Google App campaigns do not ask an advertiser to design one fixed ad. They ask for ingredients: text, images, videos, store-listing material, and—where eligible—HTML5. Google can combine, crop, modify, or omit those ingredients as it assembles ads across Search, Google Play, YouTube, Discover, the Display Network, and other app inventory. The creative job is therefore not to upload a folder of files. It is to build a portfolio that can express distinct product arguments in formats the system can actually use.
The short answer: useful Google App campaign creative assets cover four decisions. Concept coverage means distinct, truthful reasons to install and use the app. Format coverage means those ideas work as independent text, eligible image ratios, and landscape, portrait, and square video. Store-asset continuity means the listing material Google can assemble into ads and the page a visitor reaches substantiate the same promise. Reporting limits mean ratings are relative, current 14-day signals—not isolated creative experiments—and asset totals are not campaign spend. Produce the missing concept or format, repair a broken store handoff, or gradually replace a Low asset after Learning; do not treat a folder count or Best label as proof of valuable acquisition (asset types, asset reporting, creative guidance).
This guide is for app founders, heads of growth, UA leads, and creative strategists who already know the campaign goal and need to decide what to produce, what to replace, and what the reporting can honestly support. If the unresolved question is whether to optimize for installs or a post-install event, start with Google Ads for apps: choosing the campaign goal.
What are Google App campaign creative assets?
Google App campaign creative assets are the text, image, video, HTML5, and app-store materials Google can use to assemble ads for an app. Google documents that App campaigns can mix supplied assets with content retrieved from Google Play, the App Store, or eligible website pages. Supplied content may be cropped, dropped, or modified during serving, and Google may generate a video when the advertiser does not provide one (Google Ads: assets and ads in App campaigns).
That operating model changes the creative brief. A storyboard that works only when one headline, one video, and one end card appear in a fixed order is fragile. Each text line needs to make sense independently. A visual must communicate after a crop. A video needs to establish the product without assuming a preceding asset supplied the context.
The advertiser still owns the commercial logic:
which user situation the campaign should recognize;
which product promise is supportable;
which proof makes that promise credible;
which expectation the store page and onboarding can fulfill;
which in-app event makes the acquired user valuable;
which creative question should receive the next production cycle.
Automation can assemble and distribute assets. It cannot decide whether those assets represent distinct, truthful customer hypotheses.
Start with two maps: concepts and formats
Creative libraries become difficult to diagnose when teams label everything by file type. “Three videos and five statics” describes production output, not what the market was asked to believe.
Build two maps before making assets.
The concept map
A concept is a complete commercial argument. Record:
User situation: the moment, problem, or desired progress the person recognizes.
Promise: the useful outcome the app can truthfully help create.
Mechanism: what the app does that makes the promise plausible.
Proof: product demonstration, interface evidence, approved customer evidence, or another supportable reason to believe.
Objection: the concern a qualified prospect may have.
Expected post-install behavior: the activation or value event that should become more likely if the concept attracts the right user.
Disconfirmation: the evidence that would make the team revise or retire the idea.
“Fast cuts,” “UGC,” and “vertical video” are executions, not concepts. “Plan a week of meals from ingredients already at home” could be a concept if the app actually supports it and the creative demonstrates how. The same concept may appear as a product demo, a creator narration, a sequence of interface screens, or a static problem-and-mechanism image.
The format map
The format map shows whether each viable concept can participate across eligible surfaces. For the standard App campaign asset specification checked on October 13, 2026, Google lists the following production boundaries (Google Ads: App campaign specifications):
Asset Current specification Production decision --- --- --- Headlines 1–5; maximum 30 characters each Make each line independently understandable Descriptions 1–5; maximum 90 characters each Cover a distinct benefit, mechanism, proof, or next action Landscape images 1.91:1; 1200×628 recommended Compose for a wide placement rather than stretching a square Portrait images 4:5; 1200×1500 recommended Keep the app action legible in a vertical composition Square images 1:1; 1200×1200 recommended Preserve a self-contained product argument Image quantity Up to 20 images Use slots for meaningful coverage, not near-duplicate output Videos 16:9, 9:16, and 1:1; 10–60 seconds; up to 20 Translate the concept across orientations with a clear opening Video hosting YouTube-hosted; Google may auto-build video if none is uploaded Supply deliberate video when control over the demonstrated experience matters
These are documented boundaries and recommendations, not a required quota or a promise of delivery. Campaign subtype, eligibility, and controls can differ; check the current specification before commissioning files. The distinction matters especially when a team mistakes technically accepted uploads for a coherent production plan.
