How to Add Context to Improve AI Accuracy

This article explains how adding a short text description alongside your food photo or log dramatically improves the accuracy of Nutrify's AI nutrition estimates.

Why Context Matters

A photo alone tells the AI what your food looks like, but it cannot always tell the difference between similar-looking items. For example:

• A glass of whole milk and skim milk look identical
• A regular Coke and a Diet Coke look the same
• Fried rice and cauliflower rice are hard to tell apart in photos
• A small bowl and a large bowl can look the same depending on the camera angle

Adding a few words of context solves these ambiguities and can improve calorie accuracy by 20-40%.

How to Add Context

With a Photo

1. Take your food photo as usual
2. Before tapping Save, look for the Add description text field
3. Type a short description
4. Tap Save

With Text/Voice Logging

When you log by text or voice, the description itself is your context. Be as specific as you can.

What Makes Good Context

Here are the most impactful details you can add:

Brand or Restaurant Name

Brand names dramatically improve accuracy because the AI can look up exact nutrition data:

• "Starbucks grande oat milk latte" instead of "coffee"
• "Chick-fil-A spicy chicken sandwich" instead of "chicken sandwich"
• "Kirkland protein bar" instead of "protein bar"

Portion Size

Quantity is one of the hardest things for AI to judge from a photo:

• "About 2 cups of rice"
• "6 oz grilled chicken breast"
• "Small handful of almonds, maybe 15"
• "One large banana"

Cooking Method

How food is prepared changes its calorie content significantly:

• "Pan-fried in olive oil" vs. "air-fried" vs. "baked"
• "With butter" vs. "dry toast"
• "Grilled" vs. "deep-fried"

Key Ingredients

For mixed dishes, calling out key ingredients helps:

• "Pasta with cream sauce, chicken, and mushrooms"
• "Salad with ranch dressing and croutons"
• "Smoothie with banana, protein powder, and almond milk"

Diet-Specific Details

• "Sugar-free" or "diet" version
• "Whole wheat" vs. "white bread"
• "Low-fat yogurt" vs. "full-fat Greek yogurt"

Examples: Good vs. Basic Context

Context with Voice Logging

When using voice logging, just speak naturally with the same level of detail:

• "I had a large bowl of homemade chicken soup with noodles and vegetables"
• "Two slices of pepperoni pizza from Dominos, thin crust"

The AI processes your spoken description the same way it handles typed text.

Tip: You do not need to be exact. Even approximate context like "large portion" or "homemade" is better than no context at all.

When Context Matters Most

Context helps the most with:

• Drinks - Hard to judge calories visually (especially lattes, smoothies, juices)
• Mixed dishes - Stir-fries, casseroles, and soups where individual ingredients are hidden
• Similar-looking foods - Items that look alike but differ nutritionally
• Restaurant meals - Naming the restaurant lets the AI reference their specific menu items

Need More Help?

If you have a question not covered here, the fastest way to reach us is directly inside the Nutrify app:

1. Open Nutrify
2. Go to Settings
3. Tap Help & Feedback
4. Tap Report a Bug or Request a Feature

Our team reads every message. You can also email us at support@nutrifytracker.com.