How We Use AI

Last updated April 21, 2026

Clinical Nutrition Report covers AI calorie tracking apps. We use AI tools internally as well, in narrowly defined ways. We publish this page so readers can know exactly where AI does and does not appear in our editorial process, and so that readers concerned about AI-generated content can evaluate our work on its merits.

The short version

Every article on this site is written, reviewed, and approved by credentialed humans. Bylined contributors are real people with verifiable credentials (see our authors page). We do not publish AI-generated content under human bylines. We do not use AI to fabricate research summaries, citations, or clinical claims.

We do use AI as a research and editing assistant in three specific, bounded ways. These are documented in detail below.

Where we use AI

1. Research summarization (orientation only)

Contributors may use large language model tools to orient themselves on the literature in a sub-topic before beginning a piece. This is functionally similar to reading a Wikipedia overview as a starting point: it is helpful for getting a fast read on the shape of a body of evidence, but it is not authoritative and is not used as a primary source.

Every claim that ends up in published content is independently verified against a primary source by the contributor and again by our fact-checker. AI-summarized claims that cannot be verified are removed.

2. Citation finding

AI search tools are used to surface candidate citations — "what RCTs have evaluated protein intake at 1.6 g/kg in resistance-training contexts?" — that the contributor then independently retrieves, reads in full, and decides whether to cite. We have learned, like everyone else who uses these tools, that AI-suggested citations are sometimes plausible-looking but nonexistent. Our fact-checking process catches these; no citation appears in published content unless a human has retrieved and read the underlying paper.

3. Copy editing

Contributors may use AI tools to suggest sentence-level rewrites for clarity, tighten paragraphs, or surface awkward phrasing. The substance of the writing is the contributor's; AI is used the way a thoughtful copy editor would be. The contributor remains responsible for every word that appears under their byline and is the sole authority on the content's accuracy.

Where we do NOT use AI

• We do not generate full articles with AI. Every long-form piece is drafted by a credentialed human author from scratch.
• We do not use AI to evaluate apps. Our app testing — weighed reference meals, photo-recognition battery, database probes — is performed by humans against a documented protocol on our methodology page. Scoring is human-assigned; no AI assigns rubric points.
• We do not use AI to write reviews of apps that themselves use AI. The conflict of interest in having one AI review another, even an unrelated one, is too obvious to ignore.
• We do not use AI for clinical claim verification. Clinical claims are verified by the credentialed contributor and our fact-checker against primary sources retrieved and read by a human.
• We do not use AI to fabricate quotes, paraphrase patient stories, or generate case scenarios that imply real patient experiences.
• We do not auto-publish AI-generated content. Every page on the site has a human author, a human reviewer, and a human editor.

Why we publish this

The volume of low-quality, AI-generated nutrition content on the open web has exploded since 2023. Much of it is plausible-looking, citation-loaded, and wrong in ways that are hard to spot. We think the appropriate response from a publication that wants to be useful to clinicians and serious readers is to be explicit about how AI fits into the editorial process: where it helps, where it is excluded, and what guardrails exist.

We also think readers should be able to make an informed choice. If you would prefer to read content with even less AI involvement than what we describe above, that is a reasonable preference, and we are not the right publication for that preference. If you would prefer content where AI did not have to be disclosed at all because it generated everything, we are also not the right publication.

Disclosure on individual pieces

Pieces that involved more substantive AI assistance than the three categories above (research summarization, citation finding, copy editing) carry a per-piece disclosure noting the specific assistance. To date, no pieces have required such a disclosure. If that changes, the disclosure will appear in the byline area of the affected piece and the change will be noted in our update log.

Tools we currently use

We do not maintain a public list of specific AI tools or model versions in use, because the tooling changes quickly and a list would mislead about currency. Contributors use whatever current commercial AI assistants are available; the editorial constraints above bind regardless of which tool is used.

For questions about our AI policy, contact editor@clinicalnutritionreport.com.