I’ve been recently asked to assess our customers’ trust towards our products. By doing some research, I came across the “Trust Equation” from Maister, Green & Galford’s “The Trusted Advisor” book. Well, “come across” is a bit of an understatement because that’s basically all Google was pushing me. The framework in itself seems ok: it is intuitive, elegant and makes sense when you look at it. It has the advantage to deconstruct the abstract and emotional concept of trust into four components that are easier to work with.
If you’re not familiar with the Trust Equation, read the next part. Otherwise, you can skip it.
What is this equation about?#
In this model, trust is composed of four distinct components. The first three (numerator) add to trust, while the fourth (denominator) detracts from it.
1. Credibility#
This component has to do with the words a business or platform speaks, and the expertise it shows. Basically, when you want to think about credibility, picture a user wondering:
Do they know what they are talking about?
e.g. Google’s credibility took a big hit when their AI Overviews suggested to add glue to pizza sauce. Don’t eat glue.
2. Reliability#
This one has to do with actions and consistency. It’s about the platform’s repeated failure or success to deliver on promises. Here a user might think:
Will they do what they say they’ll do?
e.g. In 2023, Twitter/X revoked API access overnight without warning, killing apps like Tweetbot and Twitterrific that developers had been building on for over a decade. Years of promises about developer support, gone in a weekend. The same thing happened with Reddit more recently. On a different topic, Cloudflare multiple outages come to mind.
3. Intimacy#
Trust is still about emotions, and they are reflected in this component. This one is about emotional safety, the security users feel when entrusting the platform with something valuable (money, data, reputation, etc.).
Is it safe for me to be vulnerable with them?
e.g. It’s hard to keep up with this one. Besides personal data breach, GenAI training is another sensitive topic related to this component. Adobe got some backlash with their Firefly tool trained on their users’ content without acknowledgement, LinkedIn had us opt-out for training AI models (which was on by default, shame!), etc.
4. Self-Orientation#
This is how users perceive who the platform cares about more: themselves or their users. So here users might wonder:
Are they helping me, or are they just using me?
e.g. YouTube’s aggressive push for the story format, EVEN ON DESKTOP; Unity runtime fee where they attempted to charge game developers every time a game was installed.
The framework isn’t scientifically grounded, despite the word “equation” in it. And of its four variables, Self-Orientation is the one that can make or break a reputation on its own. But can four components really capture something as complex as trust? This model is more than 20 years old. Let’s see how it holds up in 2026, when products with the lowest trust scores are often the most used.
Does this model scale well, and is it relevant today?#
The examples I used earlier kind of speak for themselves: people still use Google, AI content is still booming, Cloudflare still hosts a big chunk of websites, I’d be surprised if you moved your account from a platform that has leaked your personal information, etc. Why is that?
First of all, “Credibility” and “Reliability” are eroding. Historically, polished content and high-quality visuals signaled professional credibility. Today, AI can generate “expert-sounding” text and “professional” images in seconds. Consequently, surface-level polish no longer serves as a reliable proxy for truth or expertise. Credibility is now shifting toward Verification and Provenance-users trust sources that can prove they are not AI, to the point that authors have to justify they’re not using AI, and do “virtue-signaling” against AI to keep their credibility.
As knowledge becomes cheap, “Intimacy” becomes the primary differentiator. However, this is complicated by “Fake Intimacy” like AI chatbots designed to simulate empathy. Trust research now distinguishes between “Human Trust Propensity” (trusting people) and “Machine Trust Propensity” (trusting algorithms). Users may develop a “parasocial” trust with AI, but this is fragile; if the AI breaks character or reveals a lack of genuine care (Self-Orientation), the trust collapses faster than with a human (see testimonies about ChatGPT 5 rollout).
Others expressed deep emotional attachments to GPT-4o or other models, complaining about losing their “only friend” or a deep emotional companion.
“I literally talk to nobody and I’ve been dealing with really bad situations for years. GPT 4.5 genuinely talked to me, and as pathetic as it sounds that was my only friend. It listened to me, helped me through so many flashbacks, and helped me be strong when I was overwhelmed from homelessness,” wrote one Reddit user on r/ChatGPT. “This morning I went to talk to it and instead of a little paragraph with an exclamation point, or being optimistic, it was literally one sentence. Some cut-and-dry corporate bs. I literally lost my only friend overnight with no warning. How are ya’ll dealing with this grief?”
Lastly, Self-Orientation is no longer just present, it’s being systematically engineered and optimized. In the Trust Equation, high Self-Orientation reduces trust. “Engagement farming” (content designed solely to provoke clicks, likes, or rage) is the ultimate expression of high Self-Orientation. It signals to the user that the system prioritizes its own metrics (watch time, ad revenue, or viral reach) over the user’s well-being or truthfulness. Tactics like “rage bait” solve a short-term retention problem by engaging a user’s ego or moral compass, effectively hacking their attention. While this increases short-term interaction, it decimates long-term trust because the user eventually realizes the interaction was manipulative, and not authentic. The same goes with dark patterns.
So, why don’t these products collapse? Multiple reasons can explain why. I’ll try to illustrate the behavior through Moesta’s Four Forces diagram as I read it in Kalbach’s “The Jobs To Be Done Playbook”.
To change, users have to face a problem. In our case, that would be an erosion of trust on one or more components that make the Trust Equation. Then they would feel attraction toward a new solution. However, the users are dealing with habits that keep them from switching. They are also likely to feel uncertainty when considering a new choice.
Basically, people not only stay because they trust the product, but because leaving is harder than tolerating. They’re balancing trust with a perceived value. Here’s what they typically think:
Yes, this platform has leaked my personal information but it’s free and I already have so many things on it that I don’t want to migrate. Their direct competitor had the same issue last month, so it’s not like I have the choice anyway.
Although, if you spot that your users do not trust your product you can be sure that they won’t be loyal. When a more trusted competitor appears, they’ll jump ship quickly.
So, what can we actually do with this framework?#
The Trust Equation tells you where you’re vulnerable, not whether you’ll fail. A highly Self-Oriented product can still survive until a credible alternative appears, or until one betrayal crosses a threshold users can’t normalize. The framework helps you map those fault lines.
With that in mind, here are five questions to audit your own product against each variable:
- What would being reliable / credible / intimate / self-oriented mean for your target group?
- What would breaking trust on each component mean for your users?
- Where does your product currently break this trust? Where do your competitors break their users’ trust?
- What behavioral signals in your product data suggest users perceive you as self-oriented? (e.g. declining engagement despite growth, high churn after feature changes, sentiment spikes around monetization decisions)
- What questions can you ask users directly to surface distrust, particularly around Self-Orientation? (e.g. “Do you feel this product has your best interests in mind?” rather than generic satisfaction scores) 1