Medical Review Process

The Short Version

Every health-critical page on this site moves through the same six stages before it goes live. A worked example of each stage follows, using a real page on this site (Cat Urinary Blockage Emergency) as the walk-through. That page was chosen because it is unambiguously YMYL, the consequences of getting it wrong are serious, and the editorial process had to be correspondingly disciplined.

1. Topic selection — is this a question real owners ask, and is our guidance actually going to change what they do next?
2. Source gathering — pull at least three tier-1/tier-2 clinical references before any drafting starts.
3. Drafting — draft from the references, not from generic knowledge. AI-assisted drafting is permitted; AI-generated publishing is not.
4. Four-check editorial review — fact, safety, clarity, bias. Documented per article.
5. Clinical cross-check — every numeric claim and every escalation threshold re-verified against the named source.
6. Post-publish lifecycle — 90-day review cadence for health-critical pages, plus reader and professional correction intake.

Stage 1: Topic Selection

We pick topics from three inputs: the AI Pet Help conversation log (what are real users actually asking about?), referring-search data from Google Search Console (what queries land on the site?), and reader corrections email (what have readers flagged as missing?). A topic only becomes an article when we believe the answer we can publish will meaningfully change what an owner does in the next 24 hours.

For the urinary-blockage page, the trigger was a cluster of chat transcripts in which cat owners described a male cat straining in the litter box — a presentation that can progress from "uncomfortable" to "fatal" inside of 24 to 48 hours. If the article did not exist, a share of those owners would go to bed assuming their cat had a "minor UTI." The article's existence is the intervention.

Stage 2: Source Gathering

Before a single sentence is drafted, the lead pulls reference material. Health-critical pages require at minimum:

• Two peer-reviewed or ACVIM-consensus references
• One professional-body client handout or guideline (AAHA, AVMA, ISFM)
• Merck Veterinary Manual section as the baseline reference

For the urinary-blockage page, the references pulled were:

• ACVIM Consensus Statement on Feline Idiopathic Cystitis / FLUTD — for the differential between obstructive and non-obstructive LUTD, risk factors, and long-term management.
• ISFM Guidelines on the Management of Feline Lower Urinary Tract Disease — for triage language, environment optimisation, and owner-facing signs.
• Merck Veterinary Manual: Feline Lower Urinary Tract Disease — baseline clinical description.
• Cornell Feline Health Center client page on urinary disease — to calibrate reader-friendly phrasing.
• AAHA/AAFP Feline Life Stage Guidelines — for prevention and risk-factor context.
• Published cost data from VPI and Nationwide pet insurance claim analyses — for the cost-range block on emergency unblocking procedures.

This is not an exhaustive list and it is not the site's marketing language. It is what actually sits open on a second monitor while the draft is being written.

Stage 3: Drafting

We are upfront about AI assistance. AI tools are used to summarise reference material and to produce an initial scaffold. A human then rewrites that scaffold against the open references, discards anything that sounds confident but is not anchored to a citation, and replaces generic language with the specific numbers, breeds, and timelines that matter for the particular topic.

Drafting rules for health-critical pages:

• No specific drug doses. We describe drug classes and the purpose of the medication. We do not publish "give X mg" because dosing is a prescribing decision.
• Any claim with a number gets a citation path. "103-105°F rectal is elevated; 106°F+ is a medical emergency" in the heatstroke page traces back to Merck and VECCS triage literature. If we cannot source a number, we soften the language or drop it.
• Escalation thresholds are explicit. Every symptom page has a "monitor at home / call today / go now" structure with concrete triggers. Vague triage language fails a reader in a crisis.
• What the vet will do. Owners who have never walked into an ER for their pet need to know what the visit looks like clinically. This reduces fear-of-cost avoidance, which is a recognised welfare problem.

Stage 4: The Four Checks

Fact Check

Every factual claim is re-verified against the sources pulled in stage 2. Numbers are double-checked against the Merck Veterinary Manual or the cited consensus statement. Claims we cannot confirm are softened or dropped. Claims that sources disagree on are noted as uncertain and resolved in the direction of the more cautious recommendation.

