Your customer does not want to trust you, they want to check

On 16 August 2026, Anthropic CEO Dario Amodei said that public rejection of artificial intelligence is, at bottom, a crisis of trust. Not fear of robots, not science fiction: people simply do not believe that companies, governments and the tech industry are on their side. He went further and admitted that the fairest criticism of the sector is that it has not delivered what it promised.

The number that comes with that statement is uncomfortable: a Gallup survey from March 2026 found that 71% of Americans oppose an AI data centre being built in their own community. This is not a debate about model alignment. These are people who heard too many promises and saw too few results.

You do not have a data centre. But the mechanism that broke there is exactly the one that decides whether a customer signs with your company.

The same crisis, at your company's scale

Think about how you are presented today. "We are a market reference." "Excellence in service." "Tailored solutions for your business." "We use artificial intelligence." Every one of those sentences is a promise, and none of them can be verified by the person reading.

Customers have learned to discount them. They have already hired someone who promised excellence and vanished, already paid for a "tailored solution" that was an off-the-shelf template, and already heard "we use AI" from someone who only had an automated form. When every company says the same thing, the sentence stops informing and becomes noise — and the buyer starts deciding another way.

The way out is not to promise better. It is to make the promise unnecessary

The best formulation I have read on this is short: make trust unnecessary. Instead of asking people to believe you, give them the means to check.

That changes the question you ask when writing anything about your company. It is no longer "how do I say I am good?" but "what do I show so the person concludes it on their own?". It is a shift of effort: it takes more work and pays far more, because proof does not compete with adjectives — it ends the argument.

The four kinds of proof that work

1. The open number. A result with the source, the period and the method in plain sight. Not "we greatly increased sales", but "from this to that, between these months, measured in this tool". An exact number is more credible than a round one, and a stated period is what separates data from folklore.

2. The demonstration. Let people use it before buying. A free tool, a sample of the work, a test with no sign-up. Whoever tries it does not need to believe you — they have seen it work. It is the most expensive proof to produce and the one that converts best.

3. The verifiable third party. A testimonial with a name, a company and a link. A review on a platform you do not control. A case the client will confirm if someone asks. Anonymous praise proves nothing — and the reader knows it.

4. Reversibility. A contract with no lock-in, a small first stage, a short window to walk away. When you lower the cost of being wrong about you, the decision stops requiring trust. It is the proof that is not about you: it is about the buyer's risk.

What no longer works

A generic badge in the footer. A testimonial with no full name. "Over a thousand clients served" without saying since when or of what. An award nobody has heard of. A stock photo of a smiling team. And worst of all: the list of adjectives any competitor could copy and paste without changing a word.

A quick, cruel test: take the sentence you introduce yourself with and swap your company's name for your main competitor's. If it still makes sense, it is proving nothing.

The test at home

We apply this here, and it is worth telling how — including the part that stings. We publish our own site's numbers — impressions, clicks, click-through rate and position — always comparing two equal periods, with the tool and the dates stated.

And we publish the bad number too: in one month, more than twelve thousand impressions and a hundred-odd clicks produced a single real enquiry. We could have shown only the rising curve. Showing the hole alongside it is what made the rest verifiable — because anyone publishing only the good parts is advertising, and readers spot that in two seconds. We wrote that whole story in our piece on what worked and what did not on our Google.

How to build your proof this week

You do not need a project. Pick one service and answer four questions in writing: what result it produced last time, with which number and in how long; which client would agree to confirm it if asked; what a person can try before paying; and what the smallest possible first step to hire you is.

The answers to those four questions, written with a name and a number, are worth more than a whole page of adjectives. And if you cannot answer one of them, you have found something more important than a new piece of copy: you have found what the service is missing.

The honest limit of all this

Proof does not fix a bad service — it only speeds up the discovery. If the work does not deliver, transparency merely makes the client notice sooner, and the bill arrives earlier. What proof does is clear away the suspicion hanging over people who do good work and are being measured against the market average.

That is why big tech's turn and the corner bakery's turn are the same turn. Nobody gets the benefit of the doubt any more for being large, or for being likeable. It goes to whoever shows.

If you want to turn this into how your company presents itself — on the site, on social, in the proposal — that is what we handle in digital marketing.

Source

The edition "Tech's Trust Problem Is a Product Problem" from Peter Diamandis's Metatrends newsletter (23 August 2026), which brought the subject and the phrase "make trust unnecessary". Dario Amodei's remarks on the crisis of trust were reported on 16 August 2026 by TechCrunch and other outlets; the 71% survey is Gallup, March 2026. The small-business angle and the four kinds of proof are ours.