Approach

Four commitments, not four adjectives.

Anyone can put the word “trustworthy” on a website. These are the four things we actually do, written so you can tell whether we have done them — and they hold for all six products, whatever each one is for.

Evidence first

A product shows where what it tells you came from, or it says plainly that it does not know. Where the ground is thin, it says so rather than rounding up to something that sounds more useful.

This is an architectural choice, not a tone of voice. The decision about what is true is moved out of the model and into a step you can inspect.

In practice

Telli 5 and avenTellix open an answer into its graded sources, claim by claim. avenTelligence shows the demand and capacity behind a recommendation. TradeIntellix shows the mechanics behind a fill, and gives no advice at all.

Methods published and versioned

The rules a product works by are written down, numbered and dated — whether they produce a score, a ranking or a recommendation. When they change, the version changes with them, and the change is recorded where you can read it.

Nothing changes quietly. A score you saw last month was produced by rules you can still go and look at.

In practice

Telli 5’s Evidence Checker is at version 2, live since 30 October 2026. Its method in full — the five bands, the per-claim verdicts and the arithmetic behind the score — is published on Telli 5’s own site.

Read Telli 5’s method at tellifive.com →

Safety before shipping

Crisis replies and clinical answers are reviewed against a written standard before they go out — by people, one at a time, not by a sample. In TradeIntellix the same rule means it never recommends a trade; in avenTelligence it means no scheduling change is made on a patient’s behalf.

Medicine comes last and with legal advice first. A stale pointer does real harm there in a way it does not elsewhere.

In practice

Nothing reaches a tester, let alone a customer, until the safety review for that release is signed off.

Your data is not the product

We do not sell your data, to advertisers or to brokers. There is no advertising in any of our products and no advertising identifier in any of our apps.

That is a business-model decision, which is the only kind that holds. We would rather be paid by the people who use our products than by anyone who wants their attention. If that ever has to change we will say so here before it ships, and nothing paid will be presented as a result.

In practice

Our apps ship without an advertising identifier at all, which is checkable in the store listing.

Privacy notice →

What every product is held to

These six hold across the family, whatever a product does. How each one meets them is its own business — and its own published method.

It shows its working

Sources behind an answer, data behind a recommendation, mechanics behind a result. You can always see where it came from.

It says when it doesn’t know

Thin ground is called thin. No product rounds up to something that sounds more useful than it is.

Its method is published

Written down, numbered and dated. When the rules change, the version changes with them.

No one can pay

Not to change what a product says, which sources it trusts, or how results are ordered.

Never its own marker

Whatever makes a decision is never the only thing that checks it. Where a product writes an answer, its claims are checked by a model from a different maker.

Nothing is called finished

Not before it is. Private pilot means private pilot. In development means you cannot use it yet.

How we treat sources

Grading a source is where most of the honesty in a system like this lives. Wherever a product relies on sources, these are the rules it grades them by — the same rules in every product that has them.

Opinion is shown as opinion. It is never presented as fact, however confidently it was written.
We cite academia, medical journals, reputable educational sources and established reporting.
Ten copies count once. Where the same author or authority sits behind every version of a claim, that is one voice repeated, not a consensus.
Retracted papers are removed, and where we can see influence over a source, we leave it out rather than discount it.