Performance Max and AI: How Google Merchant Center is Revolutionizing E-commerce in 2026

Introduction

Artificial intelligence has taken control of Google Shopping. Performance Max (PMax) is now the only available format for creating new campaigns, and its brain — Google's AI — feeds directly on your Merchant Center data to make every decision: who to target, when to bid, which format to use, which product to highlight.

The result: merchants who understand how the AI reads and uses their feed achieve 2 to 5x better performance than those who treat PMax as a black box.

This guide explains exactly how the AI-GMC relationship works in 2026 and how to get the most out of it.

---

Table of Contents

• 1. Performance Max — What It Really Is
• 2. How AI Reads Your GMC Feed
• 3. Asset Groups — The Structure That Changes Everything
• 4. Audience Signals — Guiding AI Without Constraining It
• 5. The Feed as a Creative Lever for AI
• 6. Custom Labels — Intelligent Segmentation
• 7. Google Merchant Center Intelligence
• 8. Organic Shopping and SGE — The Other Side of AI
• 9. Measuring Performance: What Really Matters
• 10. Advanced Strategies for 2026
• Performance Max and AI FAQ

---

1. Performance Max — What It Really Is

Performance Max is not an improved Shopping campaign. It's an AI-powered advertising architecture that delivers your ads across the entire Google network:

Channel	Generated ad type
Google Search	Dynamic text ads
Google Shopping	Product listings
YouTube	Video ads
Gmail	Native display ads
Discover	Image ads
Display Network	Display banners
Google Maps	Local inventory ads

One PMax campaign, all channels. The AI decides in real-time which channel, which format, which user, which moment — based on your assets and GMC data.

Why This Is a Fundamental Change

Before PMax, you controlled:

• Targeted keywords
• Bids per keyword
• Placements
• Audiences

With PMax, you delegate these decisions to the AI. In exchange, you get:

• Maximum reach across all formats
• Continuous 24/7 optimization
• Access to intent signals you didn't have (YouTube, Gmail, Maps behaviors)

The trade-off: if your input data (GMC feed, assets, audience signals) is mediocre, the AI amplifies mediocrity. Feed quality has never been more critical.

---

2. How AI Reads Your GMC Feed

The PMax algorithm treats your GMC feed as its primary source of knowledge about your products. Here's exactly what it analyzes:

Semantic Extraction from the Title

The AI breaks down your title into semantic entities:

"Nike Air Max 270 Men's Sneakers Black Size 9"
→ Brand: Nike
→ Model: Air Max 270
→ Type: Sneakers
→ Gender: Men's
→ Color: Black
→ Size: 9

These entities determine which search queries your product will appear for. A title that's too short or too generic reduces the range of matched queries.

Feed — Landing Page Correspondence

The AI crawls your product page and compares:

• Page title vs. feed title
• Displayed price vs. feed price
• Displayed availability vs. feed availability
• Schema.org structured data vs. feed attributes

High consistency = high quality score = lower bids and more visibility.

Product Quality Score

In 2026, each product receives an internal quality score (not visible in the interface) based on:

Criterion	Estimated weight
Attribute completeness	25%
Image quality	20%
Feed/page consistency	20%
Title richness	15%
Description quality	10%
Valid GTIN presence	5%
Product reviews	5%

Products with a high score receive more impressions at lower CPC. Feed optimization is directly correlated to campaign profitability.

---

3. Asset Groups — The Structure That Changes Everything

A PMax asset group is a set of assets (images, texts, videos) associated with a product group. This is the level where you influence the advertising message.

Recommended Structure for E-commerce

PMax Campaign [Main Brand]
├── Asset Group 1: Bestsellers (high target ROAS)
│   └── Listing Group: custom_label_0 = "bestseller"
├── Asset Group 2: New arrivals (exploration budget)
│   └── Listing Group: custom_label_1 = "new"
├── Asset Group 3: Sales (limited time)
│   └── Listing Group: sale_price active
└── Asset Group 4: Long tail (secondary products)
    └── Listing Group: all others

Required Assets per Group

Type	Minimum	Recommended
Short headlines (30 chars)	3	5
Long headlines (90 chars)	1	3
Descriptions	2	4
Landscape images 1.91:1	1	5
Square images 1:1	1	5
Logo	1	2
Video (if available)	0	1-3

Critical advice: Never put all your products in a single asset group. The AI doesn't have enough signal to properly optimize very different products together.

Text Assets — What the AI Does With Them

The AI automatically combines your headlines and descriptions to create text ads. It tests combinations and learns which ones perform best in context.

Best practices:

• Each headline must work independently (the AI mixes them)
• Vary the angles: benefit, urgency, social proof, feature
• Avoid redundancy between short and long headlines
• Include your main keyword in at least 2 headlines

---

4. Audience Signals — Guiding AI Without Constraining It

Audience signals in PMax are not targeting in the strict sense — they indicate to the AI a starting point, which it quickly moves beyond.

