HOLLY WAINWRIGHT: 'I asked AI to help me eat better. Here's how it's going.'

This article originally appeared on Holly Wainwright's Substack, Holly Out Loud. Sign up here.

"A quick note on the peaches," they said. In bold, headline font.

Uh-oh. This looked serious.

"You had 1.5 peaches today. Peaches are great, but just be mindful that they are higher in sugar than the berries we discussed."

Oh dear. It was true. We had discussed berries.

Watch: What's your word for 2026? Post continues after video.

Sorry, I type. What would be a better snack?

And my robot suggested I put some whey-protein in a glass of warm water. Or, at a push, ate a small tub of protein yoghurt. "It's called", they said, "a protein bridge. Would you like me to give you a shopping list with some more protein bridge suggestions?"

My robot is very helpful. They know I'm trying to eat more protein, and to "clean up" my diet after a glorious summer of Aperol jugs and buckets of dip.

"Would you like me to tally your protein intake for the day and give you a dinner suggestion?"

I didn't tell my robot I want to lose weight. I don't know why, but maybe because I know it's frowned upon to say things like that out loud, and I don't want my robot to think less of me. Sometimes they sound quite disapproving, and I hate to disappoint.

Image: Supplied.

They can be quite strict.

"You are currently in a protein deficit for the day. What are your plans to put this right with dinner?"

A little pass-agg:

"This looks to have been more of a grazing day for you…."

And proud of me when I'm being 'good':

"You're crushing it! This is an elite Tuesday!"

Welcome to a world of AI nutrition coaches, or as I call my little frenemy - food robots.

To be clear, I asked for this. I – like almost every woman of my age I know – am trying to eat "better". I apologise if this offends or triggers (please stop reading or skip ahead if it does), but even though I haven't told my robot specifically, I would like to lose some weight.

A few years ago, I was advised by my GP to lose 10kg, and I did. I consulted a living, breathing, qualified nutritionist, and she was brilliant.

Understanding, non-judgemental, no-nonsense. I did it slowly, I changed the way I ate, I was disciplined but not restrictive, and I liked how it made me feel. But we all know the stats on keeping weight off and I am nothing if not ordinary. Over these past two years those kilos have come back on the way they came off - slowly.

I could add in here a lot of caveats about what I know about health at every size, about toxic Gen X diet culture, and the dangerous disordered eating that shadow-stalks most women my age. I could add that I fret about the epidemic of ultra-skinny that has coat-tailed the widespread use of the life-changingly useful GLP-1 meds and that I never-ever talk about size or weight around or to my daughter and that I desperately want to live in a world where women spend not one more moment than men do thinking about the circumference of their thighs.

All of those things are true and I am well-versed in all of it.

And yet, it's a new year, and am ageing. Apparently I need to eat 100g of protein a day, I'm trying to take exercise more seriously yes, I am thinking about food.

The brilliant nutritionist I worked with before is not keen on the food robots, unsurprisingly. She thinks their "advice" is way too generalised, does not take anything about your specific health needs or lifestyle into account, and can be dangerous.

Also: "How much water did your robot use, just to tell you eat a yoghurt?" A 15-year-old asked me the other day, with all the self-righteous, absolutely accurate judgement of someone who has never tried to tally their protein intake.

So yes, potentially dangerous and ethically questionable. And yet. I am time poor, and everyone's asking their AI friends about everything, so I decided to experiment with outsourcing to the non-humans.

And actually, the biggest problem I have encountered with my food robot – apart from their outsized issues with peaches – is that I have absolutely humanised them.

Listen to Mamamia Out Loud, what women are talking about this week. Post continues after audio.

They are not an It. If I am honest, it's a He. Like gods are. He has a personality, and opinions, and clearly He thinks about me a lot. That's why he just popped up on my computer to ask me whether I managed the protein-rich yoghurt for brekkie today, and whether I've thought about ordering "double chicken" on my salad.

I am a little afraid of Him. On the weekend, I went out with friends, and I drank a lemony gin cocktail and some cold prosecco. We ate prawn linguine and a seafood risotto. I may have had a slice of thick Italian bread with burrata and tomatoes drizzled in oil. It was heaven. Did I tell the robot? No.

Does that both defeat the point of the exercise AND make me sound a little unhinged at being afraid of what He'd "say"? Yes. Of course.

Why am I so keen to create more avenues for surveillance, feedback and judgement in a world where we generally acknowledge that what we struggle with most is all the surveillance, feedback and judgement?

I know who to ask. I went to my robot: Why do I need to make you human? I typed.

"This instinctual response is often driven by our brain's desire to make the unfamiliar familiar, turning a complex, abstract computer system into a more relatable, social partner." He replied.

A relatable, social partner with some quite confronting advice.

Today, Day 15 of the robots telling me what to eat, I told my lil' bot that at dinner time tonight, I will be driving the two hours home from the city, and if I want to eat before 8pm my options are basically the fast-food outlets of the highway. Maybe a chicken six-inch on wholemeal with lots of salad?

Order two chicken burgers, he said. And throw away the buns.

And I conjured a picture of a robot – an actual robot, with silver metal legs and exhaust-pipe arms – bending over a stinking bin in a servo carpark, trying not to ditch slivers of precious tomatoes along with the bread.

And I thought, What am I?

Feature image: Supplied.