The Load Spike Hiding Inside a Single Long Run

Here is a training week that looks perfectly safe on paper.

Total: 280 minutes. The week before it, you ran 285. Your weekly total dropped by under 2%, and any app that judges you on weekly volume would give you a green tick.

Now look at the Sunday. If your longest run in the previous six weeks was 85 minutes, you just asked your legs, tendons and feet to stay under load for 29% longer than they have in a month and a half. The weekly number hid it completely.

That is the argument of this post: the long run needs its own ceiling, separate from weekly volume. Plenty of AI-built plans don’t have one. If yours doesn’t, you have to supply it yourself.

How a flat week hides a spike

Weekly totals are an average. Averages are good at smoothing things, and that is the problem. Redistribute the same 280 minutes and you can produce very different stress on the body.

Compare two versions of that week:

Week A (spread out)         Week B (concentrated)
Tue   55 min                Tue   45 min
Wed   50 min                Wed   50 min
Thu   50 min                Thu   40 min
Sat   45 min                Sat   35 min
Sun   80 min                Sun  110 min
------------------          ------------------
Total 280 min               Total 280 min
Longest 80 min              Longest 110 min
Longest / total = 29%       Longest / total = 39%

Same total. Week B puts 39% of everything into one session. The weekly ACWR (acute:chronic workload ratio, the figure some platforms derive from a 7-day against a 28-day load) barely moves for either. The single-session picture is wildly different.

Tissue doesn’t care about your weekly sum. A tendon or bone experiences each run as its own event: a number of loading cycles, accumulating fatigue, with the back half of a long run done in a tired, sloppier gait. Minute 95 of a run is not the same as minute 45 with a bit more added. Form degrades, cadence drops, ground contact time creeps up, and the same muscles absorb impact less well. The last 25 minutes of Week B’s Sunday are the riskiest of the lot, and they are exactly the ones that were new.

Where AI plans quietly do this

I see three patterns repeatedly in plans from general-purpose chatbots and, less often, in app-generated ones.

Percentage-of-weekly-volume logic. A prompt like “build me a 16-week marathon plan, long run about 35% of weekly mileage” sounds sensible. But if weekly mileage jumps in a step-back week and recovers, the long run can leap with it. Runna, Garmin Coach and TrainingPeaks-style plans generally handle progression better than a raw ChatGPT output, but they still tend to think in distance, and distance is a poor proxy for duration (see below).

Pace changes that stretch time without changing distance. A 20 km long run prescribed at 6:00/km takes 2h00. If your heat-affected or tired reality is 6:45/km, it takes 2h15. The plan says “20 km”, the app says the same distance, and your body gets 15 extra minutes. For slower runners, a fixed-distance long run is a hidden duration escalator.

The “just add a bit” week after a down week. Step-back weeks end with a jump, often back above the pre-step-back long run, because the plan’s weekly arithmetic says it’s fine.

Worked example: a half marathon plan that looks gentle

A runner (call her Dana) is on a 12-week half marathon plan. Easy pace 6:30/km. Here are her Garmin-recorded long runs and weekly totals.

Week 8’s total is only 2.5% above week 6’s. Nothing in the weekly figure looks aggressive. But the long run went from 98 to 117 minutes: a 19% jump in one session, and 50% longer than the step-back run a week earlier. Her Garmin training status will probably read “Productive” or “Maintaining”. It has no opinion on the fact that she has never run that long.

A rule you can actually apply

I’d suggest two checks, both doable from Strava’s activity list in under a minute.

1. Single-session ceiling: no run longer than about 10% above your longest run in the past 30 days (by duration, not distance). Some coaches stretch it to 15% for well-conditioned runners, but treat that as the upper edge.
2. Share cap: if the long run is above roughly 40% of the week’s total time, question it, unless you’re a low-frequency runner (3 days or fewer a week), where it will naturally be higher. If you run three times a week, a higher share is structural, which is exactly why the absolute-duration ceiling matters more for you.

Run Dana’s numbers through the first rule: her longest in the prior 30 days was 98 minutes, so the ceiling is about 108. Week 8 should have been 105 minutes, roughly 16 km at her pace, not 18.

Run the opening example through it as well: 85 minutes allows about 94. The plan asked for 110. That’s the flag.

What to do when the plan overshoots

You don’t need to throw the plan away. Options, in rough order of preference:

• Cap by time and keep the plan’s intent. Run 94 minutes instead of 110. Add the missing minutes back over the next two weeks if you feel good.
• Split the stimulus. Do a 75-minute run Sunday, then a 40-minute easy run Monday. Total time similar, no single tissue exposure that large.
• Use a run/walk structure for the extra minutes. Past your ceiling, walk 1 minute in every 10. Impact drops, aerobic benefit mostly stays.
• Move the extra effort to a bike or elliptical session if you’re a marathon runner needing time-on-feet fitness without more pounding.

If you’re getting niggles in the shin, Achilles or the outside of the knee in the days after a long run, that’s your cue to be stricter, not looser.

Asking an AI plan the right question

If you build plans with ChatGPT or similar, paste in your last six weeks of long-run durations and ask directly: “What is the percentage increase in my longest single run this week versus my longest in the previous 30 days? Flag anything over 10%.” Most models will do this arithmetic correctly if you hand them the data, and almost none will do it unprompted.

For app users: export or scan your Strava history, and when a Runna or Garmin Coach session appears, check the time estimate, not just the distance. If you are slower than the plan’s assumed pace, recalculate with your real easy pace.

Weekly volume and acute:chronic ratios are useful, and they’re covered in more depth in our guide to training load and injury risk. They answer one question: is this week in line with recent weeks? They are silent on the question that tends to get people hurt: is this single run in line with the longest one my body has recently survived?

Give that question its own number, check it every Sunday morning before you lace up, and let the weekly total take care of itself.