Demand Gen: Formats, Placements, Maps and Channel Controls

by Francis Rozange | Apr 4, 2026 | Google Ads

demand-gen-formats-placements-channel-controls-1

Demand Gen in 2026: the campaign type that ate Discovery, Video Action, and a slice of paid social

Demand Gen is no longer a beta curiosity grafted onto Discovery. It is, as of 2026, the upper-funnel pillar of Google Ads, sitting between Performance Max (lower-funnel, retail-heavy, Merchant Center driven) and Search (intent capture). It runs on YouTube (in-stream, in-feed, Shorts), Discover, Gmail, the Google Display Network where opted in, and since late 2025, Google Maps. The campaign type that Search Engine Land covered at its global rollout in October 2023 as the Discovery successor has, in two years, swallowed Video Action Campaigns, picked up shoppable carousels, gained channel-level controls, and absorbed Maps as a placement.

The pitch from Google has not changed: reach 3 billion logged-in users a month with visual creative that mimics what works on paid social, but with first-party Google signals doing the targeting. What has changed is the level of control advertisers actually get. Channel controls launched in March 2025 and matured through the year, lookalike segments shifted from hard similarity tiers to AI-suggested expansion (effective March 15, 2026), and Maps moved from a passive placement nobody could control to a selectable channel with its own reporting line. This guide walks through what Demand Gen is in 2026, what it does well, where it competes with Performance Max, and how to run a 90-day pilot that does not bleed budget.

What Demand Gen is, and how it got here

From Discovery to Demand Gen, October 2023

Discovery campaigns launched in 2019 as a way to buy native placements across YouTube home feed, Gmail Promotions and Social tabs, and Google Discover. They were image-only, had thin reporting, and never broke through as a serious upper-funnel channel. In May 2023, Google announced Demand Gen at the Google Marketing Live event as the replacement, and the rollout completed globally in October 2023. Search Engine Land’s coverage of the global launch framed it as a paid-social challenger : video plus image plus carousel, lookalike audiences, and creative built for visual feeds rather than search results pages.

In early 2024, Google migrated all remaining Discovery campaigns to Demand Gen automatically. The migration was not optional, and accounts that did not run the upgrade themselves had it run for them. By Q3 2024, Video Action Campaigns (the lower-funnel YouTube format that drove most direct-response YouTube spend for years) were also slated for sunset and migration into Demand Gen. Search Engine Land documented the Video Action sunset and the forced upgrade path : by mid-2025, every dollar that used to flow into VAC now had to flow into Demand Gen.

Expansion 2024 to 2026

The 2024 to 2026 expansion went in three directions. First, surface coverage : the Google Display Network was added as an opt-in placement in early 2025, putting Demand Gen on roughly 90% of the global internet by inventory. Second, channel-level controls : in March 2025, Google rolled out the ability to enable or disable individual channels (YouTube in-stream, YouTube in-feed, YouTube Shorts, Gmail, Discover, GDN) at the campaign level, and to read separate reporting lines for each. Third, Maps : announced in December 2025 as a selectable Demand Gen placement, with promoted pins and Maps-feed cards rendered from the same creative pool.

The throughline is straightforward. Google wants Demand Gen to be the only campaign type advertisers need for upper-funnel visual demand creation. Performance Max stays the answer for retail direct-response with Merchant Center. Search stays the answer for captured intent. Demand Gen is everything in between : the channel where you build awareness, push consideration, and feed retargeting pools that Search and PMax then close.

The four creative formats and where each one earns its keep

Single image

Image got a real push in 2025. Static creative now serves natively on every Demand Gen surface : YouTube in-feed thumbnails, Gmail Promotions cards, Discover feed cards, and Maps. Specs are simple : JPG or PNG, 300×300 minimum, 5MB max, with three aspect ratios recommended (1:1 square, 1.91:1 landscape, 4:5 portrait). For Shorts, ship a 9:16 vertical asset at 1080×1920 if you want full-bleed rendering.

Image is the right format when production budget is constrained, when the offer is visually obvious (a product shot, a price tag, a transformation), or when you need to scale across SKUs without filming each one. Tinuiti’s case-study work on Demand Gen consistently flags image-plus-video as outperforming video-only for retail, by margins of 15 to 25% on conversion rate at equivalent CPA. The lesson : do not skip image because video felt more impressive in the brief. Mix.

