About 53% of veterinarians in a late-2025 Brakke Consulting survey reported using some form of AI, most commonly scribe tools and cytology appsSource ↗
In the Digitail/AAHA survey of 3,968 veterinary professionals, 83% were familiar with AI and nearly 30% of those already used it daily or weeklySource ↗
70.3% of veterinary professionals named accuracy and reliability their top AI concern, and 53.9% cited data security and privacySource ↗
Veterinary AI scribe users report saving up to 2 hours per shift on documentation, with up to 75% less time spent writing recordsSource ↗
writingClaudeChatGPT

SOAP notes from exam-room shorthand

Veterinarians commonly spend one to two hours a day finishing records after appointments end, and scribe tools are the single most-used AI category in the profession. Even without an ambient scribe, turning quick exam-room shorthand into a complete, defensible SOAP note is the most repeatable AI win — and thin records are what state boards and insurance reviewers punish.

Prompt
You are an experienced veterinary scribe who writes complete, defensible medical records. Convert my shorthand into a structured SOAP note for a {{visit_type}} visit. Patient signalment: {{signalment}}.

My shorthand: {{shorthand}}

Format:
- S (Subjective): presenting complaint as the owner described it, relevant history.
- O (Objective): physical exam findings by body system, vitals, diagnostics performed.
- A (Assessment): the diagnosis or differential list exactly as I stated it.
- P (Plan): treatments given today, medications dispensed, diagnostics ordered, owner communication, and recheck plan.

Rules:
- Use only what is in my shorthand. Where a standard element is missing (weight, temperature, consent, body condition score), write [VERIFY: element] instead of inventing it.
- Never invent drug doses, routes, or frequencies — copy them exactly as I wrote them or flag them.
- Factual, past tense, no speculation. Expand ambiguous abbreviations; standard veterinary abbreviations (BAR, QAR, mm, CRT) are fine.
- After the note, list anything a state board or records audit would expect for this visit type that is still missing.

Fill in your details and the prompt updates live — then copy.

communicationClaudeChatGPTGemini

Discharge instructions owners actually follow

Owners retain little of what is said at pickup, and unclear home-care instructions are a top driver of failed recoveries and after-hours calls. Drafting discharge notes is one of the most-cited generative AI uses in veterinary practice — roughly half of surveyed professionals want AI help with exactly this document.

Prompt
You write discharge instructions for a veterinary practice. Create a take-home sheet for: {{procedure_or_diagnosis}}. Patient: {{signalment}}. Medications I prescribed (use these exactly — do not add, remove, or change any dose): {{medications}}.

Format, in this order:
1. "What we did today" — 2-3 plain sentences.
2. "The next 48 hours" — the 4-6 things that matter most, as short bullets (food, water, activity, incision or affected-area care).
3. "Medications" — one row per medication: name, what it is for, and the dosing exactly as I wrote it, reworded only for clarity.
4. "What's normal" — expected things (grogginess, mild swelling) with realistic timelines.
5. "Call us if" — clear warning signs on separate lines, with our number {{clinic_phone}}, plus when to go to an emergency hospital instead.

Rules:
- 6th-grade reading level, short sentences, no clinical jargon without a plain-word explanation.
- Use only the medications and instructions I gave you. If a standard element is missing (e-collar duration, suture removal date, recheck timing), insert [ASK DR: element] rather than guessing.
- Under 400 words so it prints on one page. Use [PET_NAME] as a merge field — never a real name.
- End with a 2-sentence version sized for a follow-up text message.

Fill in your details and the prompt updates live — then copy.

planningClaudeChatGPT

Prior-record summaries before referrals and new patients

A new patient or referral shows up with 60 pages of scanned PDFs from three previous clinics, and someone has to reconstruct the story before the appointment. Manual chart review runs 20-30 minutes per complex patient; summarization is now a headline AI feature across veterinary software, and the same job can be done with a general chatbot on de-identified record text.

Prompt
You are a veterinary internist's assistant preparing a chart review. Summarize these prior medical records for an upcoming appointment. Reason for the visit: {{appointment_reason}}.

Records (de-identified): {{record_text}}

Output, in this order:
1. Signalment and a 3-sentence case narrative.
2. Problem list — each problem with date of first mention and current status (active, resolved, unclear).
3. Medications — current and discontinued, with doses exactly as documented and the date last mentioned.
4. Diagnostics — labs and imaging with dates; flag every out-of-range value the records report and show the trend if a value appears more than once.
5. Vaccine and preventive status, if documented.
6. Open questions — conflicts between entries, gaps in the timeline, and anything I should ask the owner or the referring vet.

Rules:
- Use only what is in the records. Where something is not documented, write "not in record" — never fill gaps with what is typical.
- Do not add diagnoses, interpretations, or recommendations of your own; if two entries conflict, show both with dates.
- Keep it under one page. Bullet points, newest first within each section.

Fill in your details and the prompt updates live — then copy.

analysisClaudeChatGPT

Bloodwork explanations worried owners understand

A senior panel comes back with four flagged values, and the vet has ten minutes to call an anxious owner who will otherwise take the numbers to Google. Extracting out-of-range values and turning them into plain language is one of the documented core ChatGPT uses in practice — as long as the interpretation stays the veterinarian's.

