Can AI Help Diagnose Your Dog? How ChatGPT Caught What Two Vets Missed

If you need to know right now

Yes — AI tools like ChatGPT and Claude can be genuinely useful for dog health diagnosis, not as a replacement for a vet, but as a pattern-recognition tool that works across your full symptom history. In Luna's case, one ChatGPT conversation identified a neurological cause — diagonal limb involvement pointing to a spinal cord lesion — that two vets with x-rays had not flagged. The key is documenting everything in a single ongoing thread and holding nothing back. AI cannot examine your dog. It can help you ask better questions before the vet does.

I want to be clear about something before I tell you this story: I am not someone who thinks AI is magic. I work in content strategy and interact with AI every day — I spent two years working at an AI company when generative AI first hit the mainstream. I've watched people wildly overestimate what these tools can do. I know how they work. I know their limits. So when I finally opened ChatGPT and started typing everything I knew about my dog's symptoms, it wasn't because I thought a chatbot would save her. It was because I had nothing left to lose.

Luna had been limping for months. My partner Daniel was her primary caretaker — he works from home, so he was the one taking her to every vet visit while I was at work. He'd come home from those appointments a little distracted, a little out of sorts, and I could never quite tell if the details I was getting were complete. He'd tell me what the vet said, then go straight to his computer and spend the evening researching on his own. He was following everything the vet told him. Any improvement we saw was always short-lived.

I was getting fragments. Nothing was adding up.

The night I'd had enough

The vestibular episode broke me open. Luna woke up one morning, threw up — which she never did — and then started walking like she was drunk. By that evening, her head was tilting to one side and her eyes were shifting rapidly back and forth. We rushed her to the emergency vet. The diagnosis was vestibular disease. They gave her something for nausea and sent us home.

It took her days. Of course it did. She was a fighter.

Then, a week later, she collapsed again.

That was the moment. Too many little things had been adding up for too long, and now we were at the vet again, and I still didn't have a clear picture of what was happening to our little girl. The vets were kind. They genuinely loved Luna — everyone at her clinic did. But no one had looked at the full picture and said: this is what's happening, and this is what you need to do.

I knew from my work in AI that these tools had, in some cases, diagnosed breast cancer with more accuracy than human doctors. I figured I had nothing to lose. So I opened ChatGPT and I started documenting.

How I used ChatGPT to track my dog's symptoms

I didn't start with a prompt. I started with everything. Every symptom, every timeline detail, every small observation I'd been holding in my head for months. The limp that started on the front left leg. The way she'd struggle going downstairs but charge right up them. The paw she kept licking — not the pad, the top. The leg that had become visibly larger than the other from compensating. The fact that she limped worse right after a nap.

I told it things I wasn't sure mattered. I told it things I was embarrassed to think might be connected. I told it when I thought we might have made a mistake — the rawhide she ate before the limp started, the time she jumped off the couch before I could catch her, the supplements we were giving her. I kept nothing back, because I had no idea what piece of information would be the key.

The moment ChatGPT identified what the vets missed

Early in the conversation, ChatGPT was covering the basics — arthritis, soft tissue injury, the usual suspects for a limping senior dog. I kept adding detail. Then I mentioned that I had also noticed her back right paw dragging slightly.

The response changed completely.

It said: “Ahhh, okay — that changes things significantly and gives us a clearer clinical picture.”Then it explained that when both a front leg and the diagonal rear leg are affected, that's a classic pattern for a spinal cord lesion — not arthritis, not a soft tissue injury. The spinal cord controls both pairs of limbs, and a lesion higher up, at the neck or upper back, can affect diagonal limbs due to how the nerves branch. It named conditions I had never heard of: IVDD, cervical disc disease, brachial plexus neuropathy. It said explicitly that standard x-rays cannot show soft tissue, discs, or spinal cord compression. It said we needed an MRI, and we needed a neurologist.

Our vet had done x-rays. Seen nothing concerning. Sent us home.

The AI flagged a neurological cause within one conversation, using information our vets had — and didn't connect.

A word about vets

I don't want this story to read as an indictment of veterinarians, because that is not what I believe. Luna's vet loved her. Everyone at that clinic lit up when she walked in. That love was real, and in the hardest months of her life it mattered more than I can say to know that the people caring for her actually cared about her.

But love isn't diagnosis. Compassion isn't protocol.

Veterinarians are working with a patient who cannot describe their symptoms, in a 15-minute window, with the visual information in front of them and whatever history you've managed to communicate. Blood, urine, x-rays — those cover a slice of what's possible. There is a large space of diagnoses that simply cannot be reached without advanced imaging, without a specialist, without someone who has the time and the context to look at the full picture.

The gap between “we care about your dog” and “here is exactly what is wrong and what to do” — that gap is where families fall through. It's where we fell through. This site exists in that gap.

The awkward conversation you'll probably have

You already know the look you get when you tell a doctor you googled your symptoms. There's a version of that look — sharper, with more history behind it — when you tell a vet you asked AI.

“I went to veterinary school for four years. I have treated thousands of animals.” That's not a wrong thing to say. It's true. But it's also a deflection from the question you actually asked, and it's designed to make you feel like you overstepped by asking it.

Here is what I want you to hold onto: your dog does not have a voice. You are the only person in the room who has watched them every single day for months, who has noticed the small changes, who has logged every symptom and every pattern. That observation is not nothing. You are not a vet — but you are the expert on your specific dog in a way no vet who sees them for fifteen minutes can be.

You are also paying the vet. That is not a power move — it's just a fact that it is okay to remember. You can ask questions. You can push back. You can say “I've been tracking her symptoms with an AI tool and it keeps pointing toward neurological involvement — can we talk about that?” and if the answer is dismissal rather than engagement, that is information about whether this is the right vet for the situation you're in.

Don't insult them. Don't lead with “ChatGPT told me.” Lead with the observations — the duration, the pattern, the specific symptoms — and let the conversation go where it goes. If a doctor can't engage with a prepared, specific question from a worried owner, find a doctor who can.

What the conversation became

I kept that ChatGPT thread open for months. Every new symptom went into it. Every update, every medication change, every vet visit, every moment that felt significant. It became the most accurate record we had of Luna's decline — more accurate than my memory, more detailed than Daniel's vet notes, more complete than anything any individual doctor had seen.

When Luna was eventually diagnosed with an inoperable spinal cord tumor at her neck, we went back through that conversation and the timeline was all there. Every moment the pieces had been pointing toward a diagnosis, documented in order.

That conversation is also how Daniel had the material to start writing about what happened. The AI thread became the source document for this site.

If you're reading this in crisis mode

Someone reading this is probably where I was — exhausted, scared, getting fragments of information that don't quite add up, watching their dog struggle and not knowing what to do next. I want to say a few things directly to you.

First: the overwhelm you're feeling is real and it makes sense. You are trying to understand a medical situation with no medical training, through a communication chain that is inherently incomplete, about a creature who cannot tell you where it hurts. Of course it's hard. It's supposed to be hard.

Second: the guilt you might be feeling — about what you missed, what you didn't push for, what you didn't know — that guilt comes from love. It is not evidence that you failed. You are doing the best you can with what you have. Keep going.

Third: start documenting. Today. Open whatever AI tool you have access to and start a thread. Tell it everything. Keep adding to the same conversation so the context builds. Don't edit for relevance — that's not your job. Your job is to report. Let the AI find the pattern.

That's how we found ours. Too late for Luna — but not too late to pass it on.

Frequently asked questions

Next step

Ready to try this yourself?

We turned everything we learned into a practical guide — the exact prompts to use, how to structure your thread, and what details matter most. Free to use, no account required.

Related reading