: a crisis can end long before its digital life does.
Consider a major corporate controversy. The immediate news cycle eventually moves on. Executives change. Boards change. Customers move on. The organisation may have spent millions rebuilding trust.
But the information does not disappear. It remains in news archives, search results, social media, forums, databases and increasingly in the material that AI systems use to construct answers.
Ask a search engine what happened. Ask an AI assistant whether the organisation can be trusted. Ask it about the company’s history. The answer may be shaped by a crisis that happened years ago.
The crisis may have moved on. The digital record has not.
A new frontier of reputation management
This changes the nature of reputation. For decades, organisations have largely thought about reputation as something held by people: customers, employees, investors, regulators, journalists and communities. Increasingly, organisations also have to think about how they are represented by systems.
What does the internet say about us? What does a search engine surface? What does an AI system infer? What happens when those systems are wrong? And, perhaps most importantly, who is responsible for correcting the record?
The Air Canada case illustrates one side of this emerging problem. In 2024, an Air Canada chatbot provided a customer with incorrect information about the airline’s bereavement fare policy. When the customer relied on that information, Air Canada initially argued that the chatbot was responsible for the error. The court rejected that argument, finding the airline responsible for information provided by its chatbot.
The significance extends well beyond airline ticketing. An organisation can increasingly be held accountable for what an AI system says on its behalf, even when the organisation did not intend to say it.
But there is another, less obvious dimension to the problem.
Organisations are also being interpreted by AI systems they do not own or control. An employee, prospective customer, journalist, investor or regulator may ask an AI system a simple question: What is this organisation’s reputation? Has it been involved in any controversies? Is it a trustworthy company?
The answer may be assembled from years of news coverage, corporate websites, regulatory decisions, social media, reviews and other material available online. Some of that information may be accurate. Some may be incomplete, outdated or taken out of context. The system may then synthesise it into a confident narrative that no human being has ever actually written.
This creates a new challenge for reputation management.
The issue is not about removing unfavourable information from the record. Organisations should be accountable for their actions, and legitimate criticism should remain part of the public record. The issue is whether the information from which future judgments are formed is accurate, contextualised and representative.
The challenge is therefore no longer simply managing the news cycle. It is managing the information environment in which future audiences, customers, employees, investors and machines will encounter the organisation.
For boards and executives, these are no longer simply technology questions. They are questions of governance, risk and trust.
It is no longer enough to respond to the crisis in the moment. Organisations need to understand the information ecosystem that surrounds them: what narratives are being created, what information is being indexed and amplified, what inaccuracies are gaining traction and how those narratives may be interpreted by both humans and machines.
Because the next person making a decision about your organisation may not begin by asking a colleague.
They may ask a machine.
The question we should keep asking
That is why the answer to the question “Are algorithms the new religion?” may ultimately be less important than the question behind it. What happens when we give something authority simply because it gives us an answer? We should not stop using algorithms. We should not stop using AI. And we certainly should not retreat into nostalgia for a world before technology mediated our lives.
But we should retain one very old human habit.
Questioning.
Question the answer.
Question the source.
Question the assumptions.
Question what is missing.
Question who benefits.
And question the authority we are granting to systems whose workings we cannot see. Because if algorithms are becoming one of the most powerful sources of authority in modern life, perhaps the oldest question remains the most important.
Who is writing the scripture?