So, Did It Work ?

15 September 2026

Missed Part 1? 

Read how I asked ChatGPT to build my 14-week half-marathon training plan — and why I decided to give AI control of my training.

Fourteen weeks ago, I started an experiment.

Could ChatGPT actually help an experienced everyday runner train for a half marathon?

I gave it my running history, my recent race results, the number of days I could run, my target time and a few other bits of information.

It gave me a training plan, and I followed it for 14 weeks.

Well, mostly.

There were a few missed paces. A few tired days. A couple of runs that got dropped when I thought recovery was more important. And, rather embarrassingly, nowhere near enough strength training.

The training is done.

And on Sunday(13 September 2026), I lined up at the Great North Run to find out whether all those miles, intervals, tempo runs and long Saturday mornings had actually made me a better runner.

My target was 1:52.

So, what happened?

The Great North Run

There's something slightly strange about spending 14 weeks training for one particular day.

You spend weeks looking ahead to it.

You know roughly what shape you think you're in. You know the pace you want to run. You've done the long runs. You've done the faster sessions. You've tapered.

And then suddenly you're standing on the start line.

This year felt slightly different.

For the past nine years, I've run the Great North Run with my brother. This year was the first time I'd be doing it on my own.

Normally, I also get a great night's sleep before the race.

This year? Not so much.

I woke up what seemed like every hour and, every time I did, went to the loo. Whether that was nerves, the subconscious fear of sleeping through an alarm now that I didn't have my usual race-day companion alongside me, or simply one of those nights, I don't know.

The strange thing was that my energy levels felt good.

When I left the hotel, the weather was actually perfect. Nice and cool as I made my way to the Metro.

I was nervous, as I always am before a race, but I felt ready.

And then the weather changed.

The atmosphere in the starting pen was, as always, incredible. The Red Arrows flew overhead to celebrate the 45th Great North Run, and even after ten years of doing this race, the sheer size of the event still blows my mind.

But while we were standing in the pen, the clouds disappeared.

Clear blue sky.

Blazing sunshine and heat.

I even took the top off that I was wearing, put it on the floor and sat on it in the pen, using the runners standing around me to provide some shade.

And I was still hot.

It was while standing there that I made a decision.

Plan A was no longer the plan.

I had a Plan B.

Plan A becomes Plan B

My target was 1:52, which meant averaging roughly 8:33 per mile.

The original plan was to settle into that sort of pace and see how things developed.

But standing in that starting pen, feeling the heat building, I knew I had to be realistic.

So I changed the plan.

Plan B was simple: run the first eight miles at target pace, but don't force it beyond that. Then, from mile nine onwards, take my foot right off the gas.

Forget the pace.

Forget 1:52.

Just run, enjoy the Great North Run and make sure I got under my back-up target of two hours.

It wasn't the plan I'd spent 14 weeks training for.

But I believed it was the right decision for the conditions.

And that's the thing with running.

You can train for months. You can have a perfectly structured plan. You can know exactly what pace you want to run.

But you can't control the weather.

I don't perform well in the heat. I never have.

Some people can run brilliantly when it's hot.

I'm definitely in the can't category.

Then we were off.

How did it go?

The first few miles went pretty much according to Plan B.

The pace was tough, but it was within my control.

Mile 1 — 8:13

Mile 2 — 8:28

Mile 3 — 8:22

Mile 4 — 8:55

Mile 5 — 8:51

Mile 6 — 8:27

Mile 7 — 8:42

Mile 8 — 8:58

I wasn't cruising, but I wasn't hanging on either.

The training had done its job.

The faster sessions had made running at a decent pace feel more comfortable, and one of the biggest things I'd noticed over the 14 weeks was that I was becoming more comfortable with being uncomfortable.

To me, that's a sign of progression.

But as I got closer to mile nine, the heat was becoming harder to ignore.

And I stuck to the decision I'd made before the start. 

Plan B.

Foot off the gas.

Stop worrying about the watch.

Just enjoy the race.

The difference was noticeable.

My pace dropped, but so did the pressure.

Mile 9 was 10:48.

Mile 10 — 10:08.

Mile 11 — 9:42.

Mile 12 — 10:02.

I wasn't racing the clock anymore.

I was racing the conditions.

PLAN B - PART 1

Miles 1–8

~8:33/mile

↓

PLAN B - PART 2

Miles 9–12

Relax. Enjoy.

Sub-2 target.

↓

FINAL PUSH

Mile 13

8:56/mile

And, more importantly, I was enjoying the experience again.

The closer I got to South Shields, the hotter it seemed to become. But knowing I didn't have to force the pace made it much more manageable.

My only remaining target was two hours.

If I'd finished over that, after making the decision to abandon the 1:52 attempt, I would have been genuinely disappointed.

So the final mile became a different kind of race.

Forget 1:52.

Forget the watch.

Get under two hours.

With the finish line getting closer, I started to push again.