Available Signal Types

First-party segments:

• Customer Match lists (existing customers, cart abandoners)
• Conversion data from your site (via Google Tag)
• Google Ads remarketing lists

Google segments:

• In-market audiences (active buyers in your category)
• Affinity audiences (long-term interests)
• Custom audiences (based on keywords or URLs)

Optimal Signal Strategy

Priority signal 1: Purchasing customers (Customer Match, last 180 days)
Priority signal 2: Cart abandoners (remarketing, last 30 days)
Priority signal 3: Product page visitors (remarketing, last 7 days)
Expansion signal: In-market [your main category]

The AI uses these signals to identify similar profiles. In 2026, accounts with rich first-party data outperform accounts without signals by 40%.

Customer Match — Why It's Critical

Since 2026, Customer Match is integrated directly into Merchant Center (no longer need to go through Google Ads separately). You can sync your CRM and automatically create segments:

• High-LTV customers (exclude from acquisition campaigns, target for upsell)
• Inactive customers (reactivation with special offers)
• Category buyers (cross-sell on complementary categories)

---

5. The Feed as a Creative Lever for AI

In 2026, the GMC feed doesn't just power product delivery — it powers ad creation.

Dynamic Product Ads 2.0

The AI automatically generates display and Discovery ads using:

• The main image (image_link) and lifestyle images (additional_image_link)
• The title (title) and highlights (product_highlight)
• Price and availability for dynamic overlays
• Brand and product rating for reassurance elements

What this means: your additional images and product_highlight appear directly in dynamic ads without you having to create manual assets.

Optimizing Images for AI

Google's Computer Vision algorithm analyzes each image and extracts:

• The main product and its context
• Visual quality (sharpness, lighting, composition)
• Generated emotion (neutral, desirable, functional)

Concrete recommendations:

• Main image: centered product, white or solid background, high resolution
• 2nd image: different angle (45°, back, detail)
• 3rd image: lifestyle (product in use)
• 4th image: technical detail or material close-up
• 5th image: size/scale comparison if relevant

AI-generated lifestyle images are accepted since 2026 — tools like Google Product Studio (integrated in Merchant Center) allow generating lifestyle contexts directly in the interface.

---

6. Custom Labels — Intelligent Segmentation

Custom labels (custom_label_0 to custom_label_4) are your main control tool in PMax. They allow you to segment your catalog to apply different bidding strategies.

Recommended Custom Labels Framework 2026

custom_label_0 — Performance

bestseller     (top 20% of revenue)
performer      (top 50% of revenue)
long_tail      (the rest)

custom_label_1 — Margin

high_margin    (> 40% margin)
standard_margin (20-40%)
low_margin     (< 20%)

custom_label_2 — Seasonality

perennial      (year-round)
summer_2026    (seasonal summer products)
jan_sale       (January sales)
holiday_2026   (Christmas specific)

custom_label_3 — Stock Status

high_stock     (> 100 units)
limited_stock  (10-100 units)
last_units     (< 10 units)

custom_label_4 — Newness

new            (< 30 days)
established    (30-180 days)
mature         (> 180 days)

Strategic Application

Bestseller + high_margin → Aggressive target ROAS (priority budget)
New + high_margin → Moderate ROAS (learning phase)
Long_tail + low_margin → Capped max CPC or exclusion
Last_units → Reduced budget or exclusion if imminent stock-out

---

7. Google Merchant Center Intelligence

Since 2026, Merchant Center includes an "Intelligence" section providing insights directly drawn from Google's AI.

Price Benchmark

The AI compares your prices to competitors for each GTIN-identified product. You see:

• Your price position (above / at median / below)
• Estimated impact on your impressions
• Products where lowering price would increase volume

Tip: products "above benchmark" lose an average of 45% of impressions compared to the median. For high-margin products, this is an arbitrage to evaluate.

Demand Forecast

New 2026 feature: Google predicts demand for your product categories over the next 4 weeks, based on search trends. Use it to:

• Anticipate stock-outs during peaks
• Increase budgets before demand peaks
• Prepare your promotions at the right time

Competitive Visibility

This report shows your impression share vs. competitors by category. It identifies segments where you're losing visibility and proposes corrective actions (price, product data, budget).

---

8. Organic Shopping and SGE — The Other Side of AI

Performance Max maximizes your paid ads, but Google's AI also influences your organic presence.

Free Listings — Maximizing Free Reach

Free listings use the same feed as your PMax ads. In 2026, they appear in:

• Shopping tab
• Web results (product carousels)
• Google Images
• Google Lens
• Google Maps (for businesses with local inventory)

A well-optimized product generates free impressions in addition to its paid impressions. The marginal ROI is infinite.

SGE (Search Generative Experience) and Product Citations

Google's generative AI cites products in its responses to purchase questions. In 2026, approximately 23% of commercial queries generate a product citation in SGE.

