I Used ChatGPT to Train for a Marathon. Here's Why It Failed (And What I Built Instead)

by StrideRun Team7 min read
marathontrainingAIChatGPTrunning coach

1. Introduction

When I trained for my first marathon, I decided to use ChatGPT as my running coach. I didn't want to commit to a human coach yet, and ChatGPT seemed like a smart, free alternative. I asked it for a marathon training plan, pacing guidance, fueling tips—pretty much everything. And at first, it felt amazing. It was like having a knowledgeable coach available 24/7.

But a few weeks into training, cracks started to show. The longer the conversation got, the more things drifted. ChatGPT forgot details, contradicted itself, and eventually became more work than help. I kept thinking: Why is this getting harder instead of easier?

That frustration eventually turned into the idea for StrideRun, a purpose-built AI running coach that actually keeps track of your training and adapts when life happens. But before I talk about that, let me explain why ChatGPT falls short—not because it's bad, but because marathon training requires something it simply wasn't built to do.

Key takeaway:

ChatGPT is great for quick running questions, but it breaks down when you rely on it for months of structured coaching.

2. What Runners Try to Use ChatGPT For

If you're like me, it's obvious why people lean on ChatGPT for running advice. It can help with:

  • Training plans — You can ask, "Create a 16-week marathon plan," and it will instantly produce something structured.
  • Pacing and fueling — It can explain long-run pace, gel strategy, hydration, carb loading—anything.
  • Adjustments when life happens — Sick, injured, missed a long run? You can ask for modifications.
  • Race-day strategy — Goal pace, warm-up, taper tweaks—it knows the theory behind all of it.
  • Recovery guidance — Sleep, stretching, strength training, heart rate zones—it can talk about it endlessly.

It's accessible, free, and incredibly knowledgeable about running science. That's exactly why I started using it. But the deeper you go into training, the more the problems show up.

Key takeaway:

ChatGPT can answer almost any running question, which makes it tempting to treat it like a coach—until you need consistency.

3. The 5 Critical Problems With ChatGPT for Running Training

Here are the exact issues I ran into while trying to use ChatGPT to guide me through months of marathon prep.

Problem #1: It Slowly Forgets Your Training Details

ChatGPT's conversations get messy as they grow. After dozens of messages, it started forgetting my mileage, my goal pace, and which week of training I was in. I eventually had to start new chats just to reset its memory—meaning I had to re-explain every part of my plan.

It's nobody's fault. It's just how ChatGPT works: long chats degrade quality.

Takeaway: ChatGPT can't reliably maintain long-term training context.

Problem #2: It Doesn't Track Your Workouts

ChatGPT has no idea what you actually completed unless you manually type it in. Every time I finished a run, I had to update it. Every time I missed a run, I had to explain the whole situation. There's no calendar, no log, no history. If you forget to tell it something, it never happened.

Marathon training requires weeks of progression and pattern recognition. ChatGPT doesn't actually see that progression—you're just summarizing it for the model.

Takeaway: Without real workout tracking, ChatGPT can only guess at your training needs.

Problem #3: It Gives Inconsistent Advice

Sometimes I tested ChatGPT by asking the same question in a new chat, and the answers were completely different. One day it recommended running long runs 60–90 seconds slower than marathon pace. Another day it told me only 20–30 seconds slower. These inconsistencies are harmless when you're reading casually—but they're a problem when you're training seriously.

Marathon training thrives on consistency. ChatGPT isn't consistent because every new chat is a blank slate.

Takeaway: ChatGPT doesn't have a coaching philosophy—it improvises based on wording and mood.

Problem #4: It Can't Adjust Plans Automatically

When I got sick or missed a long run, ChatGPT couldn't simply look at my plan and make smart adjustments. I had to copy and paste the entire remaining schedule and explain what week I was on.

Real coaching requires structured logic:

  • How much mileage you can safely increase
  • Whether you should skip or replace missed workouts
  • How changes affect taper

ChatGPT can only react to whatever text you provide in that moment.

Takeaway: ChatGPT can't adapt your plan intelligently without a ton of manual effort.

Problem #5: No Accountability or Check-ins

ChatGPT doesn't check in on you. It doesn't ask how your long run went. It doesn't track streaks. It doesn't remind you to recover or adjust something. If you don't initiate the conversation, nothing happens.

During heavy training weeks, this led to gaps. I'd forget to update ChatGPT, feel overwhelmed by having to re-explain everything, and eventually stop using it.

Takeaway: Coaching requires proactive support. ChatGPT is fully passive.

4. The Technical Reason ChatGPT Fails for Long-Term Coaching

The root issue is simple: ChatGPT has a context window, which works like a short-term memory buffer. As the conversation gets longer, it struggles to keep all earlier details available. That leads to:

  • slower responses
  • forgetting your plan
  • mixing up training details
  • declining accuracy

ChatGPT was built for one-off questions, not six months of continuous coaching. It's incredible at general knowledge, brainstorming, rewriting, breaking down concepts—but not at remembering hundreds of tiny training details week after week.

Takeaway:

ChatGPT isn't bad; it's just designed for short, isolated interactions—not long-term coaching.

5. What Actually Works: Purpose-Built AI Coaching

After struggling through this for months, I realized the problem wasn't AI—it was using the wrong type of AI. So I built StrideRun, a running coach designed from the ground up for long-term training.

Here's the difference.

Stateless architecture

Instead of rereading your entire chat, StrideRun stores your plan, progress, and history separately. The AI already knows:

  • your current training week
  • your completed workouts
  • your goals
  • your preferences

Every chat is fresh, but the context is complete.

Integrated tracking

Your runs are logged in a clean, Notion-style calendar. You don't have to manually re-explain anything.

Automatic plan adjustments

If you miss a run, get sick, travel, or feel fatigued, the AI adjusts your plan based on structured training logic—not guesses.

Weekly check-ins

StrideRun actually asks how your training is going and adapts accordingly.

This isn't something ChatGPT can do, because it wasn't designed to function as a persistent coach.

Takeaway:

AI coaching works beautifully when the system is designed specifically for running, not general conversation.

6. Comparison Table

| Feature | ChatGPT (Free) | Strava | Purpose-Built AI Coach (StrideRun) | |---------|----------------|--------|-----------------------------------| | Cost | Free | $11.99/mo | Free + premium options | | Knows your plan | ❌ | ❌ | ✅ | | Tracks workouts | ❌ | ✅ (logging only) | ✅ (logging + coaching) | | Suffers context issues | ✅ | ❌ | ❌ | | Automatically adjusts plan | ❌ | ❌ | ✅ | | Weekly check-ins | ❌ | ❌ | ✅ |

Takeaway: Only purpose-built AI combines tracking, memory, and intelligent coaching.

7. Conclusion

Using ChatGPT to train for a marathon taught me a lot—not just about running, but about how AI works. It helped answer questions, but it couldn't coach me long-term. It couldn't remember my training, adjust when life got messy, or keep me accountable.

Marathon training is too complex and too personal for a tool that forgets almost everything you tell it.

That's why I built StrideRun—to solve the exact problems I ran into. If you're preparing for a race and want a coach that actually knows your plan and adapts with you, give it a try.

You can start free at striderun.app .

Related Posts