Treat specifications as eligibility, not strategy. A full set of ratios does not create conceptual diversity. Ten near-identical files can satisfy dimensions while teaching the team nothing new.
Use this matrix to find gaps:
Concept Standalone text Square image Vertical image Horizontal video Vertical video Store-page continuity Post-install event --- --- --- --- --- --- --- --- Immediate utility Clear promise Product moment Problem-to-action sequence Demo with context Fast product proof First screenshots repeat the promise Core action completed Credibility Supportable proof Evidence-led frame Proof plus objection Mechanism walkthrough Creator demonstration Listing substantiates the claim Qualified activation Differentiation Specific contrast Side-by-side use case Alternative workflow Before/after process Product-in-use comparison Store copy explains difference Repeat use or paid action
The labels above are examples, not a prescription. An app may need one concept deeply translated across formats before it needs a third concept. The matrix is useful because it makes the choice explicit.
Build assets that can survive automated assembly
Google states that text assets may be combined and that each line should work independently. It also states that supplied content may be dropped, cropped, or modified during serving (Google Ads: assets and ads in App campaigns). Those constraints suggest four production rules.
1. Make each text asset a complete thought
Do not write a headline that depends on the description to finish its grammar or meaning. Avoid five variations of the app name. Use the limited text inventory to cover different parts of the argument: situation, benefit, mechanism, proof, and next action.
Check every possible line against three questions:
Is it true without qualification supplied elsewhere?
Does it still make sense beside any other approved line?
Will a person understand the app’s job without internal product vocabulary?
A line can be distinct without being sensational. Specificity usually creates more useful variation than superlatives.
2. Design for a crop, not only a canvas
Keep the product action and essential visual proof away from vulnerable edges. Review square, vertical, and horizontal versions as separate compositions rather than mechanically extending a master file. If the concept requires small interface text to be readable, simplify the demonstration instead of assuming every placement will preserve the detail.
Google’s documentation explicitly warns that content may be cropped or modified. That is not a reason to avoid visual nuance; it is a reason to place the core product truth in a resilient area and treat secondary decoration as expendable.
3. Show the app experience early
For most app categories, a creative portfolio needs evidence that the product exists and performs the advertised job. A video that spends its entire opening on a lifestyle problem may earn attention while delaying the mechanism. A static that resembles a generic category advertisement may fail to establish what happens after installation.
Google’s current creative guidance recommends full-frame, high-density images and video with minimal text overlay, showing the app experience quickly, and captioning video because some inventory defaults to sound-off (Google Ads: App campaign creative guidance). Use captions to preserve meaning, not to turn the screen into a dense slide. Check whether the product action is understandable with audio muted and at placement size.
This guidance is directional, subject to change, and not a performance promise. The right opening depends on the category, user knowledge, and concept. The diagnostic question is whether the creative gives the viewer enough product evidence to form an accurate installation expectation.
4. Treat the store listing as part of the supplied system
Google can use app-store content when assembling ads, and a paid path may also send a person to the store listing. The store icon, screenshots, video, description, rating context, and localization can therefore participate in acquisition even when the paid team did not upload them as campaign files.
Audit the handoff from ad to store. If a vertical video promises a specific workflow but the first screenshots lead with an unrelated feature, the portfolio creates two stories. The separate guide to App Store screenshots and paid UA explains how to test that handoff without claiming screenshots directly change auction prices. If the decision is whether to route an audience to a tailored Play page or test a treatment, use the Google Play custom store listings vs. experiments guide.
Store assets are observable acquisition inputs
Google distinguishes non-store assets, such as uploaded text, images, and video, from store assets. Store assets can include app video, cover photo, screenshot, short description, title, icon, and developer name. Google says these assets are visible in the Google Ads frontend and API and in reporting; they should therefore be inspected as acquisition inputs, not treated only as background material maintained by a separate store team (Google Ads: asset types for App campaigns). Visibility does not mean every reporting view accounts for every assembled ad or establishes a causal contribution.
There are two distinct jobs here. First, store content can supply ingredients for ad assembly. Second, the listing is a conversion destination after the ad interaction. An outdated screenshot can create a misleading product expectation in either role. A campaign library that demonstrates one workflow while the listing leads with another deserves a continuity diagnosis before another round of cosmetic creative variants.
Inspect the input and the handoff separately
Start by recording which store-derived title, icon, screenshot, video, description, and developer identity are observable in the campaign’s available asset views. Check their language, product version, claims, and relationship to the current concepts. Record what is actually visible, not what the team assumes Google must be serving. Automated assembly can crop, modify, or omit supplied material, so a source file is not a guarantee of the exact rendered combination (Google Ads: assets and ads in App campaigns).