Safety Check

We read the article through one question: could any sentence here be misread in a way that harms an animal? Examples of things that get rewritten at this stage:

• Any phrasing that might lead a reader to delay veterinary care when the symptom pattern is an emergency. If the content ever risks sounding reassuring about a genuine emergency, it is rewritten to be explicit about the risk.
• Any home-remedy suggestion that could compound harm. Example: inducing vomiting with hydrogen peroxide is explicitly NOT recommended for caustics, sharp objects, or an already-neurologic animal — and the page has to say why, not just "do not do this."
• Any language that implies a site-based tool can substitute for clinical examination. The AI chat and symptom checker are explicitly positioned as triage assistance, never as diagnostic.

Clarity Check

Veterinary jargon is defined the first time it appears. "Unproductive retching," "nuclear sclerosis," "stertor versus stridor," "bilious vomiting syndrome," "granulomatous colitis" — each gets a short plain-language gloss. Readers in crisis cannot Google a term they do not know.

Bias Check

Three kinds of bias get audited at this stage:

• Commercial bias. The page is read with the affiliate/ad layer hidden to ensure no clinical recommendation has drifted toward a product category that happens to be monetised.
• Breed / species stereotypes. Breed-predisposition claims must trace to published data (e.g., CERF, VMDB, published breed-prevalence studies). "Aggressive breed" generalisations are removed.
• Unsupported confidence. Certainty language gets audited. If the literature is split, we say so.

Stage 5: Clinical Cross-Check

For the urinary-blockage page specifically, the cross-check log for the first publish looked like this:

This is the artefact that would be surfaced if a licensed veterinarian asked us to prove a specific claim. We keep the log for every health-critical page.

Stage 6: Post-Publish Lifecycle

Scheduled Review

• Emergency and symptom pages: re-reviewed every 90 days.
• Medication-adjacent pages (toxicities, drug-class overviews): re-reviewed every 90 days.
• General care and husbandry pages: re-reviewed every 180 days.
• Breed profiles and species overviews: re-reviewed annually, or immediately if new breed-predisposition data is published.

Corrections Intake

Errors come from three channels and all three are handled the same way: update the page, add a dated correction note visible at the bottom of the page, bump the "last reviewed" stamp at the top, and log the correction in the public Corrections Log.

• Reader email to corrections@petcarehelperai.com
• Professional flagging from veterinarians, RVTs, or credentialed behaviorists. We welcome this and credit the professional on the corrections log if they opt in.
• Internal re-review flagging a claim that is no longer current with the literature.

What We Do Not Do After Publish

• We do not quietly update an article to fix an error without a visible correction note.
• We do not move an article's published date forward to make it look more recent than it is.
• We do not remove an article to conceal that it existed; if content is retired, a /410-style redirect explains why.

What This Process Cannot Do

Our review process is designed to keep educational guidance safe. It is not a substitute for a physical examination, bloodwork, imaging, or any other diagnostic a licensed veterinarian would perform. We are also not in a position to tell you what is happening to your specific pet. Two dogs with the same symptom pattern can have different causes, different prognoses, and different appropriate interventions. That is why every symptom page ends with a "call your vet" pathway, not a diagnosis.

Who Reviews This Page

This very page — the one describing the review process — is audited by the editorial lead quarterly and updated if the process changes. It is the kind of page that becomes embarrassing if it falls out of sync with reality, so we keep it honest.

Related Trust Pages

• Editorial Team and Clinical Advisory References — named contributors and the clinical reference set.
• Editorial Standards — the full editorial policy.
• Corrections Log — what we have fixed, and when.
• Medical Disclaimer — the clinical and legal boundaries of everything here.
• About Pet Care Helper AI — who runs the site and why.

Page last reviewed: 2026-04-21. Process audited quarterly; any material change to the review pipeline is reflected here before it is applied elsewhere.