One-Tap Demand Gen in 2026: 9:16 Shorts Creatives, Gmail Image Ads and the New Click Math
In September 2026 Google shipped a Demand Gen update that is easy to miss, because it adds no new campaign type and no new setting. One-tap Demand Gen simply removes a step: Shorts viewers used to tap an ad, get an expanded card, and only reach your site on a second tap. Now the first tap lands them on the page.
It sounds like an interface detail, but it changes two things at once for buyers: the creative requirement (full-bleed 9:16 instead of a reworked square) and the economics of the click. Below: exactly what shipped, which assets to prepare, how Gmail billing works now, and what to check before peak season.
What was in the September drop
| Change | What it is | What to do |
|---|---|---|
| One tap in Shorts | The intermediate expansion is gone: the tap goes to your landing page | Upload full-bleed 9:16 images |
| Gmail image ads with one tap | Same logic in the Promotions and Social tabs | Add the Gmail placement to YouTube-only campaigns |
| Business Agent for YouTube | Conversational AI alongside the video ad, answering from your product feed | Beta by signup; get the feed in order first |
| Affiliate location extensions | Promoted pins on Google Maps for retail networks and dealerships | Relevant if you sell through offline distribution |
All four sit inside Demand Gen and require no new campaigns — they work in what you already run. That cuts both ways: nothing breaks by itself, and nothing switches on by itself either, beyond what you prepare at the asset and placement level. The format fundamentals are in Demand Gen in Google Ads 2026.
9:16 is not “another size”, it is a different frame
Google’s guidance is specific: full-bleed 9:16 images, a recommended minimum of 1200×2136 pixels, and minimal text inside the image. The full-bleed frame removes the surrounding interface that used to eat attention.
The practical problem is that most advertisers drop in a resized square. It shows, and it underperforms. The difference between “vertical” and “built for vertical”:
- Composition. In 9:16 the vertical axis does the work: the product large and centred, with the space above and below treated as part of the frame rather than emptiness to fill.
- Text. “Minimal text” is not a formality here. In a full-bleed frame small type is physically unreadable on a phone, and large type eats the product. Let the image carry the message and let the headline and description carry the words.
- Safe zones. The top and bottom of the screen are covered by interface and gesture elements. Keep anything critical within the middle two-thirds vertically.
- The first frame. Feed decisions are instant: if it is unclear what is being sold, there is no second chance.
The cheapest route to a decent vertical is building for the format rather than stretching what exists. If you have no production capacity, generation covers part of the gap — but check the output with your own eyes; the six mandatory checks are listed in Asset Studio and Gemini Omni. The broader testing approach is in video ads for performance: how YouTube and Shorts sell.
Why vertical became the deciding factor
Short-form vertical video stopped being “another placement” some time ago — for a large share of the audience it is the default way to consume content on a phone. The non-obvious consequence for advertising is that this user is not searching, they are scrolling. They did not formulate a query, did not compare options, and were not planning to buy anything ten seconds ago.
That is the exact opposite of search logic, where the person arrived with intent. So mechanics that work in search do not transfer directly:
- There is no ready intent — the creative has to create it rather than capture it. A bare offer with no context reads as noise in a feed.
- Decisions take a fraction of a second — clarity beats completeness. One legible product beats a collage of five.
- Format errors cost more — an ad that was obviously not built vertical looks foreign in the feed and gets scrolled past automatically.
Removing the intermediate tap amplifies all three. A weak creative used to get a second chance: the viewer expanded the card and read the details there. Now there is no second screen — there is the feed and then your site. All the work of explaining what you sell and why it matters has moved into one image and the top of your landing page. How to create demand rather than only harvest it is covered in common Demand Gen setup mistakes.
Gmail: the new click math
The most underrated part of the drop is Gmail economics. Ads appear at the top of the Promotions and Social tabs, and the first tap goes straight to your landing page. In Demand Gen you pay for that first click; subsequent clicks through to your site are not charged.
One tap removes the step where part of the audience used to drop out. It is not more interest — it is less leakage between interest and your site.
Google reports an average: advertisers who add Gmail to Demand Gen campaigns see roughly 40% more conversions from image creatives at the same ROI. That is a global average across advertiser campaigns, not a promised result — treat it as a reason to test rather than a planning number. Your own benchmark only comes out of your own split.
