Diagnosing Your Dog With AI: A Prompt Guide

This guide exists because of a ChatGPT conversation that started in desperation and ended up being the most important document we created during Luna's illness. We didn't know what we were doing when we started. By the end, we had a 40-message thread that held the complete picture of her decline — more detailed and accurate than anything any individual vet had access to.

You don't need to be a tech person to do this. You don't need a special tool or a subscription. You need a thread, a habit, and the willingness to tell it everything.

The right mindset: reporter, not diagnostician

Your job is not to figure out what's wrong. Your job is to report what you see, as accurately and completely as possible, and let the AI find the pattern. This is an important distinction. The moment you start filtering your observations — deciding what's relevant, editing out the stuff that feels embarrassing or unrelated — you start losing the signal.

The detail that cracked Luna's case was that her back right paw was dragging slightly. It seemed minor. I almost didn't mention it. When I did, the AI immediately connected it to the front left limp and named a neurological cause for the first time. Two affected limbs on diagonal sides — that's a textbook spinal cord pattern. The AI knew that. I didn't. The vet, in a 15-minute window, hadn't connected it either.

The rule: if you noticed it, include it. Even if it seems unrelated. Even if it seems like your fault. Even if you're embarrassed about it. The AI is not judging you — it's pattern-matching. Give it everything to work with.

How to start your thread

Open one conversation and keep it open. Do not start new chats for each question. The power of this approach is the accumulated context — the AI knowing everything that came before when you add new information. A fresh chat has no memory of what you told it yesterday.

A practical note: very long threads (40+ messages) can start to get unwieldy. The AI may begin to lose track of early details, and you may burn through context limits faster than expected. When that happens, the solution isn't to start over — it's to ask the AI to generate a summary of everything it knows so far, save that summary, and use it to seed a fresh thread. More on that in the tracker section below.

The opening prompt

Your first message sets the context for everything that follows. Be thorough. Include your dog's breed, age (approximate is fine), weight, any known health conditions, current medications and supplements, and a clear description of what you're seeing.

Here's a template based on exactly how I started:

I need help tracking and understanding my dog's health symptoms.

Dog: [name], approximately [age] years old, [breed or mix], [weight] lbs.
Known conditions: [arthritis / allergies / etc., or "none known"]
Current medications/supplements: [list them, or "none"]

What I'm seeing: [describe the primary symptom in plain language — when it started, which leg/area, how often, how severe]

Additional observations:
- [anything else you've noticed, even if you're not sure it's related]
- [changes in behavior, appetite, sleep, energy]
- [anything that makes it better or worse]

Vet history on this issue: [what vets have said so far, what tests have been done, what they found or didn't find]

I want to keep this as an ongoing thread. I'll add updates as things change. Please ask me follow-up questions if there's anything that would help you build a clearer picture.

What to include — nothing is too small

People consistently under-report. Here are the categories of information that matter most, including several that seem irrelevant but aren't:

Timing patterns

Is it worse after rest? After exercise? In the morning? After eating? Timing is often diagnostic. Luna's limp being worse right after a nap pointed toward neurological involvement — resting muscles reveal nerve issues that activity can temporarily compensate for.

Which exact limb, and where

Front left vs. front right matters. Whether the dog is licking the pad vs. the top of the paw matters. The specific location of a symptom is information — don't round up to 'front leg.'

What's happening on other limbs

This is the one that cracked our case. If you notice anything unusual in any other leg — dragging, weakness, occasional stumble — report it. Diagonal limb involvement is a classic neurological pattern.

Behavioral changes

Is your dog sleeping more? Less interested in play? Reluctant to do things they used to do willingly (jump up, use stairs, greet you at the door)? Behavioral changes are symptoms. Include them.

Things you think might be your fault

Did they jump off the couch? Eat something unusual? Have a rough play session? Tell it. It's not about blame — it's about building a complete timeline. If it's not relevant, the AI will say so.

What the vet said — and what they didn't

Include the exact diagnosis given (or 'no diagnosis'), what tests were done, what was ruled out, and what was left unexplained. The gaps in the vet's account are as important as what they found.

Moments of apparent improvement

Don't only report the bad days. If your dog had a good stretch, report that too. The pattern of improvement and relapse is itself diagnostic information.

Have it build you a symptom tracker

After you've had a few exchanges, ask the AI to build you a structured tracker. This serves two purposes: it creates a document you can share with your vet, and it gives you a summary you can use to seed a new thread if your conversation gets too long.

Based on everything we've discussed, can you:

1. Create a symptom timeline — a chronological list of when each symptom first appeared and how it's changed
2. Build a simple daily tracker template I can fill in each day (just the key things to watch)
3. Summarize the current diagnostic picture — what we know, what's been ruled out, and what the most likely explanations are based on what you've seen

I want to save this as a document I can share with my vet.

Save that output. Copy it to a doc. If you ever need to start a fresh thread, paste it in as the first message with a note: "This is a summary of an ongoing health tracking conversation about my dog. Please treat this as your full context and continue from here."

Going deeper: the Claude Project approach

If you're using Claude specifically, there's a feature called Projects that takes this to another level. A Project is a persistent workspace where you can upload files, save artifacts, and have conversations that all share the same context. Think of it as your dog's medical file — but one that can read, analyze, and talk back.

What to put in your Project:

📎

Vet notes and summaries

After every appointment, type up or photograph what was discussed. Upload it.

