Google's AI Mode ad formats are still tests, so fix your inputs first
At Google Marketing Live Google showed Gemini-built ad formats for AI Mode. What is live, what is still a test, and the groundwork worth doing without rebuilding your account.
Google Marketing Live on May 20 was mostly about ads inside AI Mode, Google's conversational search experience. Strip away the demos and the useful information is narrower: two formats are being tested, two more are described as coming in the next few months, and Google's advice on how to prepare is to run AI Max for Search, AI Max for Shopping and Performance Max. For an app marketer the practical consequence is mostly indirect, but there are decisions worth making now.
What Google said is live and what is not
From Google's own post on the new Search ad formats:
- Conversational Discovery ads are being tested in AI Mode. Gemini builds ad creative that answers the person's specific question, highlighting the features that fit.
- Highlighted Answers are also being tested. When AI Mode returns a list of recommendations, a relevant, high-quality ad can appear as one entry on that list.
- AI-powered Shopping ads, where Gemini selects products and writes an explainer on why a product may fit, are coming in the next few months.
- Business Agent for Leads, a Gemini-built agent inside the ad that answers questions from your website instead of sending people to a lead form, is also described as coming in the next few months.
- Direct Offers is being expanded with promotion bundling, native checkout for merchants using Google's Universal Commerce Protocol, and travel offers. Native checkout is already added, while bundling and travel are labeled coming soon.
Each format carries an independent AI explainer next to your creative, and ads stay labeled "Sponsored." That last point matters: part of what the shopper reads next to your ad is text you did not write.
Separate the signal from the stage show
Google gave no timeline for broad availability of the two tested formats and no performance data. Announcements at an annual keynote are a statement of direction, not a guide to what your account will receive next month. Our rule is to act on what changes your inputs, and wait on what changes your outputs.
Inputs are things you control: your feeds, your assets, your landing pages, your conversion tracking and your match and keyword settings. Outputs are where Google places you and what it writes around you. Spend your time on inputs.
Groundwork worth doing now
- Audit what Google would read about you. AI Mode formats are built from your website, feed and assets. Check that your product pages, app store listing copy and landing pages say accurate things in plain language. If a page is vague, an AI explainer will have little to work with, and if it is wrong, the explainer may repeat the error.
- Set up a controlled test of AI Max for Search. Google points to it as the way to prepare for the new formats. Do not convert your best campaign. Run it as an experiment on a campaign with enough conversion volume to read, against a clear control, and compare on cost per outcome you actually pay for, not on reported conversions alone.
- Tighten brand and negative controls first. The more freedom you give automated matching, the more your negatives, brand exclusions and asset pinning decide what you will not appear for. Review your negative lists and search term reports before you widen anything.
- Verify conversion tracking quality. Automated formats optimize toward whatever you report. Duplicate or low-value conversion events teach the system the wrong lesson faster. Check that the event you bid on is the one tied to revenue or retained users.
- Plan a measurement approach that does not depend on placement reports. If AI Mode results are folded into standard Search reporting, you may not be able to see them separately. A geo or time-based holdout, or a before and after on branded and non-branded search volume, gives you something to hold on to.
What it means for app advertisers specifically
Google's post is about Search and Shopping. It does not describe a change to app campaigns, so do not read it as one. If you drive installs mainly through app campaigns, the relevance is indirect: more of the high-intent research that precedes an install may happen in AI answers, where your ad and a Gemini-written explainer sit side by side. That raises the value of your store listing and web presence, since Google may use them when it describes you.
If you also run web Search for an app or an ecommerce business, then the formats are directly relevant, and the test plan above applies.
A stance on pace
We would not shift budget toward these formats yet. They are tests with no published results, you have limited control over the explainer text, and there is no evidence that they outperform standard Search ads. We would invest a small, fixed learning budget in an AI Max experiment, set a review date, and define in advance what outcome would justify a second round. If you have no experiment budget this quarter, a clean-up of feeds, negatives and tracking is still the right work.
The takeaway: Google has shown direction, not results. Fix your inputs now, run one disciplined AI Max test, and hold your budget until the formats leave testing and show data you can verify.