Articles

The inflation of digital trust

The inflation of digital trust

Imagine a business leader who requests a proposal for a new system from two IT service providers. Both documents are carefully written and professional: they contain a detailed architecture, a schedule and a reference list, and both come with a demo that looks like it works. Ten to fifteen years ago you could read a lot out of such a package, because producing it took serious work and knowledge.

The signals that underpin digital trust are losing their evidential power: behind one proposal there are years of operational experience, while the first version of the other could have been assembled by a language model in a few minutes — and on paper the buyer can no longer reliably tell them apart.

In my view this is one of the most important changes in business life, and we still talk too little about it. This article examines how digital trust got inflated: why a good appearance proves less today, what international trust surveys measure, and what rebuilds trust where signals are no longer enough — a running system, a callable reference and accountability by name.

When appearance comes for free

In the earlier phase of the internet spreading into business use, having your own website, a tidy online presence or a company email address could in itself be a distinguishing signal. A professional blog or a detailed service presentation counted as a strong signal because producing them still took significant work, knowledge and time — so it was reasonable to assume that performance stood behind it.

Generative AI loosened this relationship: a company introduction, a professional text, an outline of a technical proposal or the first version of a product demo can be produced in minutes. With the same tool, real expertise and its appearance can both be produced.

The Stanford AI Index, published in April 2026, relying on the McKinsey survey, reports that 88% of respondents say their organisation used AI in at least one business function in 2025 (78% in 2024), and that 79% of organisations using AI use generative AI regularly in at least one business function (71% in 2024).

The report itself notes that these are self-reported, indicative figures. Even so, they clearly show that the technology has become part of the everyday working of the economy.

The consequence is direct: professional appearance and good communication in themselves are already a less reliable proof of the underlying competence.

Trust does not stop at the signals

Alongside the decline in the evidential power of signals, trust in the digital environment has become more cautious. The measurements below measure different things, so it is worth looking at them separately, and none of them proves the whole picture on its own.

The 2026 edition of the Edelman Trust Barometer was conducted in 28 countries with 33,938 respondents. Trust in technology companies fell from 76% to 73% between 2025 and 2026, averaged across the 23 markets surveyed. This is a moderate decline: alongside education, the value remains the highest among the sectors. Trust did not collapse; it weakened.

According to the survey by KPMG and the University of Melbourne, covering 47 countries with 48,340 respondents (November 2024 to January 2025), 54% of people are ambivalent about or unwilling to trust AI systems. In advanced economies 39% are willing to trust them, in emerging economies 57%.

Comparable data from the 2022, pre-ChatGPT survey exists for 17 countries: the share of those who consider AI systems trustworthy fell from 63% to 56%, those willing to rely on them from 52% to 43%, while those worried about them rose from 49% to 62%. The two surveys have different respondent samples. The researchers’ interpretation is that this probably reflects, in part, more considered trust: with use, people have also learned the limits of the tools.

The information environment as a whole has become more uncertain

News is a separate area: the following data relate to journalistic content, so they do not directly prove that trust in business materials made with AI has also fallen. They do, however, show well that the information environment as a whole has become more uncertain.

The Reuters Institute’s 2026 Digital News Report was conducted across 48 markets with nearly a hundred thousand respondents. General trust in news fell to 37% after three stable years at 40% — the lowest value since measurements began in 2015, with a significant decline in 29 markets.

According to the report, one reason is that more and more people get their news through channels considered less trustworthy: 22% trust news received via social media, and 20% trust news received through AI chatbots. At the same time, trust in the most used news brands holds up better than the general figure, and the share of those worried about online misinformation rose by 4 percentage points to 62%.

In June 2025 the Pew Research Center asked 5,023 American adults; the results were published in September. 76% of respondents say it is extremely or very important to be able to distinguish images, videos and text made by AI from those made by humans — but 53% are not too confident or not at all confident that they could recognise this themselves.

The data relate to the United States and AI content in general, not specifically to business materials. People, therefore, want to know what they are looking at, but increasingly doubt that they can establish it. Taken together, the measurements show that trust in digital information has become conditional, and verifiability plays an ever greater role.

The buyer is fast, but wants verification

The trust problem also shows in B2B purchasing. Gartner asked 645 buyers between August and September 2025 and presented the results in May 2026.

70% of buyers prefer fully digital, self-service purchasing, and 67% prefer buying without a salesperson. 45% used generative AI in a recent procurement, primarily to gather data on vendors and products — while 69% would rather have the information provided by AI verified by a salesperson.

Trust in either source is unconditional for no one: 51% say misleading information is more likely from generative AI, 49% say it is more likely from a salesperson.

The buyer therefore chooses the digital route for speed, but does not entrust the risk of the decision solely to automated systems: they look for someone who can verify and who takes responsibility.

In the IT market, provable performance is the competitive advantage

The buyer of IT services traditionally could not judge the technical details, so they relied on references, communication and reputation. At the same time, the IT sector is not merely affected by the inflation of digital trust: it is able to give concrete, tangible proof of its expertise.

Today a well-written proposal and an impressive prototype are far easier to obtain, while the essential elements of responsible engineering work — modelling business processes, a secure architecture, live operation and long-term maintenance — still require professional decisions and taking responsibility. That is precisely why a running system, a real reference and professional accountability prove performance in a tangible way.

Alongside AI-accelerated development, architectural thinking, understanding the business problem and taking responsibility for a working solution become even more important — what counts in the buyer’s decision is not the technological tool or the impressive demo, but the actual result.

The profession itself does not blindly trust its tools. In the Stack Overflow 2025 developer survey (more than 49 thousand respondents, 177 countries), 46% of developers said they do not trust the accuracy of AI tool output — a year earlier it was 31%.

In the 2026 survey, 14,304 people answered the trust question: 48% trust AI output when they can easily verify it, and only 6.6% trust it in many tasks, including important workplace decisions. Another question on verification habits received 13,162 answers: 76.5% of developers run the response locally before using it, and only 9.6% use it without modification.

The questions of the two years differ, so the values cannot be compared directly. Trust in this profession depends on verifiability — and in my view the same is true of the relationship between buyer and provider.

What rebuilds trust?

If signals are not enough, trust needs a foundation that cannot simply be generated. In my view this means three things; all three are verifiable, and all three take time.

1. A running, deployed system: one that can be looked at — not a screenshot and not a presentation file, but a live solution whose operation can be verified.

2. A callable reference: a buyer who can be phoned and who tells in their own words how the collaboration works.

3. Accountability by name: someone who stands behind the result with their own name.

A personal meeting is valuable in this situation because there the buyer can ask, ask back, and see how the other thinks about a real problem.

Personal presence in itself is not proof of reliability, but it gives a more direct insight into professional thinking, attitude and competence. This is feedback that mass-automated communication can hardly replace.

This is why I believe the role of professional communities, recommendations and provable performance is rising — as a complement to digitalisation.

Credibility has become the scarce resource

Back to the two proposals: the paper still will not show the difference. The buyer will therefore look for the provider that shows more than what is on paper: a running system, a callable reference, and a person who answers for the result with their own name.

With AI, an almost unlimited amount of content can be produced, but the quantity of content is not the same as its value. In the age of information, credibility has become the scarce resource: the advantage goes to those who do not communicate the loudest, but can prove that they are able to deliver what they promise.

Related articles