The practical implication: if you run Demand Gen campaigns that are YouTube-only, adding the Gmail placement is one of the cheapest changes available this quarter. It needs no new creative — the same images work — and does not restructure the campaign. Check how it lands against your billing model in Demand Gen CPM billing.
Business Agent: what it is and how to prepare
Business Agent for YouTube is a conversational AI that sits alongside the video ad and answers questions about the product or brand from your product feed, without pulling the viewer out of what they are watching. Access is by signup and limited — this is a beta, not a mainstream format.
Preparing early is still worth it, and the preparation happens in your product data, not in the ads interface. The agent can only answer with what is in the feed: if that is a title, a price and a category, the answer will be thin. Which is why this update reads best alongside the conversational feed fields — covered in conversational attributes in Merchant Center.
The minimum preparation, which pays off regardless of beta access:
- accurate stock and pricing in the feed (a stale price inside a conversation is a complaint waiting to happen);
- real specifications instead of marketing phrasing;
- honest answers to the questions buyers actually ask;
- verified links: where the product card goes and what the person sees when they land.
One-tap Demand Gen changes what your landing page has to do
This is the consequence announcements talk about least and it is the most expensive one. The intermediate expansion used to act as both a filter and a warm-up: the viewer saw an expanded card, confirmed their interest and only then reached your site. Now there is nothing between the feed and your page.
- Traffic arrives colder. People who used to drop out at the middle step now reach the site. Visit-level conversion rate can fall while absolute conversions rise — that is the mechanic working, not breaking. Judge on cost per conversion, not conversion rate per visit.
- Speed matters more. The transition happens out of full-screen video inside a mobile app. Any load delay reads as a malfunction. Thresholds and measurement are in landing page speed and Core Web Vitals.
- The first screen must echo the creative. If the vertical showed a specific product, that product should be the first thing visible — not a promo banner or a pop-up. Mismatch here is the main driver of instant bounces.
- Pop-ups got more expensive. A mobile signup modal right after a Shorts tap kills the session almost reliably. If you must have one, trigger it on scroll rather than on a timer.
How to work through post-click losses systematically is in landing page conversion for paid traffic, and where exactly traffic dies shows up in the Landing Pages report.
A 9:16 creative checklist
To keep this out of the realm of taste, here is a checkable list. Each line is a yes or a no, not a “looks fine”.
| Check | Criterion | Common failure |
|---|---|---|
| Dimensions | At least 1200×2136, exactly 9:16 | A square with bars added top and bottom |
| Hero subject | Product fills the middle third vertically and reads in a second | Small product in an “atmospheric” interior shot |
| Safe zones | Nothing important in the top or bottom 15% | Logo or price hidden under interface elements |
| Text in image | Minimal; if present, large and under four words | A multi-line offer baked into the picture |
| Contrast | Subject separates from background effortlessly | White product on a pale background |
| Phone check | Opened on a real device at 50% brightness | Judged full-size on a desktop monitor |
| Landing match | The product in the creative is the first thing on the page | Creative shows one model, page is a generic category |
That last row deserves emphasis: creative-to-landing mismatch causes most instant exits. The format removed the screen that used to explain where the person was heading — that job now belongs to the top of your page.
If you have no production capacity
The real barrier is rarely understanding the format; it is not having hands to build fifteen verticals for different products. A workable sequence for a small team:
- Start with three products, not the catalogue. Take last month’s top spenders. Three good verticals produce more usable data than thirty mediocre ones.
- Reuse the product photography you already have. If the item was shot on a plain background, it can be recomposed into vertical without a reshoot — the point is rebuilding the composition, not padding the edges.
- Use generation for variations, not for the base. Generation is good at multiplying backgrounds and tones around a finished product shot. It is weaker as the source frame, especially when exact product appearance matters.
- One visual template per category. A shared grid — product centred, caption in one place, identical margins — saves weeks and makes results comparable to each other.
- Change one variable at a time. Background, or angle, or caption — not all three. Otherwise you get a winner and no idea why it won. Correct comparison logic is in ad A/B testing in 2026.
Affiliate location extensions: who they are actually for
The fourth item in the drop sits outside the main story, which is why it gets little coverage: promoted pins on Google Maps for retail networks and dealerships. The idea is straightforward — if your product sells through a dealer or retail network rather than directly, the pin appears where someone is already looking at a map and choosing where to go.