The final mile was 8:56.

I felt strong and started to pick up the pace as I got closer to the finish.

Then came the bit that makes the Great North Run what it is.

Prince Edward Road.

The turn onto the seafront.

That final mile into South Shields.

The crowds were incredible, as they always are. They make you feel like a superhero. The closer you get to that finish line, the louder they become.

And eventually, I crossed it.

My watch showed 1:59.

I was happy.

I'd got under two hours.

Then I saw my official time 1:58.

That made me even happier.

I'd stuck to the decision I'd made in the starting pen, adapted to the conditions and, crucially, achieved my back-up target.

But had I hit 1:52?

No.

And was I disappointed?

Of course I was.

I'm a runner.

We're always disappointed when the plan doesn't quite go the way we wanted it to.

But disappointment isn't the same thing as failure.

And I don't do excuses.

The easy thing would be to say, "I would have run 1:52 if it hadn't been so hot."

But I'll never know.

Could I have run 1:52?

Maybe.

Maybe not.

That question is now in the past, and there's no way of answering it.

What I do know is that I made a decision based on the conditions I was facing on the day, and I believe it was the right decision.

You can't control things that are out of your control.

The weather was one of them.

But did ChatGPT actually do it?

This is where things get a little more complicated.

If I'd simply wanted to prove that ChatGPT could produce a training plan, I could have stopped 14 weeks ago.

The plan looked sensible.

The paces were sensible.

The progression made sense.

It looked remarkably similar to the sort of plan I'd have written myself.

But a good-looking training plan doesn't prove anything.

The interesting bit was whether following it actually helped me on race day.

And, having now run the Great North Run, my answer is:

Yes. I think it helped.

But I certainly don't think I can say that ChatGPT was responsible for my result.

That's far too simplistic.

I was the one who ran the miles.

I was the one who turned up for the sessions.

I was the one who made the decision to back off when I was overly tired.

And I was the one who stood on the start line on Sunday.

The AI didn't do any of that.

What it did was give me a structure.

And that structure was good enough for me to follow for 14 weeks.

For an experienced runner, I think that's potentially quite valuable.

What would I change?

Interestingly, one of the biggest things I'd change has nothing to do with ChatGPT.

I'd do the strength training.

I said I was going to do it.

The plan included it.

I knew it was important.

And then I largely ignored it.

That's on me.

I don't know whether the lack of strength training affected Sunday's race. I can't honestly say that it did.

But I do know that I would have been better off being more consistent with it.

I'd also probably drop a few more of the easy or recovery runs when I was feeling particularly tired.

Not the effort sessions.

Not the long runs.

But some of the easy miles.

The important sessions were the ones I wanted to protect, and there were a few occasions during the 14 weeks when I could have listened to my body a little earlier.

But that's perhaps another interesting part of this experiment.

The plan doesn't have to be perfect to be useful.

No training plan is perfect.

A coach doesn't get everything right. An app doesn't get everything right. And I certainly don't get everything right when I write my own.

The important thing is whether the plan is sensible, realistic and appropriate for the runner.

For me, it was.

Would I use ChatGPT again?

This is probably the question I've been asked most since starting this experiment.

And now that I've actually completed it, I think I have a much better answer.

Yes.

But with an important qualification.

I wouldn't hand over responsibility for my running to AI.

I wouldn't blindly follow whatever it told me.

And I wouldn't recommend that an inexperienced runner simply asks ChatGPT for a plan and assumes everything it produces must be right.

But for an experienced runner who understands their own training, understands the difference between normal fatigue and something that isn't right, and has a reasonable idea of what they are capable of?

Yes. I think ChatGPT can be a very useful tool.

It can save you the time of writing a plan.

It can explain why you're doing a particular session.

It can help calculate training paces.

It can give you ideas.

And, as I've found over the last 14 weeks, it can produce a surprisingly sensible training structure.

But the runner still has to be part of the equation.

So, what's still possible?

That's where this whole experiment started.

After running for around 15 years, I wanted to know what this version of me could still do.

Not the 2019 version.

Not the 1:43 version.

This version.

A runner in his 50s who still enjoys running, but who has perhaps become a little more appreciative of recovery, easy miles and the fact that you don't have to smash every session to get something out of running.

My target was 1:52.

I ran 1:58.

On the drive home, I was disappointed.

But by the following day, having had time to reflect properly and taking the conditions into genuine consideration, I was ultimately pleased.

I hadn't run the race I planned to run.

But I'd run the race that the day gave me.

And I'd made what I believe was the right decision when it mattered.

I wasn't chasing 1:43.

That was never the point.

The point was to see what I could still do if I gave myself a proper training block and committed to it.

And now I know.

As for ChatGPT?

Well, I still don't think the robot is a running coach.

But after 14 weeks, I've got to admit something.

It knows its running.

Just don't expect it to do the actual running for you.

And me?

Well, maybe there's still a little more left in these legs yet.