Performance Max is Google's most demanding campaign type, and it's the most punishing for advertisers who underinvest in creative. PMax runs across every Google channel — Search, Display, YouTube, Discover, Gmail, Maps — and its automated bidding logic relies on asset quality to find the highest-performing combinations. Video is one of the most influential asset types in the mix. Advertisers who upload strong, varied video assets consistently see better conversion rates than those who rely on static images alone or upload a single video and never refresh it.
The problem: PMax's demand for video variety — multiple aspect ratios, multiple lengths, multiple creative angles — has traditionally made asset production expensive and slow. AI video generation changes that equation entirely.
What Google Performance Max Actually Needs From Video#
PMax assembles ad creatives dynamically from the asset pool you provide. The system mixes and matches headlines, descriptions, images, logos, and video to find the highest-performing combination for each placement and audience. The more assets you provide — and the stronger their individual performance signals — the more effectively PMax's automated optimization works.
For video, Google's official guidance recommends:
- At least 3–5 video assets per campaign, covering different lengths and formats
- Horizontal (16:9) for YouTube in-stream and Display placements
- Vertical (9:16) for YouTube Shorts, Discover mobile, and Gmail on mobile
- Square (1:1) for Display and Gmail where both orientations would crop poorly
- Lengths: 6-second bumpers, 15-second mid-funnel videos, and 30-second consideration videos each serve different placement types
If you upload only one video in one aspect ratio, PMax auto-generates cropped versions for other placements — and those auto-crops rarely preserve the original's visual focus. Subject positioning shifts, text overlays get cut, and the frame composition that worked for 16:9 looks awkward at 9:16. Providing native assets in each required ratio significantly outperforms letting Google handle the cropping.
The practical implication: a properly supplied PMax video asset set requires 9–15 individual files at campaign launch — roughly 3 message angles × 3 aspect ratios × at least one length per placement type. Producing that at traditional video production costs means multiple shoots and edit rounds. With AI video generation, it's one structured production session.
The AI Video Production Workflow for PMax#
The most effective PMax video production workflow treats the campaign's asset requirements as a production matrix rather than a single creative brief. You're not making one video — you're systematically populating a grid.
Step 1: Define 3–4 Creative Angles#
Each angle should address a different stage or objection in the buyer's journey. For a B2B SaaS product, that might be:
- Angle A — Social proof: "10,000 teams use this to cut reporting time in half"
- Angle B — Problem statement: "Manual reporting takes 6 hours a week you don't have"
- Angle C — Feature demonstration: "See how automated dashboards work in 90 seconds"
- Angle D — Offer / urgency: "Start your free trial — full access for 14 days"
For an ecommerce brand, the angles shift: lifestyle outcome, product detail, review-led social proof, and time-limited offer. The specific angles depend on your product, but the principle holds: launching with variety gives PMax's automated system something to optimize across.
Step 2: Write Prompts Across the Three Core Lengths#
The prompt structure for each length:
6-second bumper: Single claim, high contrast, brand end card. Example for a reporting tool: "6-second product ad, bold white text on deep navy background, single claim 'Cut reporting time by 80%', logo end card with website URL, clean minimal animation, no voice narration, 16:9."
15-second conversion video: Problem-solution structure, one specific benefit, explicit CTA. Example: "15-second ad, opens on a professional staring at a cluttered spreadsheet, cuts to the same person with a clean automated dashboard and an expression of relief, voice narration 'Stop building reports manually', call to action text overlay 'Try free for 14 days', 16:9 format."
30-second consideration video: Longer story arc, feature details, trust signals. Narration over product demonstration, two or three specific benefits, end card with CTA. This format earns the time by going deeper than a bumper can.
Step 3: Generate Across All Three Aspect Ratios#
For each length × angle combination, generate horizontal (16:9), vertical (9:16), and square (1:1) versions. Vertical and square prompts require minor adjustments to keep visual focus centered — text overlays and subject positioning need to account for the narrower frame. A prompt that specifies "subject and text centered in vertical frame, 9:16" consistently produces cleaner crops than relying on automated reformatting.
A systematic run produces 36 assets from 4 angles × 3 lengths × 3 ratios — far more than PMax requires at launch, but the production time is hours rather than weeks, and you now have a testing pool to draw from as performance data comes in.
The Creative Formats That Perform Best in PMax Placements#
Not all video types convert equally across PMax's placements. Three formats drive consistent results:
Direct-response product videos (Display and Gmail): Short, text-heavy, benefit-first. These don't rely on narrative — they deliver a claim, show the product or outcome, and direct to a CTA. For Display placements especially, the first two seconds need to communicate the core offer before the viewer can scroll past. Static-feeling AI video — a clean product over a solid background with text overlay — often outperforms cinematic treatments in these placements because it reads clearly at small sizes.
Testimonial-style social proof (YouTube in-stream): The YouTube in-stream placement gives you five seconds before the skip button appears. AI-generated videos that open with a specific result — "I booked 42 calls last month using this tool" or "We reduced returns by 34% after switching" — hold significantly more attention past the five-second mark than videos that open with a logo animation or slow brand intro. Creating Google video ads with AI covers the specific hook structures that earn attention in pre-roll formats.