Then inspect the landing listing as a visitor would. Does its opening sequence substantiate the promise in the ad? Is the relevant feature available in the advertised market and app version? Does the store copy explain the mechanism and resolve the objection rather than introduce a different offer? These checks create a bounded brief for a store-page correction; they do not justify a claim that the store team controls auction pricing.
Follow the measurable path without inventing a direct CPI lever
Store screenshots do not directly change auction CPI or CPA as a platform input. Do not describe a screenshot replacement as an auction-bid adjustment or attach a guaranteed acquisition-cost saving to it. The relevant path is store-page continuity and conversion, then install volume, then the quality and value of those users. At a comparable level of attributable paid traffic and spend, a change in the number of attributed installs can change an observed spend-per-install ratio. That arithmetic does not establish why the ratio moved or prove that screenshots caused it.
Keep the evidence chain explicit:
Record the ad concept, destination listing, market, language, device context, and dates of each material change.
Where available, compare listing visitors and installs using the store’s documented definitions. Distinguish a store conversion rate from a campaign-attributed install count; their denominators and attribution rules may differ.
Read campaign spend, install volume, and cost per attributed install at campaign level. Explain changes in traffic mix, bids, targeting, budget, and conversion timing before assigning a creative cause.
Check activation, retained behavior, purchase, or the app’s chosen value event in the approved product-measurement stack. More installs are not automatically more useful users.
State the remaining uncertainty: privacy limits, cross-device identity, view-through attribution, delayed events, small samples, and overlapping creative or product changes can prevent a clean ad-to-listing-to-user reconciliation.
This is an editorial measurement framework, not a Google guarantee that store-page changes improve delivery or quality. The screenshot and store-conversion guide owns the fuller conversion-testing decision; this page owns how store inputs fit the Google asset portfolio.
Use custom details pages only within the documented scope
Google currently documents custom details pages for App campaigns for installs promoting Android apps, using a Google Play custom store listing. This is a current eligibility statement, not a vague promise that a feature is coming soon and not a claim that all accounts or campaign types have identical controls (Google Ads: asset types for App campaigns). Check the controls available to the account before planning an audience-specific destination.
A tailored page can help a team maintain a coherent message between an eligible campaign and its listing. It still needs a truthful product promise, enough observable traffic to interpret, and a downstream event definition. Use the custom store listing versus experiment guide to decide whether the job is routing to a relevant page or testing a listing treatment. Neither mechanism substitutes for a causal acquisition experiment.
Use an asset portfolio, not a production quota
A portfolio should cover uncertainty. A quota covers file count. The two are not the same.
Use four portfolio roles:
Role Question it answers Production implication Evidence to protect --- --- --- --- Core concept Can the strongest current product argument continue to acquire useful users? Maintain healthy format coverage without constant cosmetic churn Campaign economics and post-install quality Challenger concept Does a different user problem, promise, or proof deserve investment? Change the commercial argument, not just the opening frame Incremental learning against the core concept Translation Can a viable concept work in another orientation, duration, language, or placement context? Preserve the hypothesis while adapting execution Whether the concept survives the new format Diagnostic asset Is a specific handoff or objection causing the loss? Isolate a mechanism, proof, store continuity, or onboarding expectation The transition the asset was designed to change
The production plan should state how many active concepts can be supported with coherent translations and enough downstream data to evaluate. It should not inherit a universal weekly output from another advertiser.
More files can expand eligibility and provide more combinations, but they can also fragment the learning record. If new work cannot be tied to a concept, a format gap, or a diagnostic question, it is not yet a production priority.
Define the measurement contract before reading the asset report
Google’s enhanced App asset reporting can show asset, campaign, and network views, with dimensions such as asset ID, state, status, orientation, and type. Google warns that asset-level reports may show higher totals because multiple components in one ad impression can each be counted, that served cost can differ from billed cost, and that auto-created asset costs may be excluded. For accurate spend and performance totals, Google directs advertisers to campaign-level totals (Google Ads: enhanced asset reporting).
This warning should shape the analysis. An asset report is not a ledger of isolated ads. Each asset may have appeared in combinations, networks, and user contexts that the operator did not assign evenly.
Create a four-layer measurement contract:
Asset layer
Use asset ID, type, orientation, approval status, delivery, conversion metrics, and the current reporting view to understand participation. Compare like with like. A vertical video and a headline do not compete for the same role.