📎

X-rays and test results

Yes, you can upload images. Claude can look at an x-ray and tell you what it sees — not as a diagnosis, but as a starting point for better questions.

📎

Your symptom log

The running document from your chat thread, updated regularly.

📎

Medication and supplement list

With dosages, start dates, and any changes. Context for interpreting new symptoms.

📎

Vet conversation guides

Before every appointment, ask Claude to generate a list of questions based on your current symptom picture. Print it and bring it.

📎

Anything that feels relevant

A note from a specialist. A Facebook group post that sounded exactly like your dog. A study you found. Upload it and ask Claude what it means for your situation.

The goal is to give the AI more complete information than any single vet has ever had about your dog. Your vet sees your dog for 15 minutes every few months. Your AI project sees everything — every symptom, every test result, every medication change, every good day and bad day — and holds it all at once.

Getting a second opinion: vet-specific AI tools

General-purpose AI like ChatGPT and Claude is powerful, but it's generalist — trained on everything, specialized in nothing. There are also AI tools built specifically for veterinary medicine, trained on clinical literature, case studies, and diagnostic frameworks used by actual veterinarians. These can be worth layering in, especially when you have a complex or puzzling situation and want a domain-specific second look.

The distinction matters: a general AI is pattern-matching across everything it knows about medicine, behavior, biology, and your conversation. A vet-specific AI is pattern-matching against a more focused body of veterinary knowledge. They can catch different things. Using both is not overkill — it's due diligence.

Tools worth knowing

Ask.Vet

A veterinary-specific AI assistant trained on clinical knowledge. Good for symptom triage, medication questions, and getting a second read on a diagnosis. Useful when you want veterinary framing rather than general medical pattern-matching.

Try it for: "My dog has these symptoms. My vet said X. Does this diagnosis fit, and what else should be ruled out?"

VetGPT / VetAI (check current name)

There have been several veterinary-specific GPTs and tools in this space — Melissa used one during Luna's illness that has since been renamed or updated. Search 'veterinary AI second opinion' or check the GPT Store for current vet-focused tools. The landscape changes fast.

Use the same approach you'd use with any AI: give it everything, don't filter, ask it to flag what it's uncertain about.

Your general AI, with a vet-specific prompt

You can also shift the framing of your existing conversation. Ask ChatGPT or Claude to "respond as a veterinary specialist reviewing this case" or "evaluate this from a veterinary neurology perspective." It won't give you the same depth as a purpose-built tool, but it can shift the angle usefully.

A caveat on all of this:none of these tools — general or veterinary-specific — are a replacement for a licensed vet who can actually examine your dog. What they are is a way to arrive at that examination better prepared: with better questions, a clearer picture of what's already been ruled out, and a stronger sense of what imaging or testing to push for. The goal is to walk into that 15-minute appointment having already done the work a 30-minute consult would normally require.

Before every vet visit: generate your questions

One of the most practical things you can do is ask the AI to prepare you for your appointment. This takes five minutes and can completely change the quality of your 15-minute window with the vet.

I have a vet appointment on [date]. Based on everything we've tracked about [dog's name], can you:

1. Give me a 3-sentence summary of the current situation I can read to the vet at the start of the appointment
2. List the 5 most important questions I should ask, in order of priority
3. Tell me what I should push for if the vet doesn't suggest it (specific tests, imaging, referrals)
4. Flag anything in our symptom history that the vet should know about but might not ask about

Print it. Bring it. Don't be embarrassed — vets respect prepared owners. And if your vet seems annoyed that you've done your homework, that's information too.

Prompts you can copy right now

There is no perfect prompt. The best thing you can do is be candid and thorough — more than you think you need to be. These templates are starting points, not scripts. Edit them to sound like you, add the details that feel relevant even if you're not sure they matter, and don't worry about phrasing anything perfectly. The AI is not grading you. It just needs the information.

A quick-reference list for the most common situations:

Starting fresh with a new symptom

My dog [name] just started [symptom]. They are [age], [breed], [weight]. Here's what I've noticed: [describe]. It started [when]. Here's what I've tried: [list]. Here's what the vet has said: [or "nothing yet"]. Please ask me follow-up questions.

Updating an existing thread

Update: it's been [X days] since my last message. Here's what's changed: [describe]. [Dog's name] has been [better/worse/the same] overall. New things I've noticed: [list]. Does this change anything about the picture?

After a vet appointment

We just got back from the vet. Here's what happened: [summary of visit]. The diagnosis/recommendation was: [what they said]. Tests done: [list]. What was NOT addressed: [anything you feel was glossed over]. Based on our history, does this make sense to you? What should I be watching for?

When something gets suddenly worse

Something changed suddenly today. [Dog's name] is [describe what's different]. This is new as of [when]. On a scale of urgency, should I call the vet today, go to an emergency clinic, or monitor at home? What signs would tell me it's getting worse?

When you think the vet missed something

My vet said [diagnosis/conclusion]. But I'm not convinced, because [reason]. Given everything in our history, is there something else this could be? What would I need to ask for or test to rule it out?

The most important thing isn't which tool you use or how perfectly you phrase your prompts. It's that you start. Open a thread today. Type what you're seeing. Keep adding to it. The record you build will be there when you need it — whether that's at a vet appointment next week, a specialist consultation next month, or the moment when one small detail finally makes everything click into place.