Asset Groups and Listing Groups in Performance Max 2026: Limits and Structure
Performance Max gets criticised for being a black box: targeting is handed to the algorithm, placements are only partly visible, search terms likewise. But one layer stays entirely under the advertiser’s control, and it decides a good half of the outcome — Performance Max asset groups and the listing groups inside them. That structure determines which creative can meet which product, which landing page receives traffic, and where a campaign starts bidding against itself.
Below: the exact limits for every asset type, the rules that make a listing group tree valid, the splitting patterns that work in retail, and the mistakes that send an entire budget into a single SKU.
Asset groups and listing groups in Performance Max: two separate layers
A single PMax campaign contains two independent grids, and they get conflated constantly:
- An asset group is a set of creative — headlines, descriptions, images, logos, video — plus a final URL and audience signals. It is how the campaign shows up.
- A listing group is a filter over Merchant Center products inside one asset group. It is what the campaign sells.
Listing groups exist only in campaigns connected to a product feed. A lead-gen PMax campaign without a feed has one layer only — asset groups. We covered that setup in Performance Max for lead generation — where intake quality matters as much as structure, see fake leads in Google Ads.
Asset group limits: required and optional
These are the numbers the interface enforces on save. Keep them in the creative brief. They did not change in 2026, but Google has revised them several times over the format’s life, so check the live help pages before a large build.
| Asset type | Limit | Min | Max |
|---|---|---|---|
| Headline | 30 characters | 3 | 15 |
| Long headline | 90 characters | 1 | 5 |
| Description | 90 characters | 2 | 5 |
| Business name | 25 characters | 1 | 1 |
| Landscape image 1.91:1 | 1200×628 recommended, 600×314 minimum | 1 | 20 |
| Square image 1:1 | 1200×1200 recommended, 300×300 minimum | 1 | 20 |
| Portrait image 4:5 | 960×1200 recommended, 480×600 minimum | — | 20 |
| Square logo 1:1 | 1200×1200 recommended, 128×128 minimum | 1 | 5 |
| Landscape logo 4:1 | 1200×300 recommended, 512×128 minimum | — | 20 |
| YouTube video | 16:9, 1:1 or 9:16, 10 seconds or longer | — | 15 |
Other constraints: 5,120 KB maximum file size, GIF, JPG and PNG accepted. A media bundle must stay under 150 KB, one per group. One call to action, automated by default.
A trap that bites teams building through the API or bulk imports: aspect ratio is validated when the asset is linked to a group, not when it is uploaded to the library. A 1200×620 image uploads happily and only fails at link time.
What brand guidelines change
With brand guidelines enabled, business name and logos stop being asset-group assets and move to campaign level. The 4:1 landscape logo can only live inside an asset group when brand guidelines are off. The practical consequence: decide on the brand profile before you build groups, or you will be relinking assets by hand. How the profile itself works is in Brand Guidelines in Google Ads 2026.
For retail campaigns connected to Merchant Center, Google will generate the minimum asset set from the feed. Convenient for launch, poor for performance: auto-generation gives the system exactly the minimum number of combinations. On keeping generated text inside your rules, see text guidelines for AI assets.
How many asset groups to build
The ceiling is 100 asset groups per campaign, minimum one. An asset group belongs to a single campaign and cannot be shared: if the same creative is needed in two campaigns, it has to be assembled twice.
The practical answer almost always lands between one and eight. There is exactly one splitting criterion: split when groups need different creative, different landing pages, or different audience signals. If the only difference is a product variant sharing the same images and the same page, keep it in one group.
| Situation | Call |
|---|---|
| Two ranges with distinct visuals and distinct landing pages | Separate asset groups |
| One product in three colours, one product page | One group |
| Best sellers vs clearance stock | Separate groups — different message, different economics |
| Seasonal collection with its own promotion | A dedicated group for the season |
| B2B and B2C from the same site | Separate groups, more often separate campaigns |
The opposite error is more common than over-consolidation: ten thin groups with five conversions each per month. The algorithm has nothing to learn from in any of them, and reporting fragments into statistically empty rows. When in doubt, start with fewer groups and split as data accumulates.
Listing groups: how the tree works
Listing groups are defined inside an asset group and decide which feed products that group may advertise. The structure is a tree:
- A subdivision node creates a new level.
- A unit node is a leaf — a final set of products, included or excluded.
- Every subdivision must be fully partitioned, which means it must contain an Other node. That guarantees each product falls into exactly one unit.
