How Performance Max Consumes Your Feed (And Why Feed Quality Matters More Than Ever)

PMax is not just Shopping with extra steps

Performance Max campaigns replaced Smart Shopping as Google’s AI-first campaign type. But many advertisers still treat PMax as a renamed Smart Shopping — submit the same feed, set a tROAS, and hope for the best. That approach misses most of what PMax actually does with your product data.

Standard Shopping shows a product image, title, price, and store name. PMax builds full ads: display banners, YouTube video thumbnails, Discover cards, Gmail promotions, and more — all assembled automatically from your feed data and the assets you provide. The quality ceiling for PMax ads is directly bounded by what’s in your feed.

How PMax reads your feed: the asset assembly model

When Google’s AI constructs a PMax ad unit, it draws from two sources: your uploaded asset group (manual headlines, descriptions, images, videos) and your product feed. Feed data supplements and sometimes overrides what you’ve manually uploaded.

Feed attribute	How PMax uses it	Missing = ?
title	Used as headline source. Google picks phrases from title to build responsive ad headlines when asset group headlines run out of combinations.	Generic titles lead to generic AI-generated headlines.
description	Seeded into ad description slots. Phrases matching the shopper’s query are prioritized.	AI falls back to generic category descriptions. Lower relevance.
lifestyle_image_link	Used in Display, YouTube, and Discover placements. Google prefers lifestyle over product-on-white for these surfaces.	PMax falls back to asset group images or auto-cropped product image, often poorly composed.
product_highlight	Bullet-point highlights shown in free listings and rich Shopping panels. ~95% of merchants leave this empty.	Your product panel looks sparse vs. competitors who fill it in.
product_detail	Structured key-value specs surfaced in product detail views and rich product panels.	Less rich product panel. Competitors with specs filled win the comparison click.
additional_image_link	Multi-image carousel in Shopping placement. Also pulled into Display creative pool.	Single image display. Lower engagement in carousels.
custom_label_0-4	Used for asset group product filters. Controls which products enter which asset group.	All products land in the default “All products” asset group with no differentiation.

The lifestyle_image_link gap most advertisers miss

Standard Shopping ads use your image_link: product on a clean background, filling the frame. That format is correct for the Shopping tab. But PMax serves across Display, YouTube, and Discover — and on those surfaces, a product-on-white image looks out of place and underperforms.

The lifestyle_image_link attribute (minimum 600×600px, aspect ratio between 2:0 and 2:3) lets you provide a separate image showing the product in context: a sofa in a room, a jacket on a model, a coffee maker on a kitchen counter. Google prioritizes this image for non-Shopping placements in PMax.

If you don’t provide lifestyle_image_link, Google either uses your asset group images for these placements or auto-crops your product image — frequently producing awkward compositions. The resulting ads look amateurish compared to competitors who’ve provided proper lifestyle imagery.

What to do: For your top 20% of SKUs by revenue, commission or curate lifestyle images and add them as lifestyle_image_link. Use stable URLs — avoid timestamp-based CDN URLs that change on re-upload.

Asset groups and feed segmentation

A PMax campaign can contain multiple asset groups, each with its own set of creative assets and its own product filter. The product filter is built using: google_product_category, product_type, brand, item_group_id, id, or custom_label_0-4.

This architecture means your feed segmentation strategy directly determines how precisely you can tailor ad creative to product groups. An account with no custom labels and generic product types will have one or two asset groups covering the entire catalog. An account with clean custom labels and deep product types can have:

• An asset group for “Hero products” (custom_label_0 = hero) with tailored headlines about bestsellers

• An asset group for “New arrivals” (custom_label_1 = new_arrival) with freshness messaging

• An asset group for “Clearance” (custom_label_2 = clearance) with sale-oriented copy and creative

• Category-specific asset groups (product_type = Apparel > Women > Dresses) with relevant images

Without this segmentation, all products share the same creative — the AI shows the same ad messaging for a clearance handbag and a full-price luxury coat. The creative mismatch reduces conversion rates and gives the algorithm conflicting signals about product value.

How the title becomes a headline

When Google’s responsive ad generation exhausts your manually uploaded asset group headlines, it mines your product titles for additional headline candidates. A title like “Men’s Leather Belt — Brown — Size M” becomes potential headline material: “Men’s Leather Belt”, “Brown Leather Belt”, “Belt — Size M”.

But a title like “Belt_SKU_4821” or “Leather Belt - Our Price” generates headlines that sound bizarre in ad context. The AI doesn’t distinguish between a product title and an ad headline — it treats your title as a content source.

Practical implication: Optimize titles for both Shopping query matching and the sub-phrases they contain. A title optimized for both: “Trafalgar Men’s Genuine Leather Dress Belt — Brown — 1.25-Inch Width” contains natural-sounding headline fragments and keyword-rich phrases simultaneously.

Feed quality directly caps PMax performance

In standard Shopping, a weak feed underperforms but the damage is contained. In PMax, a weak feed compounds across all placements. The AI has fewer high-quality signals to work with, produces lower-quality ad creative across all surfaces, and generates less relevant audience signals to bid against.

The most impactful feed attributes for PMax, beyond the standard Shopping basics:

• lifestyle_image_link — enables non-Shopping placement creative without relying on asset group images

• product_highlight — free bullet points in product panels; visible across 6 surfaces at zero cost

• product_detail — structured specs; critical for electronics, appliances, and technical products

• custom_label_0-4 — enables meaningful asset group segmentation

• additional_image_link — feeds the multi-image carousel and Display creative pool

Common mistakes in PMax feed configuration

• One giant “All products” asset group: No custom labels = no segmentation = same creative for every product. Implement at least a margin-tier and performance-tier label scheme before launching PMax.

• Using product-on-white images for lifestyle placements: Works in Shopping, looks awkward in Display and YouTube. Provide lifestyle_image_link for top SKUs.

• Ignoring product_highlight and product_detail: ~95% of merchants leave product_highlight empty. Fill 4-6 bullets per product.

• Weak titles that generate poor AI headlines: Review titles as headline source material, not just Shopping query bait.

• No feed segmentation for bidding: Without custom labels, PMax can’t differentiate high-margin products from clearance items.