Campaign layer
Use campaign totals for spend, installs, in-app actions, and conversion value. Record the bid strategy, optimization event, conversion windows, geo and language scope, and material changes. This is the layer for budget and aggregate efficiency decisions.
Product layer
Use first-party analytics or an approved measurement stack to evaluate onboarding, activation, retention, subscription, purchase, or another product-defined event. The campaign’s chosen conversion should map to a real event definition rather than a convenient label.
Economic layer
Use the finance-approved logic for net revenue, platform fees, refunds, variable costs, payback, and contribution. Do not call platform conversion value profit. Do not call an early cohort’s observed revenue lifetime value.
The layers do not need to agree perfectly. They need defined ownership and a reconciliation process.
How to interpret performance labels and combinations
Google’s single-campaign performance ratings are relative within the same campaign and asset type, based on the last 14 days. Changing the report date range does not change the current performance rating: the column always shows the current rating, not a reconstructed historical label. If comparing the underlying metrics with that label, use the last-14-days window and still preserve the campaign context. A Low or Best label is not a universal quality score, a post-install quality verdict, or an isolated experiment (Google Ads: App campaign asset reporting).
The cross-campaign report shows active assets from active ads, ad groups, and campaigns that have had impressions in the last 30 days. Asset data updates daily, not in real time; an asset may remain visible after a pause until the next update. Those rules describe inclusion and freshness, not a promise that every row represents an equally exposed test (Google Ads: App campaign asset reporting).
Auto-generated assets are excluded from asset reporting, so the sum of asset spend should not be expected to reconcile to overall campaign spend. This is separate from enhanced reporting’s component-counting and served-versus-billed-cost caveats. Use campaign totals for spend and aggregate performance; do not force a reconciliation by assigning the unexplained difference to a particular creative (asset reporting, enhanced reporting).
Keep a dated export and your decision notes if you need to review what the team knew at a prior point. A later report’s current label does not recover that historical state. Insufficient data can also prevent a useful rating; absence of a label is not proof that the underlying concept failed.
Use labels as prompts:
Learning: preserve stability unless an integrity problem requires intervention.
Low: investigate concept, execution, eligibility, format, and downstream quality before deleting.
Good: retain useful coverage; do not replace merely because it is not labeled Best.
Best: examine the concept and combination context, then create a deliberate challenger—not a superficial clone.
Unrated: decide whether the asset lacks enough comparable exposure, approval, or portfolio context.
Google’s top-combinations report displays a limited set of high-performing image, video, and text-only combinations and may remain empty without sufficient data (Google Ads: top combinations for App campaigns). A top combination shows what worked together under the observed system. It does not isolate which component caused the outcome, prove incrementality, or guarantee that a copied combination will retain performance.
Write conclusions at the level the evidence supports. “This concept deserves a new vertical execution” is often defensible. “This headline caused the lower CPA” may not be.
A practical weekly decision workflow
The cadence can be weekly, biweekly, or tied to conversion maturity. What matters is that the same decision sequence repeats.
Step 1: check integrity
Confirm approvals, broken links or deep links, store availability, event firing, value transmission, campaign changes, localization, and inventory constraints. Do not diagnose creative while the measurement or destination is broken.
Step 2: reconcile totals
Compare campaign spend and conversions with the product and economic views. Explain material differences in attribution, event timing, identity, refunds, and value definitions before ranking creative work.
Step 3: map delivery to portfolio roles
Label each supplied asset by concept, format, orientation, language, and role. Identify concepts with narrow delivery, formats with no approved asset, and files that duplicate an existing idea without answering a new question.
Step 4: read transitions
Move from delivery to install, activation, retained behavior, and monetization. Find the earliest material divergence. Strong install volume with weak activation suggests a different creative or store-expectation problem than weak delivery with strong observed user quality.
Step 5: choose one decision
Use one of five actions:
Maintain: preserve a useful core asset and its measurement conditions.
Translate: adapt a validated concept to a missing format or market.
Challenge: introduce a different commercial hypothesis.
Diagnose: isolate a store, onboarding, event, approval, or combination issue.
Retire: remove an asset whose role is no longer useful or whose claim, eligibility, or economics fails the agreed gate.
Step 6: update the learning ledger
Record asset IDs, concept, execution, dates, campaign goal, event, delivery context, observed result, uncertainty, action, and next question. Preserve what was learned even when the campaign later changes.
Common failure patterns
The library is diverse visually but identical commercially
Different creators, colors, and durations repeat the same promise. Add a challenger concept grounded in a different user situation, mechanism, or objection.
