Nutrition as Feedback: What AI Misses, According to Culinary Nutritionist Marina Rubel

The market for AI-driven personalized nutrition is expected to hit nearly $5 billion in 2025 and climb to $21 billion by 2034. Algorithms now parse biomarkers, genetic data, microbiome reports, sleep and activity patterns, then generate real-time dietary advice. Continuous glucose monitors—once reserved for diabetics—are now worn by otherwise healthy people tracking their blood sugar after every meal. The technology is impressive. And for a lot of people, it's not quite working.

They get glucose curves, microbiome breakdowns, genotype-based recommendations—and stall. The app suggests changes. They swap out foods. Nothing really shifts.

Marina Rubel, a Ukrainian-born culinary nutritionist practicing in the U.S., runs into this every day. "The data is there," she says. "But between the chart on the screen and what actually happens in someone's kitchen, there's a huge blind spot. The algorithm doesn't cover it."

Rubel operates differently. Not a fixed meal plan, but a loop: observe, form a hypothesis, make a food decision, track the response, adjust—then start again.

Every meal becomes a data point. What happens to energy an hour later? How does the rest of the day go? Do sugar cravings show up in the afternoon? Those signals can be read—if you know what you're looking for.

Years in professional kitchens across the U.K. and Europe gave her a feel for what happens to food before it ever hits the plate—how temperature changes protein structure, how cooking methods affect nutrient bioavailability, why the same ingredients can produce different outcomes depending on how they're prepared. She later earned nutrition credentials (IPHM), which put those same processes into physiological terms. Two frameworks describing the same thing. Once they lined up, the gap between a "correct" diet and an actual result became easier to find.

Where the Algorithm Falls Short

CGM charts, microbiome reports, nutrigenomic guidance, sleep trackers—all of them share one limitation. They deal with what's already happened or what can be directly measured. But between buying food and actually absorbing nutrients, there's a chain of decisions—on the counter, at the table, throughout the day—that no app fully captures.

An app can identify chicken breast on a plate, but it doesn't know if it was cooked at 320°F or 425°F. A microbiome algorithm might call for more fiber, but it won't catch that someone eats spinach raw—where oxalates can bind minerals before they reach the gut. Nutrigenomics might flag a tendency toward vitamin D deficiency, but miss that the diet lacks enough fat for proper absorption. Accurate data. Incomplete picture.

Two people with identical labs and identical diets can end up with very different nutritional outcomes depending on how food is cooked and in what order it's eaten. The system works with what can be measured. The kitchen hasn't been digitized yet.

"A sensor will show you a glucose curve," Rubel says. "It won't tell you why it looks that way, or what to change in the kitchen." The data is an incoming signal. After that, you need someone who knows what to do with it—what question to ask, which detail in a client's routine to check, what single change to test first. Most nutritionists can read labs and build meal plans. Far fewer understand what happens to food between the cutting board and the plate.

How the System Works in Practice

A client wears a CGM for two weeks. The glucose chart repeats: a sharp spike after lunch, then a drop below baseline. That dip brings fatigue, brain fog, a reach for coffee or something sweet.

The algorithm flags the pattern and offers the standard fix: lower the glycemic index of lunch. The client cuts back on rice, adds more vegetables. The curve barely moves.

Rubel reads the same data differently—not the spike, but what comes before it. The client eats lunch at a restaurant near the office. Bread arrives first while the main course is being prepared. He eats it. Then comes the protein. A salad is ordered but ends up last, out of habit. So every lunch starts with fast carbs on an empty stomach. The issue isn't the food. It's the order.

One change: protein and vegetables first, carbs after. Same menu. A week later, the curve flattens. The post-lunch crash softens. The 3 p.m. coffee stops feeling mandatory.

The sensor showed a symptom. Rubel found the mechanism. And the fix had nothing to do with eating differently—just eating the same things in a different sequence.

Where the Demand Is Coming From

Nutrition is moving in a direction medicine took decades ago—from general advice toward testable hypotheses and measurable outcomes. The tools to do that are finally here. Specialists who know how to use them are still catching up.

Rubel is one of the few working at that intersection. Kitchen practice, nutrition science, and patient monitoring data—in her method, these aren't three separate inputs. They're one process. Clients are showing up with charts and the same question: why isn't anything changing? That question now has an answer. Finding someone who can give it is still the hard part.