Google Discover — AI Image Style Guide

Google Discover · Image Strategy 2026

Make Images
They Can’t
Ignore

A complete AI image style guide for Google Discover. The right visual style can be the difference between 2% and 12% CTR — here’s exactly what works and how to prompt for it.

8.9%
Avg Discover CTR
4×
vs. Search CTR
67%
Of publisher traffic
Cinematic Drama
Strategic Contrast
Bold Emotion
Minimal Object
Mystery Curiosity
Before you generate
a single image
01
📐
Minimum 1200px Wide
Google will not show large preview cards without this. Posts with large images get 45% higher CTR. Aim for 1600px at 16:9 ratio.
1600 × 900px ideal
02
🏷️
Max-Image-Preview Tag
Add this to every article page head: <meta name="robots" content="max-image-preview:large"> — Without it, you’re invisible.
Required meta tag
03
📵
No Text Overlays
Unlike YouTube thumbnails, Discover crops images unpredictably on mobile. Text on images gets cut off. Keep the image clean — the headline does the text work.
Discover specific rule
04
Use WebP Format
Smaller file size = faster load = higher Google ranking. Convert all AI outputs to WebP before publishing. JPEG is the fallback minimum.
Speed signal
05
🎯
Single Clear Focus
One dominant subject. No busy backgrounds. High contrast between subject and background. The image must communicate in under half a second on a phone screen.
Visual clarity
06
🚫
No Generic Stock
Smiling people in offices, handshakes, globes — Google and users recognize and ignore stock instantly. AI-generated originals consistently outperform stock imagery.
Originality matters
AI image styles
that stop the scroll
🎬
Cinematic · Drama · Tension
Something big is happening
Style 01 · Best Overall CTR
Cinematic Drama
Dark, moody backgrounds with a single high-contrast subject lit dramatically. Uses film-grade lighting — a singular light source casting deep shadows, creating immediate tension and emotional pull. This style has the highest cross-niche CTR because it triggers an emotional “I need to know more” response in the brain within milliseconds. Works for almost every topic.
News & Current Events Finance Health Politics True Crime Tech
📋 AI Prompt Template
Cinematic editorial photograph, [SUBJECT], dramatic single-source side lighting, deep shadow on opposite side, ultra-sharp focus, dark background with subtle color gradient, high contrast, photorealistic, 8K, shot on 85mm lens, moody atmosphere, editorial quality, no text
👤
The Answer You’ve Been Missing
Style 02 · Editorial Trustworthy
Strategic Contrast Split
Half dark, half light. A person or subject on one side, clean negative space on the other. Instantly communicates authority and credibility. Great for opinion pieces and expert analysis.
Opinion Analysis Interviews
📋 AI Prompt Template
Editorial split-composition portrait, [PERSON/SUBJECT] on left third, dramatically lit against dark background, right two-thirds clean light background with subtle texture, high contrast, professional photography, no text, 16:9
💊
The one thing that changes everything
Minimalist · Clean · Focused
Style 03 · Curiosity Trigger
Minimalist Hero Object
One single object, perfectly lit, centered on a clean or gradient background. The isolation creates mystery — the object becomes the question. Pairs perfectly with curiosity-gap headlines. Deceptively simple, but extremely clickable.
Lifestyle Product Science Food & Health
📋 AI Prompt Template
Product photography, [OBJECT] centered on minimal [COLOR] gradient background, studio lighting from above, ultra-sharp, soft reflection beneath, hyperrealistic, clean composition, no text, no clutter, 16:9 format
😞
Before
🚀
After
Style 04 · Transformation Power
Before / After Transformation
Two distinct visual states side by side. Taps directly into the desire for transformation. Psychologically irresistible — the brain naturally wants to close the gap between the two states. One of the highest CTR styles for self-improvement, finance, and health content.
Health & Fitness Finance Home / Decor Career
📋 AI Prompt Template
Split-panel image, left side showing [NEGATIVE STATE] with muted desaturated tones, right side showing [POSITIVE STATE] with vibrant warm tones, clean center divider, photorealistic, no text, high contrast between panels, 16:9
🕵️
What nobody is telling you…
Style 05 · Curiosity Gap King
Mystery Silhouette
A partially obscured or silhouetted subject in near-darkness, with one dramatic light source revealing just enough to create intrigue. Exploits the information gap — the brain must click to resolve the tension. Pairs with “what nobody tells you” style headlines.
Investigative Secrets & Reveals Conspiracy Deep dives
📋 AI Prompt Template
Atmospheric silhouette photograph, [SUBJECT] in near-darkness, single dramatic backlight creating rim lighting, deep blacks, mysterious fog or haze, cinematic color grade with [COLOR] tones, high contrast, no face visible, no text, film grain, editorial
The 4-step formula
for every image
🎯
1. Match Style to Topic
Pick the style that matches your article’s emotional promise. Drama for news. Minimal for curiosity-gap. Split for authority. Transformation for aspiration. Never use a random style.
🤖
2. Generate with AI
Use Midjourney, FLUX, GPT-Image, or Ideogram. Generate 4–6 variations per article. Always specify “no text, 16:9, photorealistic, editorial quality” in your prompt.
✂️
3. Crop for Mobile First
Discover is 95% mobile. Make sure the main subject is centered or slightly left. No critical detail near the edges. Test at 375px width — if it reads clearly, you’re good.
📊
4. Track and Iterate
Google Search Console shows Discover impressions and CTR per URL. If CTR is below 5%, swap the image. What works compounds over time — keep records of winning styles.
Pick your weapon
🖼️
Midjourney v7
Best Overall Quality
Unmatched cinematic and editorial quality. Ideal for Styles 1, 2, and 5. Use –ar 16:9 flag. Best for dramatic, moody, photorealistic work.
→ Best for: Drama, Mystery, Contrast styles
GPT-Image (GPT-4o)
Best for Specific Concepts
Exceptional at following complex, specific prompts. Great for before/after composites and object isolation. Best text-to-image instruction following available.
→ Best for: Minimalist Object, Before/After styles
FLUX 1.1 Pro
Best Speed + Value
Fast, affordable, and produces excellent photorealistic editorial images. Available via Replicate, fal.ai, or Freepik. Great for daily publishing volume.
→ Best for: High volume publishing, all styles
🎨
Ideogram 3.0
Best Composition Control
Outstanding composition understanding. Great for split designs and structured layouts. Strong at following aspect ratio and framing instructions precisely.
→ Best for: Split Contrast, Minimalist styles
🔥
Adobe Firefly 4
Best for Brand Safety
Commercially safe with full IP indemnification. Best choice if you’re a publisher worried about copyright. Quality is excellent for editorial work.
→ Best for: Publishers needing full IP safety
🌟
Leonardo AI
Best Free Tier
Strong free tier with consistent quality. Good style controls and presets. Use “PhotoReal” and “Cinematic” modes for Discover images. Great starting point.
→ Best for: Getting started, low budget
Your Discover Image Checklist
Run every AI-generated image through this before publishing. One missed item can tank your CTR.
  • Image is at least 1200px wide (1600px preferred)
  • 16:9 aspect ratio — Discover’s native card format
  • Max-image-preview:large meta tag is in page <head>
  • Zero text overlaid on the image
  • Single clear subject — no visual clutter
  • High contrast between subject and background
  • Saved as WebP for fast load times
  • Readable as a thumbnail at 375px mobile width
  • Style matches the emotional promise of the headline
  • Looks original — not like a stock photo