Evidence & Validation
What the science supports, how PAWSCHECK works, how vet review works, and what we are still validating. We would rather under-claim than overclaim — so this page is deliberately honest about the limits.
Why objective gait analysis
The case for measuring movement rather than relying on the eye is well established in the veterinary literature. Peer-reviewed studies consistently show that subtle lameness is easy to miss and hard to agree on:
77% of experienced vets disagreed on the presence of mild lameness by eye.
Keegan et al., 2010
Around 64% of lameness was missed by visual observation compared with objective force-plate measurement.
Evans et al., 2005
A dog’s gait naturally varies day to day, so a single observation is a weak baseline.
Nordquist et al., 2011
Stress in the clinic environment changes how a dog moves, masking or exaggerating signs.
Döring et al., 2009
How PAWSCHECK works
- What it measures. From your video it looks at movement asymmetry — stance timing, head and pelvic movement, and joint range of motion across walk and trot — and reports which limb and pattern are most affected.
- Consistency. The AI analysis runs at a fixed, deterministic setting so the same video yields the same measurements, which is what makes tracking change over time meaningful.
- Scoring. Findings are expressed on the familiar 0–5 veterinary lameness scale so your vet can read them at a glance.
- Video quality matters. Confidence depends on the footage. Poor lighting, following shots, or too few gait cycles reduce what can be measured, and the report says so.
Every report is vet-reviewed
No report reaches an owner automatically. Every one is checked by a UK RCVS-registered veterinary surgeon before it is delivered. Review means a vet confirms the affected limb and severity read sensibly against the video, corrects them if not, and can withhold a report that the footage does not support.
Our reviewing vets include Dr Alastair Greenway (BVM&S, MRCVS, ABVA) and Claire Greenway (BVM&S, MRCVS), who also write much of the clinical education on this site.
What PAWSCHECK does not do
- It is a screening and monitoring tool, not a diagnosis. It cannot replace a hands-on veterinary examination.
- It cannot perform a neurological examination. It can only note whether video-visible signs (such as knuckling or scuffing) are present.
- It can produce false positives and false negatives, especially from limited footage.
- It measures movement, not cause. Identifying why a dog moves the way it does requires your vet.
Product validation — in progress
We deliberately do not publish a single headline accuracy percentage, because we do not yet have a robust, independently comparable figure for this specific system measured against a gold standard. Quoting one would overstate what we can currently prove.
What we are doing instead: measuring the consistency of repeated analyses, tracking how often our reviewing vets agree with or correct the AI’s reading, and working towards comparison against objective reference methods. As those results mature we will publish them here, including the parts that are less flattering.
The better your footage, the better the read
Consistent, well-recorded videos are the single biggest factor in a reliable result.
Read the recording guide