Does Running Form Actually Cause Injury?

Short answer: much less than the running-shoe-and-video-analysis industry would like you to believe, and much less than your plan’s workload does. If you’re following Runna, Garmin Coach or a ChatGPT-built schedule and you’re wondering where to spend your worry, spend most of it on the numbers in your training log, not on your knees-over-toes angle.

What the evidence says about form and injury

Runners love a clean explanation. Heel striking, overstriding, a dropped hip, a crossover gait: each has been blamed for shin splints, runner’s knee or Achilles trouble. The trouble is that when researchers follow large groups of runners forward in time, the form-to-injury link is weak and inconsistent.

A few findings worth knowing:

• Foot strike. Daoud and colleagues (2012, Harvard cross-country team) found rearfoot strikers had roughly twice the yearly rate of repetitive stress injury compared with forefoot strikers. That is the study people cite. But it was a single team, and later work, including a large Dutch cohort, found no reliable protective effect from changing strike pattern. Forefoot strikers just move the stress to the calf and Achilles.
• Prospective gait studies. Reviews of prospective biomechanics research (Ceyssens et al., 2019, American Journal of Sports Medicine) found only a handful of kinematic and kinetic variables with any association with injury, and the results were mixed by injury type. Nothing close to a “bad form” checklist emerged.
• Retraining trials. Gait retraining can change a variable (cadence, loading rate, hip adduction). Evidence that it prevents injury in healthy runners is thin. Evidence that it helps some people with existing pain, such as patellofemoral pain, is better.

Form differs hugely between runners who never get hurt. Elite runners overstride, cross over and land on heels. Bodies adapt to the pattern they repeat. What breaks them is usually that something changed too fast.

Why load history is the better predictor

The stronger signal comes from training load, specifically how quickly it changed relative to what the body was used to.

• Nielsen and colleagues (2014, JOSPT) followed novice runners and found those who increased weekly distance by more than 30% between sessions or weeks had a clearly higher rate of injury than those who increased by under 10%, particularly for distance-related injuries.
• Previous injury is consistently one of the strongest predictors in the literature. If you were hurt in the last 12 months, your risk is higher regardless of how you land.
• Spike research in team sport (Gabbett’s acute:chronic workload ratio work) popularised the idea that a week far above your recent average is risky. The ratio itself has been criticised statistically, and I wouldn’t treat 1.5 as a magic line, but the underlying idea survives: sudden jumps hurt.

Notice that none of this needs a camera.

A worked example: reading your own plan

Say a ChatGPT-built marathon plan gives you this (km per week, with your last four weeks of actual running before it started):

Week 1 of the plan is the problem. You jumped from 32 to 42 km, and 42 against a 29 km average is a 46% bump. The form of your stride has not changed between week 4 and week 5. Your tissue just got asked for 10 extra kilometres it hadn’t been prepared for.

A quick sanity check you can do in a spreadsheet or by eye in Strava’s weekly view:

this_week_km / average_of_last_4_weeks_km

< 1.10   comfortable
1.10-1.30 fine if you feel good and the week after isn't also a jump
> 1.30   ask why, and consider capping it

These are rules of thumb, not validated cutoffs. They are still more useful than a gait score.

Also check intensity, not just distance. Runna and Garmin Coach often add intervals or tempo sessions in the same week as a long-run extension. If the long run goes from 16 to 20 km and you add a second hard session, you’ve stacked two loads. Moving one of them a week is a free fix.

A few other things to look for in an AI plan:

1. No cutback weeks. A good plan drops volume about every third or fourth week (roughly 20 to 30%).
2. Long run share. If your long run is more than about 40% of your weekly distance, a single run carries too much of the load. At 30 km a week, a 14 km long run is 47%.
3. Ignoring your actual data. ChatGPT plans in particular are built from what you typed, not from what your Garmin recorded. If you said “I run 40 km a week” and you actually averaged 25, the plan starts in the wrong place.

So is form work pointless?

No, but be precise about what it’s for.

Running economy. This is where form changes have the most credible payoff. Cadence, vertical oscillation and ground contact time are related to how much oxygen you burn at a given pace. Small gains of 1 to 3% are plausible from drills, strength and plyometric work. That’s a real race-day effect, not an injury shield.

Specific symptoms. If you have persistent pain at a particular site, look at how you move. Patellofemoral pain is the clearest case: raising cadence by around 7.5 to 10% has been shown to reduce knee joint load, and hip-focused retraining has helped some runners with pain. Here you’re treating something that already exists, with a mechanical reason to think it matters.

Obvious, extreme asymmetry. Garmin’s ground contact time balance can show a gap. If your watch (Forerunner 265, Fenix and similar with a running dynamics pod or wrist estimate) shows 53/47 on every run, that’s within normal variation. A swing to 60/40 after a calf strain is worth noticing.

What form work isn’t is general injury insurance. Spending six weeks changing your strike pattern while your weekly mileage jumps 35% is trading a speculative risk for a proven one. Changing strike pattern itself is a load: calves and Achilles take on new demand, and runners who switch abruptly often pick up calf or foot problems.

If you do want to look at your mechanics, phone video is a reasonable low-cost start for spotting something specific such as a pronounced hip drop or a very low cadence. Our guide to AI gait and form analysis from phone video covers what the apps can and cannot reliably tell you.

What I’d do with this

For a recreational runner on an app-generated plan, in order of priority:

1. Audit the load. Compare each planned week to your real recent average. Cap jumps at around 10% where you can, and push back on any week above 30%.
2. Respect history. If you had an injury in the last year, start the plan lower and ramp slower than the app suggests.
3. Strength train twice a week. Calf raises, single-leg squats and hip work have better trial support for injury reduction than technique changes.
4. Use form for a reason. Work on cadence or economy drills if you want to run faster, or if a specific symptom points you there. A cadence nudge of about 5% is a modest, low-risk experiment.
5. Leave your stride alone otherwise. If you’re healthy and progressing, your current form is the one your body has already adapted to.

The camera can wait. Open your weekly totals first.