In this article
What Google recommends
Google’s October creative guidance recommends supplying varied formats and working from existing brand assets, creators or AI-assisted production. Its rule of thumb is at least three videos and three images across horizontal, vertical and square ratios. [1] Treat this as platform guidance, not a guarantee of better results for every account.
The agency’s next question is what those assets are meant to communicate. A folder full of resized images may cover dimensions while repeating the same weak message. Production volume and the quality of a test are different things.
Build a matrix around buying questions
Use rows for buyer questions and columns for useful formats. A software product might need a demonstration, a setup explanation and an objection response. A retail offer might need product use, a specific benefit and a delivery or fit explanation.
For each cell, name the claim, proof and destination. Do not fill every possible cell automatically. Prioritize the messages most likely to resolve uncertainty for a real prospect, then choose the formats needed for the campaign’s intended placements.
Keep invariants across the assets
Write down the offer, approved product facts, logo treatment, language and prohibited claims. These should remain consistent even when the story or crop changes. A comparison becomes difficult when each asset quietly advertises a different discount or product capability.
Prepare placement-specific compositions rather than blindly shrinking a horizontal image into a vertical frame. Review text at the size it will actually appear, check the opening video frames and confirm that critical information survives the intended placement treatment.
Make the production brief measurable
Our suggested brief contains the buyer question, concept, format, evidence, expected next step and test outcome. For a hypothetical agency product, a demonstration could show how a marketer finds the country driving a cost increase. Its success question might be whether it attracts relevant trial users who complete that review.
That is more specific than “make it modern and high converting.” The designer can make a clearer asset, and the marketer can explain what the test taught. Avoid manufactured interface results or simulated customer success presented as a real demonstration.
A small creative test matrix
| Review item | Question or evidence | How to use it |
|---|---|---|
| Demonstrate | One task and its real result | Prepare suitable horizontal, square and vertical compositions. |
| Explain | One practical buying question | Keep the offer and factual claims fixed. |
| Reassure | One genuine objection and supplied proof | Use authorized evidence. |
| Review | The useful campaign outcome | Inspect delivery context before comparing assets. |
A hypothetical software team starts from one recorded workflow and an approved still image. It can create a demonstration, an explanation of setup and an objection response without pretending each is a customer success story. The team then chooses appropriate formats for the intended campaign. This is a more coherent production plan than ordering nine unrelated AI images and deciding what they mean after delivery. Preserve the message and source asset for each version.
Interpret allocation before declaring a winner
Different assets may receive different amounts of spend or appear in different contexts. Review delivery volume, campaign settings and business outcomes before comparing their cost figures. An asset with one early result is not automatically the most reliable performer.
Keep an archive of final assets, dates and destinations. If a message works, the archive helps the next production cycle preserve what was actually tested. If the result is inconclusive, record why: sparse outcomes, unequal exposure or a destination change can all limit interpretation.
Use 1ClickReport to connect evidence and production
Start with the Analyst’s review of the connected client data and identify the largest commercial question. For still-image concepts, use AI Creative Director with the client’s reference asset and a clear brief. Review the output before exporting it into the campaign workflow.
Video production is a separate capability and may involve external tools or credits; do not describe a still-image workflow as automatic free video generation. The goal is a coherent path from observed problem to an approved asset and a measurable campaign test.
A first task to try
Using this client’s goals and connected campaign results, propose a Demand Gen creative matrix with three distinct buyer questions. For each, define the claim, proof, destination, suitable formats and business outcome. Keep approved brand facts and offer consistent. Do not invent customer results.Questions readers ask
Is one design in three sizes enough variety?
It covers sizes but may not test different messages. Separate format coverage from concept variety.
Does Google’s guidance guarantee a lift?
No. Use your own account’s qualified outcomes and a defensible comparison.
Sources and editorial method
Research checked October 26, 2026. Platform facts link to primary sources below. The worksheets and operating examples are our editorial recommendations; hypothetical examples are not customer results. This article is published by 1ClickReport.