AI Running Series | #1
How I Built My Own Marathon Training Plan in Under 2 Minutes — Using AI
When you set a goal for a marathon — say, finishing under 4 hours — you start running. But somewhere along the way, a quiet anxiety sets in: Am I actually training the right way?
What you really want is someone to look at your current fitness level and map out exactly what to do between now and race day. Turns out, that's exactly what AI can do now.
I fed ChatGPT, Gemini, and Claude the same prompt. All three gave me different results — and one of them became the plan I'm actually running with.
· · ·
I. All You Need Is a Few Basic Facts
The more specific your input, the more personalized the output. Here's what I prepared:
| Item |
My Example |
| Current marathon time |
4 hours 24 minutes |
| Goal time |
Sub 4:00 |
| Target race |
Mid-October |
| Weekly mileage |
~90km (56 miles) |
| Training style |
Easy Run focused, morning runs |
| Heart rate data |
5:40/km pace → HR drifts from 130 to 145–150 after 10km |
| Training principle |
Injury prevention first |
| Days per week |
6 days |
No full marathon time? No problem — a 10K or half marathon result works just as well. For heart rate, any GPS watch will do.
💡 No heart rate data?
That's fine. Just describe what you feel: "I start gasping after 10km" or "I can't hold my pace past 20km." AI picks up on that too.
· · ·
II. The Prompt — Just Swap in Your Own Numbers
Copy the prompt below and replace anything in [ orange brackets ] with your own information. Paste it directly into any AI chat.
Tip: paste it into your notes app first, edit the brackets, then copy the whole thing into the AI.
I'm a recreational runner.
Based on the information below, please build a personalized marathon training plan for me.
📌 Current Status
| Current marathon time |
[4 hours 24 minutes] |
| Weekly mileage |
[~90km / 56 miles, mostly Easy Runs] |
| Heart rate data |
[At 5:40/km pace: HR ~130 for first 10km, drifts to 145–150 after that] |
| Training pattern |
[Morning runs, 6 days a week] |
🎯 Goal
| Goal time |
[Sub 4:00 — 5:41/km pace] |
| Race date |
[Mid-October] |
🛡 Training Principles
| [Injury prevention is the top priority] |
| [Prefer mileage-based approach over intervals] |
| [Please structure the plan in clear Phases] |
📋 What I'd Like
| Break the plan into Phases from today through race day. |
| For each Phase, include weekly mileage, training intensity, and key goals. |
| Give me clear, numeric criteria for moving from one Phase to the next. |
| Present the results in a table format. |
💡 Tip
Try this prompt in all three — ChatGPT, Gemini, and Claude. The results are more different than you'd expect.
· · ·
III. Same Prompt, Three Very Different Results
Gemini
When I asked for a visual follow-up, it generated a full card-style infographic with color-coded phases and bar charts. The most visually polished of the three, and the friendliest tone. That said, it scaled down my weekly mileage significantly — so the output needed a careful review before I could use it.
ChatGPT
The most detailed of the three. Phase breakdowns, day-by-day schedules, intensity ratios, phase-transition criteria, and training principles — all on one page. A follow-up visualization request produced a clean infographic as well. Best for runners who want to see every detail laid out.
Claude
The most concise. Clean structure, just the essentials — easy to scan and act on. If you want to understand the big picture at a glance without digging through details, this one lands best.
All three did a solid job. The difference is in what you prioritize.
I ended up using Claude's version as my main plan — easy to follow at a glance — and referencing ChatGPT's for the finer details.
· · ·
IV. The Input Is What Matters
The more specific your data, the more useful the plan. That was the clearest takeaway.
AI won't assess your body for you.
But it's surprisingly good at turning the information you already have into a structured, actionable plan.
Heart rate data made the biggest difference in my case.
❌ Vague input
"Make me a Sub-4 marathon training plan."
✅ Specific input
"At 5:40/km, my HR drifts from 130 to 145–150 after the first 10km."
To an AI, those are completely different requests. The more specific you are, the more realistic the plan.
What is Heart Rate Drift?
A term even experienced runners sometimes haven't heard
When you run at a steady pace, your heart rate gradually climbs over time — even though your speed hasn't changed. For example: 5:40/km feels manageable at 130 bpm for the first 10km, but by 20km, you're at 150 bpm and starting to struggle. Same pace, harder effort. The more your HR drifts, the more likely you are to fall apart in the second half of a race. Reducing drift is one of the most important markers of marathon fitness.
What is a Tempo Run?
Useful even if you're just getting started
A tempo run sits right at the edge of comfortable — not easy, but not all-out either. You can't hold a conversation, but you could manage a word or two. Think of it as running at a "comfortably hard" pace for 3–5 miles. If Easy Runs build your engine, tempo runs improve its efficiency.
· · ·
V. Want a Visual? Make One More Request
The first response is usually a wall of text. Useful, but hard to scan. One more prompt fixes that:
"Can you turn this training plan into a visual card layout, with each Phase color-coded?"
One heads-up: the visualization step sometimes changes a number or two. Always cross-check the final image against the original text plan before committing to it.
· · ·
I'm currently in Phase 1 of this plan.
My target race is a small local marathon in mid-October — a community event, nothing fancy. Entry fee was about $35. But it has a finish line, and that's enough.
Whether this plan works or not, I'll share the full process — training logs, pace data, and the final result — right here.