- A tree is valid once it has at least one unit. The root itself can be that unit, which is the “All products” case. Until the tree is valid, ads do not serve.
Available dimensions: brand, category, channel, condition, custom attribute, item ID and product type. “Custom attribute” in Google Ads is what Merchant Center calls a custom label; how to lay an assortment out across them is in custom labels in your product feed.
The key difference from Shopping
In standard Shopping campaigns, listing groups carried bids. In Performance Max they only include and exclude products — bidding belongs to Smart Bidding. Two consequences follow:
- Splitting the tree “to bid up on best sellers” achieves nothing. There is no bid to raise.
- The only way to give a product different economics is a separate asset group or campaign with its own tROAS target. Relative product priority is handled elsewhere — see Product Value Optimization in PMax and Shopping.
Google recommends grouping by dimensions — brand, product type, custom label — rather than enumerating individual item IDs. A hundred-node tree built on item IDs turns into manual labour on every assortment change and collapses the first time SKUs are renumbered.
Structures that work
Four patterns cover most retail cases:
| Pattern | Split by | Best fit |
|---|---|---|
| By margin | custom_label_0: high / mid / low margin | Broad assortment with uneven economics |
| By category | Top-level product_type | Stores with clearly distinct departments |
| By brand | brand | Multi-brand retail where the brand is part of the creative |
| By lifecycle | custom_label: new in / best sellers / clearance | Seasonal and fashion categories |
Whichever pattern you pick, one rule holds: one product, one asset group within a campaign. If the same SKU ends up in two asset groups, the campaign starts choosing between its own creatives on the same product, data for that product splits in half, and creative conclusions stop being trustworthy.
Overlap with standard Search campaigns is a separate mechanism — campaign type priority and match quality decide it — and we covered it in search themes in Performance Max.
Audience signals: the third splitting criterion
Audience signals are set at asset group level and are routinely mistaken for targeting. They are not targeting: the system treats them as a starting hint about where to look and moves beyond them as soon as it finds conversions elsewhere. The limits of that expansion are covered in optimised targeting and audience signals.
For structure, the implication is this: if two asset groups share the same signals and the same landing page, you have built a duplicate. So check all three parameters together when splitting — creative, URL, signals. If only the creative differs, consider whether adding assets to the existing group would do the job.
A workable signal set per group: CRM customer lists for the matching segment, site audiences such as category viewers and cart abandoners, and one or two in-market audiences. Mixing signals from unrelated segments inside one group hands the system a contradictory hint and lengthens the learning period.
Scaling the structure as the account grows
The most common launch question is where to start. A working sequence:
- Launch. One campaign, one or two asset groups, a listing tree at “All products” or split across two or three broad categories. The goal of this stage is data, not elegance.
- Weeks 2–4. Look at which products and themes consume spend. If 80% of the budget went into one product_type, you have a candidate for a dedicated group with its own creative.
- First split. Break out what already has statistics, not what you wish were selling. A group built for an aspirational segment that does not convert will simply sit without impressions.
- Different targets. Once segment economics diverge enough to need different tROAS values, asset groups are no longer the right tool — you need separate campaigns with their own budgets.
- Rebuilds. Every structural change costs accumulated learning. Between rebuilds, change creative, not the frame.
What to exclude at the listing group level
Exclusions are the underused half of the tree. Sensible candidates: out-of-stock items when the feed updates with a lag, products with negative margin after shipping, items priced below the point where a click can pay for itself, and SKUs sold only as part of a bundle. Each exclusion frees budget for products that actually return it — one of the few direct spend controls Performance Max leaves you.
Ad strength, video and auto-generation
Reaching “Excellent” ad strength requires your own videos in all three orientations — horizontal, square and vertical — each at least 10 seconds long. Without them, Google assembles video from your other assets. It works, but the output is not yours to control: typography, pacing and emphasis will be whatever the generator chose.
Treat ad strength soberly. It is a diversity indicator, not a revenue forecast. A “Good” group with a strong offer routinely beats an “Excellent” one padded to satisfy the meter. That said, launching on the bare minimum — three headlines, two descriptions, two images — leaves the system almost no combinations to work with.
Migrating a structure from Shopping to Performance Max
Teams moving off standard Shopping campaigns are tempted to recreate the old tree node for node. Don’t: in Shopping, granularity existed to carry bids, and PMax has no bids at that level. A sensible migration order:
- Put the old structure in a spreadsheet: node, share of spend, ROAS over the last 90 days.