Problem-agitation videos (YouTube Shorts, Discover): Vertical placements respond to content that feels native to the feed — conversational pacing, direct address, problem-first framing. AI video for these placements works best when it mimics the organic content style of the platform. A video that looks like a polished commercial performs worse in a Shorts feed than one that opens mid-thought and gets to the point in the first three seconds.
Structuring the Asset Mix for Automated Optimization#
Google's PMax system optimizes toward the best-performing combinations, but it needs signal to do that — which makes the first four to six weeks of a campaign a signal-gathering period as much as a performance period.
Launch with variety, not volume on a single angle. PMax performs better over time when the initial asset pool spans meaningfully different creative directions. Uploading 15 versions of the same message gives the system very little to work with compared to five versions each of three distinct angles.
Let low-performing assets be deprioritized naturally. PMax's asset reporting shows "Learning," "Good," "Best," and "Low" ratings for each asset. Remove Low-rated assets only after 30+ days of meaningful impressions. Removing them earlier disrupts optimization before the system has found the right audience segments for each creative direction.
Add fresh assets on a 2–4 week cycle. The system rewards new creative. Adding 2–3 new video assets every few weeks — particularly angles informed by performance data, extending what's already working — consistently lifts performance versus campaigns where assets are set once and left unchanged.
Matching Video Length to Conversion Stage#
The length of a PMax video asset signals the placement system about where in the funnel it belongs — and each length requires a different creative execution to work.
6-second bumper ads are interruption, not persuasion. The only job of a 6-second ad is to plant a brand memory or reinforce a claim for someone who has already encountered longer-form content. Write them as a single complete thought: "The only tool that [specific claim]. [Brand name]." They're most effective as companion assets to 15-second and 30-second videos rather than standalone persuasion.
15-second conversion videos are the workhorse format. They're long enough to deliver a problem-solution structure and a CTA, short enough to hold attention across placements. For high-intent audiences in PMax, 15-second assets typically produce the best direct conversion rates. The structure: problem in 3 seconds, solution in 8 seconds, CTA in 4 seconds.
30-second consideration videos are for mid-funnel audiences who've seen the brand or category but haven't committed. They can carry more context, feature demonstrations, and trust signals like customer counts or partner logos. They perform better in YouTube in-stream than in Discover or Display, where shorter formats hold attention more reliably.
Testing AI video ad creative goes deeper on setting up an asset testing methodology that extracts accurate performance signal from PMax's limited reporting.
Reading PMax Asset Performance Reports#
Google provides asset-level performance ratings in PMax that, while less granular than standalone campaign reporting, give enough signal to make informed creative decisions.
Asset rating (Best / Good / Learning / Low): A relative quality signal comparing assets within your pool. "Low" means the system is deprioritizing the asset against others you've uploaded — it doesn't necessarily mean the creative is bad. It may mean similar assets are outperforming it, or the audience segment it's well-suited for isn't receiving enough budget to generate meaningful impressions.
Impressions by asset type: Shows which placements your campaign is actually winning. If video impressions are low relative to image impressions, it's usually a quality signal issue — the system is finding it harder to place the video assets profitably. Adding vertically-oriented assets often resolves this for campaigns that launched with only 16:9 video.
Conversion rate by asset group: PMax asset groups let you organize assets by audience or creative theme. Comparing conversion rates across asset groups reveals which combination of creative angle and audience targeting is driving the most efficient conversions — the data you need to decide which angle's assets to expand in the next production cycle.
What PMax reports don't give you: individual creative performance at a placement-by-placement level. For that granularity, running dedicated YouTube campaigns in parallel using the same video assets — where placement and format performance are fully visible — provides the breakdown data to inform future PMax creative decisions.
Scaling the Creative Refresh With AI Video#
The practical ceiling on most PMax campaigns isn't budget or bid strategy — it's creative refresh rate. Campaigns where assets go unchanged for 60+ days consistently underperform campaigns with regular creative additions, because the automated system has fewer combinations to test across evolving audience segments.
Scaling AI video ad creative production becomes the lever that unlocks PMax's full potential over time. A systematic workflow — one 90-minute session every two weeks producing 10–15 new assets — keeps the creative pool fresh without ongoing agency involvement:
- Pull the asset performance report and identify which angles are rated "Best" or "Good"
- Write 3–4 new prompt variations extending the high-performing angle — a new hook, a different visual treatment, or a variation on the offer
- Generate new assets in the required aspect ratios and lengths
- Upload to PMax and tag with the production date to track refresh timing
- After 30 days, remove assets that received "Low" ratings for the full review period
Over six months, a campaign maintained this way builds to a creative pool of 60–100 video assets — enough for PMax's automated system to genuinely optimize across audiences, placements, and funnel stages simultaneously. The advertisers who consistently outperform in PMax aren't the ones who built the best single video. They're the ones who built the most capable and regularly refreshed asset pool.
PMax's automated optimization is only as good as the creative it has to work with. Treating video asset production as a one-time task is the most common reason PMax campaigns plateau.
If you need a tool built for the multi-format, multi-ratio, batch production that Performance Max requires, Mango generates the full asset matrix in a single session — so your PMax creative stays fresh and your campaigns keep improving.