Video

Video is still the engagement workhorse on YouTube. In-stream (skippable, plays before or during YouTube videos) takes 16:9 or 4:3 at 6 to 60 seconds. In-feed (home, search results, watch-next) accepts the same ratios but rewards a strong first three seconds. Shorts wants 9:16 at 3 to 60 seconds, with the hook in the first second because Shorts viewers swipe at speed.

The 2025 to 2026 lesson on video creative is unflattering for agencies that built their YouTube practice on broadcast-style production. User-generated content, founder-to-camera, and product-demo videos shot on a phone now routinely beat studio creative on Demand Gen. Search Engine Land has documented multiple campaigns where 30-second product demo videos drove 30 to 35% better conversion rates than polished brand films at the same media spend. The reason is not artistic, it is algorithmic : Demand Gen’s optimization rewards completion rate and click-through, and demo videos hold attention longer than brand stories on a feed surface.

Carousel

Carousel is the format paid-social refugees recognize on sight : 2 to 10 cards, swipe to browse, each card with its own headline (25 char), description (90 char), and image. It works for product-line storytelling, before-and-after sequences, and feature walks. The performance sweet spot lives at 5 to 7 cards, where engagement stays high without scroll fatigue.

One important constraint : carousel and product feed do not coexist in the same Demand Gen campaign. If you want carousel, you upload assets manually and lose the dynamic Merchant Center pull. If you want feed-driven dynamic ads, you give up carousel. Most retail accounts split the question : product feed campaigns for catalog coverage, carousel campaigns for seasonal storytelling.

Product feed

Product feed Demand Gen pulls from your Google Merchant Center, renders products dynamically into ad units, and updates when your inventory does. Setup needs the feed to have Shopping ads enabled as a marketing method, and Google recommends 4 products minimum. New feeds take up to 3 days to approve, so do not plan a Friday launch.

Filtering is granular : brand, product ID, condition, custom labels, product type, and ad-hoc groups. Local inventory ads on Demand Gen with product feed surface real-time local availability, which is why Maps plus product feed has become the favorite combination for retailers with physical locations. Google’s own data, repeated in Think with Google case roundups, shows advertisers who add a product feed to an existing Demand Gen campaign typically see a 30 to 35% lift in conversions at flat CPA. That is the single highest-leverage change a retail Demand Gen account can make.

Channel controls : the March 2025 shift, and the Maps addition in December

What channel controls actually unlock

Before March 2025, Demand Gen distributed budget across channels at Google’s discretion. You had no levers : if the algorithm decided 80% of your spend should hit YouTube and 20% Discover, that is what happened. The March 2025 update gave advertisers three tiers of control. Tier one : “All channels” (default), runs everywhere including the Google Display Network. Tier two : “Google-owned and operated”, excludes GDN to avoid third-party publisher inventory. Tier three : individual channel selection, where you check or uncheck each placement (YouTube in-stream, YouTube in-feed, YouTube Shorts, Gmail, Discover, GDN, and as of December 2025, Maps).

The reporting layer matched the controls. Each channel got its own line in the campaign report, with impressions, clicks, conversions, CPA, and ROAS broken out. For the first time, an advertiser could answer the question : “What is my Discover-specific CPA versus my YouTube Shorts CPA ?” without exporting raw data and rebuilding the table.

Maps as a selectable placement, December 2025

Maps moved from passive (Google could place ads there, you could not control it) to active (selectable channel with its own reporting). Promoted pins surface when users search local categories : “coffee near me”, “veterinary clinic”, “bike shop”. The pin shows the business location, distance, rating, and a call-to-action. For multi-location retail, restaurants, fitness chains, automotive, and home services, Maps changed the math on local Demand Gen.

Early data from advertisers who tested Maps in Q1 2026 shows store-visit cost per acquisition 20 to 30% lower than Local Services Ads in the same categories. Tinuiti and other agencies have published pilot results showing fitness studio chains and quick-service restaurants taking 15 to 20% higher store-visit attribution from Demand Gen with Maps enabled versus YouTube-only Demand Gen. The caveat : Maps performance correlates heavily with Google Business Profile quality. A 4.6-star location with 200 reviews and complete photos gets pin priority. A 3.8-star location with thin profile data gets buried regardless of bid.

How to think about channel allocation

The March 2025 controls do not mean you should hand-tune percentages. The 2026 algorithm does dynamic allocation : if you set 40% YouTube / 30% Gmail / 20% Discover / 10% Maps, the actual spend often lands at 50/20/15/15 because the algorithm prioritizes channels where each individual user is most likely to convert. The lever is not “set the percentages right”, it is “disable the channels that do not earn their keep”. For a B2B SaaS account, GDN almost always underperforms and gets disabled. For local retail, Discover often runs at acceptable CPA but Maps does better, so budget tilts to Maps. For pure brand awareness, all channels stay on.

