Ijcit briefings are written and edited by humans. We do not publish AI-generated draft text under the name of an editor or contributor. We do not use AI to invent quotes, citations, or details that a human writer did not verify.

The pledge

The following commitments hold for every briefing published on this site, every issue of the weekly newsletter, and every sector report:

• No briefing body copy is generated by a large language model and published without rewriting by a named human editor. Where a writer uses AI to summarise a long source document for their own reading, the briefing is then written from the source, not the summary.
• No quote attributed to a named person is generated, paraphrased, or composited by an AI tool. Quotes come from interviews, on-the-record statements, public filings, or other verifiable sources.
• No citation, source link, or factual reference in a briefing is suggested by an AI tool without human verification against the underlying source. “The model said X is true” is never a sufficient basis for a claim.
• No bylined article is written end-to-end by an AI tool. The named author has read every primary source cited, written the draft, and stands behind the claims made.
• No AI-generated imagery is used to depict real people, places, or events. Stock imagery, commissioned photography, and editorial illustration follow our usual sourcing standards. Where AI-generated illustration is used for purely abstract or decorative purposes (a hero graphic, a pattern background), it is clearly labelled.

Where we do use AI

We use AI tools in the editorial process for tasks that do not produce published prose. Specifically:

• Transcription. Interview audio transcribed by speech-to-text tools (Whisper, AssemblyAI, etc.). Transcripts are reviewed against the source recording by a human before any quote is used.
• Citation lookup. AI tools to find the URL or DOI for a regulation, paper, or vendor document we already know we want to cite. Every citation is opened and verified by a human before publication.
• Grammar and copy-edit checks. Tools like Grammarly or LanguageTool for typo and style consistency. The editor accepts or rejects each suggestion individually.
• Internal research summarisation. Editors may use AI tools to summarise a long source document for their own reading, but the briefing is then written from the source, not the summary.
• Translation assistance. Where a source document is in a language an editor does not read fluently, machine translation may be used to identify the relevant passage. The relevant passage is then translated by a human (the editor, a colleague, or a commissioned translator) before any quotation or claim is published.

Why this matters

Our readers are senior IT decision-makers in healthcare and finance. Decisions made on the basis of a briefing affect patient outcomes, regulatory compliance, and financial settlement risk. Those readers need to know that a human with a name and a track record has signed up to the claim and the citation.

AI-generated text is plausible by design. It is also confidently wrong by design — not occasionally, but routinely. Hallucinated regulatory citations, invented vendor capabilities, and fabricated quotes are the standard failure mode of contemporary large language models. A publication that treats AI output as a publishing-ready source becomes, by definition, an unreliable source.

We do not believe AI tools will never produce reliable journalism. We do believe that as of today, the only way to be confident that a published briefing represents reality is to require a human writer who has read the sources to stand behind it.

How we tell

The editor reviewing a draft asks: “Could the writer point me to the source for every claim in this briefing?” If the answer is yes, the draft is fit to publish. If the answer is “I think I read it somewhere” or “the model said so”, the draft goes back for sourcing.

Detection and accountability

If a reader believes a briefing contains AI-generated copy that violates this pledge, we ask them to write to [email protected] with the specific passage and reasoning. We will review, respond, and where the pledge has been broken, we will publish the correction under our corrections policy.

We do not run AI-detection tools on our own drafts, because contemporary AI-detection tools are unreliable. We rely on the editorial process — sourcing, review, accountability — rather than on technical detection.

What this pledge does not cover

• Sponsor copy. Sponsor-supplied newsletter and banner copy is the sponsor’s responsibility, not ours. We do not edit sponsor copy for AI-generation. Sponsor copy is clearly marked as sponsorship and is not editorial.
• User-generated content. Where we publish reader letters, event-feedback notes, or other user-generated content, we do not screen for AI generation. Such content is presented as the reader’s submission, not our editorial work.
• Future product surfaces. If we add an AI-powered search, summary, or recommendation feature to the site in future, that feature will be clearly labelled as AI-generated. The pledge applies to editorial content, not to product features built on top of it.

Review

We review this pledge annually. If the underlying technology changes materially — if AI tools become reliable enough to write source-verified prose, or if the failure modes change — we will update the pledge and document the change here.

Last reviewed: 2026.