It is niche but underrated for manufacturers and brands with offline distribution: the traffic goes to a store, not a site, so it has to be measured differently — through visits and calls rather than carts. How to count that is covered in store visit conversions. If you have no offline sales, skip this one entirely.
What to do before peak season
- Audit your assets. Export current images and mark which are genuinely 9:16 at 1200×2136 and which are resizes. Replace the resizes first in your highest-spend categories.
- Add Gmail to YouTube-only campaigns. No new creative required; the change takes minutes.
- Check the feed link. It underpins product formats and any future Business Agent access — and clear your disapprovals while you are there, see feed diagnostics.
- Rebuild the top of your landing pages. For your top five by spend: product visible immediately, no blocking pop-up, action button within thumb reach.
- Record a baseline. Write down current performance before you change anything, or in December you will not be able to separate format effects from seasonality.
- Stagger the changes. Do not launch verticals, Gmail and new landing pages on the same day. A week between steps is enough.
- Revisit audiences. The format changed; audience logic did not — current approaches in custom segments and intent audiences.
Not breaking what already works
- Do not delete old assets immediately. Add verticals alongside and let the system choose. Removing a working creative for experimental purity during peak season is expensive principle.
- Watch the placement mix. Adding Gmail will shift budget. That is expected, but if the YouTube share drops sharply, check whether you hit a budget ceiling rather than a creative problem.
- Keep your evaluation window longer than the learning period. Any meaningful asset or placement change restarts adjustment. Conclusions on day three are reading noise.
And the thing most teams skip: log the changes — date, campaign, what exactly changed. In six weeks, when the numbers diverge, that log is the only way to connect cause and effect; see Google Ads change history.
What to expect and what not to
- Expect more site visits for the same budget: a step where people dropped out is gone.
- Expect a different traffic mix, with more impulse taps.
- Expect a higher share of mobile conversions — and therefore more sensitivity to mobile layout and mobile payment methods.
- Do not expect a resized square to deliver. The format rewards creative built for it.
- Do not expect to replicate the 40% figure. It is a global average for one placement, not a target for your account.
- Do not expect the format to compensate for a weak landing page. It exposes one faster.
- Do not expect immediate results from Business Agent: it is a limited beta, and readiness depends on your feed rather than your campaign settings.
One caveat on measurement: if you enable this before Q4, seasonality overlaps everything. The only clean read comes from a geo split or twin campaigns, not a month-over-month comparison. If you need the infrastructure to run those tests in parallel, see Google Ads agency account rental and the rest of the PPC Rebels services.
FAQ
What changed in Demand Gen in September 2026?
The intermediate tap is gone: in Shorts and Gmail the first tap on an image ad goes straight to your landing page. The drop also announced the Business Agent beta for YouTube and affiliate location extensions on Maps.
What size should Shorts vertical creatives be?
Full-bleed 9:16, with a recommended minimum of 1200×2136 pixels and minimal text inside the image.
Do I need a new campaign for the new format?
No. Upload correct 9:16 assets to your existing Demand Gen campaigns and add the Gmail placement if you are not already using it.
Can I just crop a square creative to 9:16?
Technically yes, in practice almost always badly. A full-bleed vertical is composed differently: hero subject in the central zone, minimal small type, legible on a phone.
How are Gmail clicks billed in Demand Gen?
You are charged for the initial click; subsequent clicks through to your site within that interaction are not charged.
What is the 40% Gmail figure?
Google reports that advertisers adding Gmail to Demand Gen see on average 40% more conversions from image creatives at the same ROI. It is a global average, not a guaranteed outcome.
What is Business Agent for YouTube?
A conversational AI beside the video ad that answers product questions from your feed without the viewer leaving the video. It is currently available by signup in limited release.
Why might visit conversion rate fall after enabling this?
Because the intermediate step that acted as a filter is gone. More people reach the site, including less-committed ones. Judge on cost per conversion rather than conversion rate.
What should I fix on the landing page first?
Mobile load speed, a first screen that matches the creative, no timer-triggered pop-up, and an action button reachable without scrolling.
How do I measure the effect if I launch before Q4?
With a split only: by geography or twin campaigns. Month-over-month comparison in peak season mixes format effects with seasonal demand.
Do these formats need a product feed?
Image formats do not, but a feed is required for product formats and for Business Agent. If those are on your roadmap, get the feed in shape first.
Should I turn everything on at once?
No. Stagger verticals, the Gmail placement and landing page edits across weeks, otherwise the result cannot be decomposed.