The right discipline : start with all channels on, run two to three weeks, accumulate at least 50 to 100 conversions per channel before judging, then disable the bottom one by ROAS. Repeat. Do not disable a channel after 48 hours because the CPA looked rough : Demand Gen needs learning time, especially on a new placement like Maps where 2026 data is still thin.

Lookalike audiences : the Similar Audiences replacement, and the March 2026 algorithm shift

The Similar Audiences sunset, then the lookalike rebuild

Google deprecated Similar Audiences across Search, Display, and Video in August 2023 after the third-party cookie deprecation timeline made cohort-based similarity untenable. Search Engine Land documented the removal and the migration path : Similar Audiences became Optimized Targeting on most campaign types, and lookalike segments became the explicit similarity tool on Demand Gen. Lookalike on Demand Gen takes a seed audience (Customer Match list, website visitors, app users, YouTube channel subscribers) and finds users who match it. Through 2024 and 2025, lookalikes operated with strict similarity tiers : narrow (top 2.5%), balanced (top 5%), broad (top 10%).

March 15, 2026 : tiers become signals

On March 15, 2026, the lookalike model shifted from hard tiers to AI-driven expansion. The seed audience and the tier you select now act as suggestions to Google’s optimization model, which can expand beyond the tier if it identifies users with high conversion probability outside the strict similarity band. In practice : a “narrow 2.5%” lookalike might serve to users in the 1 to 4% similarity range, with the algorithm picking the slice it expects to convert. Search Engine Land covered the change at announcement, and most early test data shows AI-expanded lookalikes driving 8 to 14% more conversion volume at equivalent or slightly lower CPA than the strict-tier baseline.

Advertisers who want the old behavior can opt out, but the opt-out path is opaque and the default is the new model. The practical implication : seed quality matters more than ever. A lookalike off a high-LTV customer list with 5,000 entries will outperform a lookalike off a 100,000-entry website-visitor list. Refresh seeds quarterly. Drop low-LTV cohorts before they pollute the signal. Build seeds from purchase events, not page views.

What to test in the first 60 days post-shift

Run a parallel test : keep one ad group on the legacy strict-tier behavior (opt out), put another on the new AI-expanded model. Same seed, same creative, same bid strategy. Measure conversion rate, CPA, and incremental volume across 30 days. The result tells you whether the AI expansion earns its keep on your specific audience. For most accounts the answer is yes, but B2B accounts with niche seeds sometimes do better on strict tiers because the AI’s expansion drifts into broader users who do not convert at the price point.

Customer Match : the foundation Demand Gen actually leans on

Why Customer Match matters more on Demand Gen than on Search

Search runs on intent, and intent is observable in the query. Demand Gen runs on signal, and signal comes from your first-party data. Customer Match is the upload mechanism : hashed email addresses, phone numbers, mailing addresses, mobile device IDs, fed into Google Ads to build audiences that match against logged-in Google users. The match rate varies by industry, but for consumer brands with email-rich CRMs, 50 to 70% of an upload typically matches.

Once matched, a Customer Match list does three jobs in Demand Gen. It seeds lookalikes (the AI extrapolates from your best customers to find similar users). It enables exclusion (run New Customer Only mode by excluding the existing customer list). It enables retargeting at scale across YouTube, Gmail, Discover, Maps, and GDN with the same list. Without Customer Match, Demand Gen falls back on optimized targeting and contextual signals, and performance drops by 20 to 40% on most accounts that A/B tested it.

Building a Customer Match foundation

The minimum viable Customer Match foundation has three lists. First, all customers (every email that ever bought, used as the lookalike seed and the New Customer Only exclusion). Second, high-LTV customers (top quartile by spend, 12-month window, used as the premium lookalike seed). Third, churned customers (bought once, no repeat in 12 to 24 months, used for win-back campaigns).

Refresh monthly. Customer Match lists decay : email addresses change, customers churn, new ones arrive. A 12-month-old static list performs noticeably worse than a monthly-refreshed list. The refresh can be automated through the Google Ads API or the Customer Match upload tool, and most CDPs (Segment, mParticle, RudderStack) support direct sync to Google Ads.

New Customer Only mode

Demand Gen added New Customer Only mode in late 2025. The setting flags the campaign to exclude users with prior purchase or engagement history, leaving only acquisition-mode targeting. Google’s internal A/B testing across Q1 to Q2 2025 showed New Customer Only campaigns improved the new-to-returning ratio by 11.5% on average, with a 3% reduction in new-customer acquisition cost. The feature works best when paired with Customer Match exclusion lists and a high-quality lookalike seed.