- Collapse anything that differed only by bid. If two nodes sold the same product type at different bids, that is one group in PMax — express the priority difference through conversion value instead.
- Keep the split where creative or landing page differed. That is the only reason to split groups here.
- Port the exclusions first. Products you deliberately kept out are included by default in a new campaign. This is the single most forgotten step, and the usual source of “PMax dumped budget into products I never advertise”.
- Don’t run both in parallel for long. While legacy Shopping campaigns and a new PMax campaign share a feed, PMax takes priority and a clean comparison is not available anyway.
One number worth recording before the move: the share of spend going to brand queries. It almost always rises afterwards, and without a baseline you cannot distinguish genuine incremental growth from traffic that was already yours being re-labelled.
Reading asset reporting without fooling yourself
The most common analytical mistake in PMax is adding up conversions by asset. Every asset that appeared in a serving ad is credited with the full conversion, so the asset-level sum always exceeds the real conversion count. Those numbers are for comparing assets against each other, never for totalling.
What is worth checking regularly:
- Where spend actually goes across channels inside PMax — our breakdown of channel reporting in Performance Max.
- Which landing pages receive traffic when final URL expansion is on: final URL expansion in PMax and AI Max.
- Whether a new creative set is genuinely better, using asset experiments in Performance Max instead of before-and-after eyeballing.
- Whole-campaign incrementality, once volume allows it: Conversion Lift for Search and Performance Max has been self-serve since October 2026.
Pre-launch checklist
- Brand guidelines decision made before groups are built, not after.
- Every asset group exceeds the minimum: 8–12 headlines, 3–4 descriptions, 3–5 images per required format.
- Own video in at least two orientations.
- The listing group tree is valid — every subdivision has an “Other” node.
- No SKU appears in two asset groups of the same campaign.
- Splits built on feed dimensions, not enumerated item IDs.
- Each group has its own final URL and its own audience signals, otherwise the split has no meaning.
- URL expansion exclusions set for pages that should not receive ad traffic.
Structure is the part of Performance Max where a human still decides. Get it right once and ongoing optimisation becomes creative and feed work rather than a fight with your own campaign. For the rest of the operating stack — stable ad accounts and infrastructure that holds at volume — see Google Ads agency accounts and the full PPC Rebels services list.
FAQ
How many asset groups can one Performance Max campaign have?
Up to 100, with a minimum of one. The ceiling is almost never relevant: in practice the working range is one to eight, depending on how many genuinely different creative themes you have.
Can an asset group be reused in another campaign?
No. An asset group belongs to one campaign. The assets themselves live in a shared library, but the group has to be rebuilt.
What happens if a listing group tree has no “Other” node?
The tree is invalid and ads do not serve. Full partitioning is a structural requirement, not a recommendation.
Do listing groups set bids like they did in Shopping?
No. In Performance Max they only include and exclude products. Bidding is Smart Bidding’s job, and differing product economics are expressed through separate campaigns or conversion value.
Can the same SKU sit in two asset groups?
The tree structure prevents it within a single partition, but careless splitting across different dimensions can create overlap. Avoid it: product data fragments and creative conclusions stop being reliable.
How many images should launch with?
The minimum is one landscape and one square, but that is a floor, not a plan. A workable launch is 3–5 images per required format plus 4:5 portraits, so the system has combinations to explore.
Is my own video mandatory?
No, but without it Google generates video from your assets and you do not control the result. “Excellent” ad strength requires your own videos in three orientations, each 10 seconds or longer.
Why do asset-level conversions add up to more than campaign conversions?
Because every contributing asset is credited with the full conversion. The data is comparative, not additive.
What do I do when budget concentrates in one product?
That is Smart Bidding working as designed — it goes where conversions are. If you need distribution, separate product groups into their own campaigns with their own budgets, or differentiate conversion value.
Why does the interface reject an image that uploaded fine?
Aspect ratio is checked at link time, not upload time. An image with the wrong proportions enters the library and fails only when you attach it to a group.
Should every product category get its own asset group?
Only if the categories need different creative, landing pages or audience signals. Splitting for its own sake produces thin groups with unreadable statistics.
How often should the structure be revisited?
Quarterly, and whenever the assortment or seasonality shifts materially. Between reviews, change creative inside groups rather than the structure itself — restructuring throws away accumulated learning.
Performance Max automates buying, not architecture. Asset groups decide what the campaign says and listing groups decide what it sells; everything else the algorithm will fill in for you, including your mistakes.
