Best Performance Max Asset Types, Ranked

Vertical video is the PMax asset most teams still skip, and the one that opens YouTube Shorts and other tall placements without a letterbox. Ten seconds is the usual minimum; 10–20s with a first-frame product action beats a 45-second brand film. Ranked first because this inventory behaves like paid social — if the first three seconds are a presenter in a studio, you paid for a skip. UGC-shaped verticals, including AI avatar spots built as ads, belong here more than they belong in a 16:9 pre-roll.

Spec the file as a complete ad, not as a crop. Captions on, product in the first beat, no assumption of sound, no tiny legal text. Google will run it next to organic Shorts; a logo sting is a swipe. If you generate UGC for Meta anyway, export a clean 9:16 without platform-specific UI and upload it as a PMax video rather than hoping auto-generate will invent one from stills. Keep a unique first frame from your Meta winner so you are not fatiguing the same open in two auctions without noticing. If you only have landscape, do not upload a pillarboxed file and call it vertical. Rebuild the first frame in 9:16 or leave the slot empty until you can. A fake vertical trains PMax that your brand looks like a mistake.

Limits: PMax will still mix this with Search text; a vertical that only makes sense after a 20-second story will fail in isolation. Put the brand and the offer in the first five seconds. Upload more than one vertical if you have them — different hooks, not three encodings of the same take. And do not let the asset group 'Excellent' score replace a watch of the first frame on a phone. Review actual YouTube placements in the report, not only ad strength. If Shorts-like inventory is zero, you still have a vertical problem even if the UI is green. Fix the file, not the bid.

If you upload no video, Performance Max will often auto-generate one from images and text. That clip fills the requirement and usually looks like a requirement. A real landscape video — 10 seconds or longer, offer in the first five, brand visible early — is the asset that makes in-stream inventory usable. Ranked second only because so much PMax ecommerce growth now sits in tall mobile surfaces; you still want both orientations.

Structure for skippable: the first five seconds must contain the product and the reason to stay. 15 seconds is a sweet length for a lot of in-stream and Discovery use; 60-second explainers belong in a different campaign if they belong at all. Square (1:1) video is worth adding if you have it, but do not let a square crop be your only file. Horizontal remains the in-stream native. Brand in the first five seconds does not mean a logo sting; it means the product is recognisable. A three-second mood shot before the SKU is a skip. Cut it. If legal needs a disclaimer, caption it — do not spend the open on a fade.

Production shortcut: take the winning UGC script, recut with product B-roll that fills 16:9, captions safe for a wider frame. Do not upload a TV commercial from 2019 with a new end card. PMax will serve pieces of it. Asset groups with only auto-video plus a feed are how 'PMax doesn't work' case studies get written. Keep auto-generated video off as soon as a real clip is approved. Leaving both in the group lets PMax spend on the slideshow because it is 'available'. Remove the junk asset the same day you upload the real one.

Image assets are still how Display, Gmail and a lot of Discovery actually look. Ranked third because a vertical video will not save you if the only stills are marketplace PNGs. Supply landscape (about 1.91:1), square (1:1) and portrait (4:5) of the product in use. Lifestyle that still shows the SKU clearly beats a beautiful room where nobody can find the item. These stills also become the frames of auto-video if you fail rank 1–2 — so they have to be honest.

Google's own checklists still want multiple images per ratio, not one hero cropped three ways. Vary the scene: in-hand, on-body, before/after if you can claim it, pack shot as a minority. Text on image should be sparse; PMax will add headlines. If AI product-photo tools are in the mix, keep the SKU accurate — warped logos and extra buttons are a policy and a returns problem. Portrait 4:5 is optional on paper and useful in practice on mobile Discovery-like surfaces. Shoot or generate on a phone first; PMax stills are seen small. If the SKU needs a crop-in to be recognised, you do not have a lifestyle image yet — you have a room. Add a tighter in-hand frame to the same ratio set.

Retail PMax with a feed already has product images. Do not treat that as a full stills strategy. Feed shots are pack-forward; they underperform as the only Display creative in many accounts we see. Add lifestyle as campaign assets even when the listing is complete. Reject any AI still where the label type, colourway or count of components does not match the PDP. PMax will combine that image with a true headline and you will buy returns that look like a bidding problem. Feed images can stay pack-forward; campaign stills should not. If a stranger cannot name the SKU at phone size, replace the file before you add another ratio.

Short headlines are the Search half of PMax and they get glued onto your images everywhere else. Ranked fourth because they are cheap and they are where ecommerce brands still paste the same benefit three times. Fill toward the top of the allowance (up to 15, 30 characters each) with distinct angles: problem, proof, offer, identity, logistics. Pin only for legal. Pinning for taste is how you kneecap the combiner.

