Field Note 03: How trust, authenticity and personalisation pull against one another
Summary: A newsroom built for the AI era holds three forces in tension — trust, authenticity and personalisation — each pulls against the other two. The task is to find the configuration in which all three balance.
In the first issue I called Newsroom 5.0 a railway rather than a better canal. A railway is only as good as the principles it runs on, and a newsroom built for the AI era runs on three of them, held in constant tension.
Trust, authenticity and personalisation — the three points of what I call the TAP Triangle — are three ingredients 21st century journalism must deliver. Too much of one ingredient sours the dish. Too little of another makes it equally unappetising. The aim is to find a balance where the flavour of each one complements the others.
How the three points compete
Personalisation competes with trust. When algorithmic selection narrows what a reader sees into only what they already believe, trust in journalism as a source of shared, verified reality declines. But personalisation can also strengthen trust: when someone who prefers to consume information visually receives the same verified reporting as someone who prefers text, comprehension of an issue improves. This is what personalisation is for.
Personalisation competes with authenticity too. At scale, when automated systems mediate the relationship between newsroom and reader, they can manufacture the appearance of personal connection without its substance. If readers believe they are forming a relationship with an individual journalist when they are in fact engaging with an algorithm, the relationship rests on a false foundation. When that becomes clear, as it always does, trust collapses more completely than if nothing personal had been promised.
Authenticity has two elements. Editorial authenticity is the transparency of the journalistic process — reporting kept separate from opinion, clear sourcing, honesty about what is known and what is not. Emotional authenticity is the human register: the sense that a real person stands behind the writing. The institutional distance that traditionally built trust — professional objectivity, editorial gatekeeping — systematically suppresses the second. The resolution is not to choose between trust and authenticity, but to separate authenticity’s two elements: editorial authenticity protected without exception, because it is trust’s ally; emotional authenticity cultivated deliberately, through journalists who are visible as people and a voice that does not mistake distance for credibility.
What separates journalism from a platform
Underneath the competing flavours of trust, personalisation and authenticity is what separates journalism from a platform. A platform asks what will keep this person engaged. A journalism product asks what this person needs to understand their world and act on it, delivered in a way they can use.
The first holds attention. The second leaves the reader informed and equipped. A reader who spends longer on the site is not automatically better served. But a reader who acts on what they have read is.
Putting the triangle to work
The TAP Triangle earns its place when it is used as a forward test — applied at commissioning, at the point of strategic decision, and before an AI tool is deployed, rather than as an explanation produced once the consequences are already visible.
It starts with the portfolio, because different kinds of journalism weigh the three points differently. Civic journalism (government, policy, elections) has a different equilibrium of the three forces than practical journalism (personal finance, consumer advice), cultural journalism(reviews, critiques, experiences) or situational journalism (news updates, weather, markets). Where one demands trust as the dominant dimension, others require more emotional authenticity or stronger personalisation.
The TAP Triangle does not solve those problems. It makes them visible before the investment is made rather than after. For AI specifically, the triangle gives editors two tools. The first is governance. Automation changes a content type’s configuration in predictable ways, so for each application the questions are plain: which point dominates here, what does automation do to it, and what level of human oversight does that require? Civic journalism demands the most stringent oversight, because trust dominates and an AI error there carries the highest cost; situational content is the most amenable to automation, for the same reason inverted.
The second tool is design. The distinction between platform personalisation and journalistic personalisation is not a philosophical position. It is a product brief: it tells the team building a recommendation engine which signal counts as a successful outcome, and which engagement behaviour should trigger an editorial review rather than be counted as a win.
Used this way, the TAP Triangle answers one question and one only: what values must the system protect? It does not, by itself, tell you how closely to watch any given tool — a weather summary and an election briefing can run on the same technology and demand entirely different scrutiny. That is a second question, and it needs a second instrument. We’ll talk about that in a future Field Note.he hierarchy is the map. What follows is the territory.
FROM THE FIELDS – TWO THINGS WORTH READING
The strategies meant to keep newsroom AI safe are coming apart in practice.
The strategies meant to keep newsroom AI safe are coming apart in practice. A new study from Aarhus University, presented at CHI 2026 in Barcelona, studied how ten journalists across four Scandinavian media companies actually use generative AI. All four had pursued the same three strategies — build an in-house tool, write ethical guidelines, run AI training — and all three came apart on contact with the work. The journalists preferred ChatGPT to the in-house tools, one company having already scrapped its own after concluding ChatGPT did the job better. Five of the ten could not say what their guidelines contained; some who did know feed material into ChatGPT in ways those guidelines did not permit. Even journalists who had taken the training could describe how a language model works only in the vaguest terms.
This is the clearest field evidence yet for the argument the first issue made: a policy cannot do the work of an architecture and an operating system. Guidelines that journalists do not know, do not rate, or quietly route around give an organisation the appearance of governance while leaving the practice ungoverned. And the failure is structural, not a matter of slack discipline or thin training — hand people a rule with no system beneath it and AI use goes private: individual, invisible, unaccountable, the exact condition the frameworks in this book exist to prevent. The lesson is not to write a firmer guideline. It is to build the architecture the guideline was always meant to sit on, and the operating system that tells people how to decide once the guideline runs out.
Audience trust is shifting from the masthead to the journalist.
FT Strategies and WAN-IFRA surveyed 448 newsroom leaders across 86 countries for their Future Newsrooms Study 2026, and built one of its four ‘gaps’ entirely around trust. The finding: audience trust is moving from institutional authority toward relational signals — the credibility, affinity and transparency that individual journalists carry — and most newsrooms are not structured to supply it. Reporters still spend 38 per cent of their week on production and only 11 per cent on the post-publication work, the community engagement and direct audience contact, that makes expertise visible. And 78 per cent of newsrooms now think prominent journalists may need a pay structure that sets them apart from the room.
What the study calls the trust gap is, in the language of this issue, the trust–authenticity tension at industry scale. The reading to resist is the easy one: that trust is migrating to individuals, so the fix is to turn reporters into personalities. That mistakes one face of authenticity for the whole. Editorial authenticity — transparent process, clear sourcing, honest uncertainty — is trust’s ally, and is protected without exception. Emotional authenticity — the sense of a real person behind the work — is the face the study is measuring, and it does need deliberate cultivation. A newsroom that chases relatable voices while loosening its standards has pulled the rubber band too far in one direction, and will lose the trust it was reaching for. The shift is real. Holding it without snapping is the editorial task.
The TAP Triangle tells you what to protect; it does not tell you how closely to watch each AI tool you put in front of readers. The next issue takes up the instrument that does — the Visibility–Editorial Control Matrix — and the four quadrants every AI application falls into, one of which is more dangerous than it looks.
Until then, I would value your perspective from where you sit: which of the three points does your own operation lean on hardest, and which one quietly gives way when they pull against each other? Hit reply.
If this was useful, pass it on.
Until next time,
Dietmar
The Field Note: Twice a month, for editors, publishers, and heads of product working out what AI means for their newsrooms. One piece of original analysis. Two things worth a senior reader’s time.