A “Best” asset attracts low-value users
The platform label and campaign outcome reflect the selected optimization and reporting context. Inspect activation, retention, purchase quality, and the expectation created by the asset before scaling its lesson.
Store-derived creative tells the wrong story
The paid team has strong supplied assets, but auto-created or store-sourced combinations emphasize an outdated feature. Update the store listing and verify what Google is using; do not assume uploaded campaign assets are the only public inputs.
A missing ratio is treated as a strategy failure
First distinguish eligibility from concept. Produce the missing orientation when it unlocks meaningful coverage, but do not expect a new crop alone to solve weak product positioning.
Every low-rated asset is removed at once
The campaign loses coverage and the team destroys its comparison context. Google recommends gradual replacement of Low assets after Learning rather than removing them all at once (Google Ads: App campaign creative guidance). Replace deliberately, preserve useful concept and format coverage, and record which question the new asset answers. A misleading claim or broken destination still warrants correction rather than waiting for a label.
The team copies a top combination
The recreated files mimic the visible execution but miss the user situation or mechanism that made the original combination coherent. Extract the concept before reproducing the style.
When this framework does not fit
Do not run a creative-portfolio program when the app is unavailable in the target market, the store listing is materially misleading, the event stack is unreliable, the product cannot retain the acquired users, or the business has not defined the action it values. Fix the active constraint first.
The framework also does not make asset-level reporting causal. Use a controlled experiment or incrementality method when the decision requires a causal answer. Automated delivery, combination effects, sparse events, and changing auction conditions limit what an observational asset report can establish.
Finally, do not apply one portfolio design to every app. A game, subscription utility, marketplace, fintech product, and health app have different demonstration, policy, trust, and value requirements. The concept map must begin with the product’s real customer decision.
How this page fits the Sharply Labs app-growth system
This page owns the reader job around Google App campaign creative assets: portfolio architecture, automated assembly, asset reporting, and production decisions. It does not replace the Google App campaign goal framework, which owns the install-versus-in-app-action decision. It complements the mobile app creative testing framework for Meta and TikTok, where the platform controls and testing mechanics differ, and the Google App Campaigns vs. Meta Ads comparison, which owns channel allocation.
Sharply Labs’ mobile app growth practice and performance marketing services connect campaign setup, creative systems, app-store conversion, event quality, post-install behavior, and acquisition economics.
When a Sharply Labs conversation is useful
A conversation is suitable for an app founder, growth lead, or UA team already running—or preparing to run—Google App campaigns when creative production is active but the asset library cannot answer which concepts, formats, and store inputs deserve the next investment.
Sharply Labs can examine the current asset portfolio, campaign goal, reporting definitions, store handoff, post-install event, and learning ledger, then define a bounded production and measurement plan. The useful output is a concept-and-format map, the first evidence gap to fix, and a next creative decision the team can interpret.
The conversation does not promise a lower CPI or CPA, a Best rating, more scale, or future revenue. Those outcomes depend on the product, market, auction, measurement, and execution.
Sources
The specification, asset-type, creative-guidance, assembly, reporting, and combinations sources below were checked on October 13, 2026. Platform documentation establishes current capabilities and reporting boundaries, not expected acquisition performance. Recommendations in this guide are editorial decision rules; eligibility, guidance, and interfaces can change.
Google Ads Help: About assets and ads in App campaigns — automated assembly, cropping, modification, and generated video; not a guarantee of an exact served combination.
Google Ads Help: App campaigns specs and format requirements — current text, image, and video boundaries; specifications and campaign-subtype eligibility can change.
Google Ads Help: Asset types for App campaigns — non-store/store assets, reporting visibility, and Android install-campaign custom details pages; not universal account availability or causal performance evidence.
Google Ads Help: App campaign creative guidance — independent text, gradual Low-asset replacement after Learning, full-frame visuals, app demonstration, and sound-off captions; directional guidance, not a promised result.
Google Ads Help: About enhanced asset reporting for App campaigns — reporting views and component/cost definitions; asset totals are not a campaign accounting ledger.
Google Ads Help: About asset reporting for App campaigns — current 14-day relative ratings, 30-day active-asset inclusion, daily updates, and excluded auto-generated assets; observational signals do not isolate causality.
Google Ads Help: About assets top combinations reports for App campaigns — observed top combinations subject to sufficient data; not a component-level causal test.
Google Ads Help: Best practices for App campaigns — retained background source from the original article, accessed September 13, 2026; recheck current setup guidance before implementation.