Write them to stand alone. 'Ships tomorrow' is a headline. 'Discover our story' is a waste of 18 characters. Include at least one short line (Google still likes a subset under ~15 characters for tight slots). Do not stuff the exact match keyword into every line; the asset group already has listing groups or search themes to hint intent. Refresh when a query cluster is converting and the headlines never mention the reason. Read the asset combinations or listing-group reports for headlines that never serve. Replace those lines; do not add a sixteenth synonym. Dead headlines are how you think you have variety when you have a pile.

AI copy is fine as a first pass if you paste the claims fence. It is not fine as fifteen paraphrases of 'premium quality'. Treat headlines like a hook matrix: one idea per line. The combiner will mix them with descriptions you did not storyboard, so never write a headline that only makes sense as line two of a jingle. Align one or two headlines to the same angle as your vertical video so Search and YouTube do not describe two different products. PMax will still mix freely; you can at least avoid a contradiction in the sources.

Long headlines (up to 90 characters) and descriptions (two to five, also ~90) are the sentences Display and larger units actually read. Ranked fifth because they rarely unlock new inventory the way video does, but a missing long headline is a self-inflicted 'Average' ad-strength tax, and a description that repeats the short headline is a wasted slot. Write a complete value statement plus a proof plus a next step.

One long headline should be sayable as a full answer: problem, product, proof. Descriptions can carry returns policy, shipping, and a qualifier you cannot fit in 30 characters. Keep a shorter description in the set for tight units. Do not put the only honest disclaimer exclusively here if a short headline might serve without it — compliance belongs in the places that will actually show, which sometimes means a pinned short line plus this block. Put the return window and the shipping promise in different descriptions so a large unit can show either. Do not stack every qualifier into one 90-character block that none of the combiners can fully use.

These fields are where ChatGPT-style copy goes generic. Paste three customer sentences and force one of them into a description. If every description could sit on a competitor's ad, you do not have descriptions yet. Refresh descriptions when the offer changes; stale '20% off' in a PMax group is a policy and a trust problem. Assign an owner. Headlines get love; these fields rot. A quarterly pass is enough if someone owns it; never is the default. If the offer is gone, the description is a policy issue, not a copy issue.

Signals (Customer Match, custom segments, your data) and search themes are not creative, but they are assets in the operational sense: they tell PMax where to start looking. Ranked sixth because a beautiful video into a blank signal still eventually finds buyers — slower — while a perfect signal into marketplace PNGs still looks like a banner. Use them as a starting hint, then judge the asset group on search term and placement reports, not on the signal you wished was targeting.

Upload customer lists you actually have rights to use. Build custom segments from URLs and terms that map to the asset group's angle, not to the whole brand. Search themes in 2026 can be numerous per group; treat them like exact-ish hints, not a dumping ground. If a group is meant to sell a starter kit, do not theme it with every SKU in the catalog. Re-upload lists on a schedule so Customer Match does not silently decay. An old purchaser list as your only signal is how PMax hunts yesterday. Pair it with a custom segment that matches this asset group's angle.

The failure mode is using signals as a substitute for a Search campaign you were afraid to pause, then blaming PMax when branded queries soak spend. Watch the search term view. If you need control, keep a Search campaign for the queries you must own and let PMax hunt the rest with assets that can survive Display. If branded queries dominate, use negatives where the product allows, or keep brand in Search. Do not 'fix' it by stuffing more themes. Themes are hints; they will not out-shout your own brand name if that is the cheapest conversion.

You need a square logo; a landscape logo helps some banner shapes. You do not need to romanticise them. Auto-generated video from stills is the asset type we rank last because it exists to prevent an empty slot, not to win YouTube. If this is your only 'video', you have not done video. Upload the logo, then go back to rank 1. Ranked last because they are required hygiene, not a growth story. Upload a sharp square logo, add the landscape mark if you have it, and treat auto-video as a spare tyre. If the spare is still on the car after week two, you do not have a Performance Max creative programme — you have a checklist.

Logo rules are boring and worth following: high-res, clear on light and dark, not a full lockup so tiny it becomes a smudge. Landscape 4:1 if you have a horizontal wordmark. That is the end of the logo playbook. For auto-video, watch it once on a phone. If you would not pay for that impression, upload a real clip. Google's ad-strength checklist will still smile at you. The auction will not. Watch the auto-video once on a laptop and once on a phone. If either looks like a template, remove it after the real clip is in. Ad strength may dip overnight; that is acceptable. Paying for a slideshow is not.

Business name should be the brand people would search, not a slogan. CTA can stay automated unless you have a hard legal preference. None of these are where an ecommerce team should spend the design week. Spend it on a 12-second vertical of the product working. Set a reminder to replace auto assets whenever Google regenerates them after an edit. Groups have a habit of re-growing junk. Hygiene is recurring, which is why this type still cannot be the strategy. Budget the 12-second vertical instead; the logo was never going to find the customer.