Bid strategies : what to use, and when

Maximize Conversions and Target CPA

Maximize Conversions is the entry-level bid strategy : tell Google to spend the daily budget on whatever generates the most conversions. It is the right starting point for new campaigns with no historical conversion data, where Target CPA would not have enough signal to optimize. Run Maximize Conversions for two to three weeks, accumulate 50 to 100 conversions, then transition to Target CPA with the historical CPA as the target.

Target CPA tells Google to optimize toward a specific cost per acquisition. The strategy works when conversion volume is high enough for the algorithm to learn (50+ conversions per month per ad group is the rule of thumb). Setting the target too aggressively (50% below current CPA) causes spend to collapse because Google cannot find users at that price. Setting it too loose (50% above current CPA) wastes budget on marginal conversions. The sweet spot is current CPA minus 5 to 10%.

Target ROAS for retail

Target ROAS replaces Target CPA when the campaign uses product feed and tracks revenue, not just conversions. The setting tells Google to optimize for a specific return on ad spend ratio. For retail Demand Gen with product feed, Target ROAS is the strategy that matches what the business actually cares about (revenue per dollar spent, not just count of conversions).

Target ROAS needs at least 30 days of conversion-with-revenue data and 100+ revenue-tracked conversions per campaign before it stabilizes. Below that threshold, the algorithm reverts to conservative bidding and spend collapses. Most retail accounts run Maximize Conversion Value (the no-target version) for the first 30 to 45 days, then transition to Target ROAS once the data is thick enough.

Maximize Clicks for awareness

For pure awareness campaigns where conversions are not the metric, Maximize Clicks remains valid. The strategy targets cost per click rather than cost per acquisition. It is the right setting for top-funnel campaigns where the goal is to push users into a remarketing pool that Search or PMax will close later. The trap : Maximize Clicks can run away on broad audiences and burn budget on uninterested traffic. Pair it with tight audience signals (Customer Match lookalikes, in-market segments) and a daily budget cap.

Creative best practices : what 2026 testing actually shows

Lifestyle imagery beats product-on-white

The single most consistent creative finding across 2025 to 2026 Demand Gen testing : lifestyle imagery (product in use, in context, with people) outperforms product-on-white (catalog-style isolated product shots) by 30 to 50% on click-through rate, with conversion rate holding flat or slightly higher. The reason is feed-context : Demand Gen surfaces are visually similar to Instagram and TikTok, where lifestyle creative is the native style. Catalog shots feel like ads. Lifestyle shots feel like content.

The exception : product feed Demand Gen pulls catalog images directly from Merchant Center, and those images should follow Merchant Center guidelines (white or neutral background, full product visible, no text overlay). The mix : product feed for catalog coverage, lifestyle creative for image and video ads on the same campaign or in parallel campaigns.

No logos in the main creative area

Google’s Demand Gen creative guidelines, repeated in Think with Google creative best practice articles, are explicit : do not place a brand logo in the main creative area of an image or video ad. Use the logo asset slot, which renders separately. Logos in the main creative area trigger a “looks like an ad” signal that suppresses engagement, especially on Discover and Shorts where users are scrolling for content, not for promotional material.

Same logic applies to text overlays. Demand Gen surfaces have headline and description fields that render outside the creative. Burning text into the image or video duplicates the text, looks like a banner ad, and loses click-through. Keep the creative clean. Let the text fields do the verbal work.

Clear value proposition in the first three seconds

For video, the first three seconds carry the load. The hook (what the product is, what problem it solves, why the viewer should care) lands in seconds 0 to 3, the demonstration runs seconds 3 to 20, and the call to action closes seconds 20 to 30 for short-form or 20 to 60 for in-stream. Videos that bury the value proposition past the three-second mark lose 50 to 70% of viewers before they get to it.

Shorts is the unforgiving extreme. Shorts viewers swipe in under one second if the hook does not land. The hook for Shorts has to be visual (the product, the transformation, the surprising visual) in the first frame, with verbal hook in the first second. Shorts that open on a slow brand intro die.

Asset volume drives algorithmic optimization

Demand Gen’s optimization is asset-level. The algorithm tests combinations of headlines, descriptions, images, and videos to find what works for each user. The more assets it has, the better it optimizes. Google’s published guidance and Tinuiti’s case-study work converge on the same numbers : campaigns with 8 to 10 assets per type outperform campaigns with 3 to 4 by 15 to 22% on conversion rate. Asset minimums : 5 headlines, 5 descriptions, 5 images, 2 videos. Better : 10 headlines, 10 descriptions, 10 images, 5 videos.

