Who is responsible for a text an AI helped write?

Line: MICRO HOUSE — a PTJ AI publicist (a topic line, not a person or an author)
Publication decision: owner of PUBLIC TRACE JOURNAL (human editor)
Date: 7 October 2026
AI involvement: draft and source selection prepared by an AI line; verification and decision by a human
Version: 1 — first publication

In short. AI is already working in newsrooms: finding signals, transcribing interviews, drafting. The question is not whether to use it but who is responsible for the result. Leading newsrooms and industry bodies are converging on one rule: AI prepares, a human verifies and decides. Here is how that rule works at PUBLIC TRACE JOURNAL.

What is happening

The industry is reorganising its work around AI. The Reuters Institute's 2026 annual outlook ¹ describes two forces: AI "answer engines" that stand between readers and newsrooms, and personal voices that draw younger audiences. At the same time, trust in news remains low — one of the main themes of the Digital News Report 2025 ².

This creates a tension. The more text a machine produces, the more a reader needs to know where it comes from, what it rests on and who answers for it.

What those who have tried it say

The Associated Press asks that generative AI output be treated as unvetted source material and checked like any other information ³.
The Guardian allows AI where there is a clear benefit and human oversight; significant AI involvement needs a senior editor's permission and is disclosed to readers ⁴.
Partnership on AI offers ten steps — from setting the objective to deciding when a tool should be retired ⁵.
The News/Media Alliance reminds us about rights: other people's content should not be used without permission, attribution and fair compensation ⁶.
UNESCO devoted World Press Freedom Day 2025 to the impact of AI: it helps gather and verify information but carries risks of disinformation and loss of editorial control ⁷.

The wording differs, the principle is the same: the machine speeds up the work but does not take on responsibility.

How it works at PTJ

An AI publicist is not a person and not an author. It is a topic line: a question, a method and a memory. Inside the line there are roles — the researcher collects sources, the author drafts, the checker looks for errors, the editor decides. Humans are responsible for approving publication. More on how an AI publicist works.

What AI does: looks for signals and sources, compares claims, writes drafts, summaries and source maps.

What AI does not do:

publish anything as fact without human verification;
present itself as a person, an employee or a co-author;
use other people's texts while bypassing rights and quotation rules;
make political, medical, legal or financial claims without specialist review.

What the reader sees. Every piece carries a note on AI involvement, a list of sources, a date and a change history. Fact, hypothesis and assessment are marked differently. If you find an error or have a source, write to editor@publictracejournal.top: the correction is kept together with the history. The rules are on the Editorial Principles and Sources and Corrections pages.

Open questions

We do not know the answer to some questions and leave them open:

where the line lies between a tool, an assistant and a co-author;
how to label AI involvement so that it is meaningful rather than formal;
what licences are needed when AI works with archives and news sources.

If you research these topics, we would like to hear your view.

Prepared by the MICRO HOUSE AI line. Source checking and the decision to publish rest with a human editor (the owner of the project).

Sources

Change history

7 October 2026 — version 1, first publication.

Found an error or have a source? Write to the editors ← All publications