Prompt
You help a veterinarian explain lab results to a pet owner. Patient: {{signalment}}. Lab values (with reference ranges): {{lab_results}}. My assessment as the attending veterinarian — this is the only interpretation you may present: {{vet_assessment}}.

Produce:
1. A table of every out-of-range value: value, reference range, direction, and one plain-English line on what that value measures (what the organ or system does — not what disease it implies).
2. A summary for the owner, 120-150 words, 6th-grade reading level, that explains the findings using only my assessment. Use [PET_NAME] as a merge field. Warm, calm, honest — no sugarcoating, no alarm.
3. Three questions the owner is likely to ask on the phone, with short answers consistent with my assessment.

Hard rules:
- Do not add possible diagnoses, prognoses, or treatments beyond what my assessment states. If my assessment does not cover a flagged value, list it under "Dr. to address" instead of explaining it away.
- Do not invent normal ranges — use the ones I provided.
- No percentages of survival, no cost estimates, no drug suggestions.

Fill in your details and the prompt updates live — then copy.

writingChatGPTClaudeGemini

A client-education library for prevention season

Every practice hands out the same 15 topics — heartworm prevention, dental disease, obesity, senior screening — and most handouts were written years ago at a reading level owners skim past. Client-education material is a documented ChatGPT use case for vets, and better education is one of the top reasons practices say they want AI: it drives compliance.

Prompt
You write client-education handouts for {{practice_name}}, a veterinary practice. Create a one-page handout on: {{topic}}. Audience: {{audience}}.

Structure:
1. "Why this matters" — the risk in everyday words, specific to this audience and region where relevant. 3-4 sentences.
2. "What you can do" — 4-6 concrete owner actions, as short bullets.
3. "Myth vs fact" — the three most common owner misconceptions on this topic, each corrected in one sentence.
4. "When to call us" — specific signs that warrant an appointment.

Rules:
- 6th-grade reading level. Friendly and direct, never preachy or scary.
- Name no drugs, brands, or doses — write "ask us which prevention option fits [PET_NAME]" instead.
- Do not invent statistics or study findings. Where a number would strengthen the point, write [STAT: what to look up] so I can fill it from AVMA, AAHA, or the American Heartworm Society.
- Do not give medical advice beyond widely accepted preventive-care guidance; anything condition-specific becomes "ask us at your next visit."
- Under 400 words so it prints on one page, plus a 2-3 sentence version for a social media post.

Fill in your details and the prompt updates live — then copy.

communicationClaudeChatGPT

Review responses that never confirm someone was a client

Veterinary reviews run hot — big emergency bills and end-of-life decisions produce one-star reviews written in grief or anger, and prospective clients read every reply. AAHA describes teams pasting reviews into ChatGPT for response options; the drafting is easy, and the constraint list is what keeps a public reply from becoming a confidentiality problem.

Prompt
You are the practice manager of {{practice_name}}, a veterinary practice, responding publicly to an online review. Draft a response.

Review ({{star_rating}} stars): {{review_text}}

Hard rules — these protect us legally and ethically:
- Never confirm or deny that the reviewer or their pet was seen at our practice. No "at your visit," no pet names, no "when you brought Max in."
- Never mention any treatment, diagnosis, outcome, bill, or date — even details the reviewer disclosed themselves.
- Never argue, correct their account, or explain what "really happened," and never blame the pet's condition or the owner.
- If the review involves the loss of a pet, lead with genuine, unhurried compassion — one full sentence of it before anything else.

Then:
- For positive reviews: thank them warmly in one or two sentences; reference only generic things (kind team, clear communication).
- For negative reviews: acknowledge the frustration or grief without admitting fault, state our general standard in one sentence, and move it offline: "Please call our practice manager at {{practice_phone}}."

Under 80 words, no emojis. Give me two versions: one warmer, one more formal.

Fill in your details and the prompt updates live — then copy.

Common questions from veterinarians

HIPAA doesn't cover animals — so can I paste patient records into ChatGPT?

HIPAA doesn't apply, but confidentiality still does: most state veterinary practice acts and the AVMA Principles of Veterinary Medical Ethics make client information and medical records confidential. The workable pattern is to strip client names and contact details before using a consumer chatbot, and use purpose-built veterinary tools with security certifications for anything that touches the full chart.

Can I trust AI to read radiographs or make diagnoses?

Treat it as a second opinion at most. Unlike human medicine, veterinary diagnostic AI faces almost no regulatory review, so products can reach market with little published evidence — and under the VCPR, the diagnosis and treatment decisions remain yours regardless of what the software said. The profession's own top concern in surveys is exactly this: accuracy and reliability.

Will AI get drug doses right?

Assume it won't. Dosing is species-specific, and a language model will confidently produce a dose that is wrong for the species or the formulation. Never let AI supply a drug name, dose, route, or frequency in a record or client document — those come from your prescription, verified against a formulary like Plumb's.

Do I need a veterinary AI scribe, or is a general chatbot enough?

They solve different problems. A general chatbot handles de-identified writing well — discharge instructions, handouts, note structuring, review replies — for the cost of a subscription. Purpose-built scribes earn their price where they record exam-room audio, integrate with your practice software, and carry security certifications; users report saving up to two hours per shift.

Related professions