Selection criteria to be cited:

1. Feed completeness: all required attributes filled in
2. Domain trust score: clean approval history
3. Product reviews: average rating and number of reviews
4. Semantic relevance: match between the query and your product data
5. Competitive price: compared to other matched products

For SGE:

• Fill in aggregate_rating or activate Google Customer Reviews
• Ensure your product_highlight answers common questions
• Maintain a feed with 0 disapprovals (trust score accumulates)
• Verify that Schema.org structured data is consistent with the feed

GEO (Generative Engine Optimization) for E-commerce

GEO is optimization for AI engines — Perplexity, ChatGPT Search, Google SGE. In 2026, a growing share of purchase journeys starts with a question posed to an AI.

GEO principles for product listings:

• Direct answer: your title and description must directly answer "what is the best [product] for [use]"
• Facts first: start your description with key specs, not marketing copy
• Clear structure: use lists, precise technical data
• Authority: mention certifications, labels, quality proof
• Freshness: regularly update your descriptions (AI engines favor recent content)

---

9. Measuring Performance: What Really Matters

Key PMax Metrics to Monitor

Metric	What it measures	Good value
ROAS	Revenue / Spend	Depends on your margins
Conversion rate	Clicks → Purchases	> 2% e-commerce
Average CPC	Cost per click	Monitor the trend
Asset performance	Asset score	"Good" or "Best"
Product partition insights	Performance by product group	Identify outliers

GMC Diagnostics to Cross-reference with PMax

Each week, cross-reference:

1. Feed approval rate (via GMC) with PMax impressions (via Google Ads)
2. Disapproved products with products with no impressions → these are often the same
3. Product quality score (via MyGoogle audit) with ROAS by category

The PMax Learning Phase

PMax requires a minimum 6-week learning phase. During this period:

• Don't modify target ROAS by more than 15% at once
• Don't touch listing groups
• Leave enough budget for the AI to have data (minimum 50 conversions/month)

Common mistake: intervening on a PMax campaign after 2-3 weeks because results aren't there yet. The AI needs time to learn your conversion patterns.

---

10. Advanced Strategies for 2026

Strategy 1 — PMax Campaign + Separate Brand Campaign

Create a separate search campaign for your brand terms with higher priority. This prevents PMax from absorbing easily attributable brand conversions and artificially inflating your ROAS.

Strategy 2 — Segmentation by Product Lifecycle

New arrivals (< 30 days) → PMax with 200% target ROAS (accept learning phase)
Bestsellers → PMax with 400% target ROAS (maximize)
Clearance → PMax with 150% target ROAS (clear stock)

Strategy 3 — Dynamic Feed for Promotions

Don't wait for your daily feed update for flash sales. Use the Content API to push real-time price changes:

• sale_price updated in real-time via API
• sale_price_effective_date with precise end time
• PMax detects the change in < 2h and adapts ads automatically

Strategy 4 — Local Inventory Ads If You Have Stores

If you have physical locations, enable Local Inventory Ads in GMC. PMax will automatically integrate them for users near your stores — conversion is 3x higher for "available nearby" than for pure e-commerce.

Strategy 5 — Test Video Assets

PMax campaigns with video have an 18% higher conversion rate than those without. In 2026, Merchant Center integrates Google Video Builder which automatically generates product videos from your images and feed data. No technical skills required.

---

Performance Max and AI FAQ

Can you still create Standard Shopping campaigns? No. Since early 2026, only Performance Max campaigns are available for new Shopping advertisers. Existing Standard Shopping campaigns continue to work but can no longer be created.

How do I prevent PMax from running on certain channels? You cannot exclude channels directly in PMax. The solution is to create dedicated campaigns on channels to exclude with higher priority (e.g., dedicated Display campaign to reclaim display budget outside PMax).

My ROAS has dropped since switching to PMax — is that normal? The transition period may lower ROAS for 4-8 weeks. If the drop persists after 2 months, check: feed quality, presence of audience signals, sufficient budget for conversions, and asset group segmentation.

Can AI spend my budget on low-margin products? Yes, if you don't segment with custom labels. The AI optimizes conversions, not your margins. Use custom labels to exclude or cap budgets on low-margin products.

What's the minimum budget for PMax to be effective? Google recommends a budget allowing at least 50 conversions/month. In standard e-commerce, that represents approximately $2,000-$5,000/month depending on average order value. Below that, the AI lacks data to learn.

Is Google Product Studio free? Yes, it's integrated for free in Merchant Center Next. It allows generating lifestyle images, backgrounds and image variants directly from your account.

---

Conclusion

Performance Max and Google's AI have fundamentally changed the rules of e-commerce. The power is no longer in manual keyword or bid management — it's in the quality and richness of the data you provide to the algorithm.

An incomplete GMC feed is now a double penalty: disapprovals + degraded PMax performance. Conversely, an optimized feed is a durable competitive advantage: better positions, reduced CPCs, SGE citations, and superior ROAS.

The good news: unlike the opaque algorithms of the past, PMax gives you clear levers — feed quality, custom label segmentation, first-party audience signals, and quality creative assets.

Start by auditing your feed quality with MyGoogle — the tool identifies missing attributes, inconsistencies, and priority optimizations in 30 seconds to properly fuel your PMax machine.