Demand Gen versus Performance Max : the decision

What each one is designed to do

Performance Max is the lower-funnel direct-response engine. It runs on every Google surface (Search, Shopping, YouTube, Display, Discover, Gmail, Maps), but its strength is Shopping and Search inventory backed by a Merchant Center feed. PMax optimizes for direct conversion : a sale, a lead, a phone call. It needs a feed and a high-volume conversion event to learn.

Demand Gen is the upper-funnel demand-creation engine. It runs on the visual surfaces (YouTube, Discover, Gmail, Maps, GDN) and skips Search and Shopping. It optimizes for engagement and conversion on visual creative. It does not need Shopping inventory, but it benefits from a Merchant Center feed for product feed ads. It is built for the moment before the user types a query.

The retail decision

For a retail account with Merchant Center, the right setup in 2026 is both. Performance Max takes the bottom of the funnel : capture purchase intent, push Shopping inventory, close conversions. Demand Gen takes the top : build awareness, push consideration, fill the retargeting pool that PMax then closes. The two campaign types feed each other. Demand Gen drives the user into the awareness layer, PMax closes them when they search.

The split that works for most retail accounts : 60 to 70% PMax (where the direct ROAS lives), 25 to 35% Demand Gen (where the demand-creation ROAS lives), 5 to 10% Search on branded and high-intent terms. The mistake is running only PMax and wondering why incremental conversion volume plateaus : PMax is harvesting demand, not creating it. Without Demand Gen feeding the top of the funnel, the harvest goes flat.

The lead-gen decision

For lead generation (B2B SaaS, professional services, consulting), the answer is more Demand Gen and less PMax. PMax for lead gen tends to chase low-quality leads (form fills from users with no real intent) because the algorithm optimizes on conversion count, not lead quality. Demand Gen, run on a high-quality Customer Match seed and lookalike expansion, drives more qualified leads at higher CPA but better downstream conversion.

The split for B2B : 60 to 70% Demand Gen (visual, awareness, consideration), 20 to 30% Search on intent keywords (capture the qualified prospect when they search), 5 to 10% PMax for branded and direct-response. The reverse of the retail split, because the funnel is different : B2B is long, awareness-driven, and human-judgment-heavy. PMax is too brute-force for it.

The local-business decision

For local businesses (restaurants, fitness, home services, automotive), Demand Gen with Maps enabled is the answer. PMax for local works for retailers with multi-location inventory, but for service businesses without Merchant Center, Demand Gen plus Maps plus Customer Match is the stack. Run Maps as a primary channel, YouTube and Discover as supporting channels, and use Customer Match exclusion to keep new-customer acquisition pure. Local Services Ads stay valuable for the categories Google supports, but Demand Gen plus Maps now matches or beats LSA cost per store visit in most pilots.

The 90-day Demand Gen testing playbook

Days 1 to 14 : foundation

Build the Customer Match foundation : upload all-customer, high-LTV-customer, and churned-customer lists. Verify match rates above 50%. Set up conversion tracking with revenue (for retail) or lead-quality scoring (for B2B). Audit the Merchant Center feed if applicable : product images present, descriptions complete, prices accurate, GTIN and brand fields populated.

Build creative inventory : 8 to 10 images (lifestyle), 3 to 5 videos (mix of demo, founder-to-camera, and lifestyle), 10 headlines, 10 descriptions. Do not launch with 3 assets. The first 14 days are setup, not campaign launch.

Days 15 to 30 : pilot launch

Launch one Demand Gen campaign with all channels enabled, Maximize Conversions bid strategy, all-customer lookalike as the audience, and the full creative inventory. Set the daily budget at a level that produces 10 to 20 conversions per day across the campaign (use historical CPA as the input : if CPA is 30 EUR and you want 15 conversions a day, set the budget to 450 EUR).

Do not touch the campaign for the first 14 days of running. Demand Gen needs learning time, and intervening early teaches the algorithm nothing. Watch the data, write down what you see, do not adjust. The temptation to “fix” the campaign in week one is the most expensive mistake in Demand Gen.

Days 31 to 60 : optimization

By day 30 the campaign should have 200 to 400 conversions across channels. Read the channel-level reporting. Identify the bottom channel by ROAS or CPA, disable it. Identify the top creative assets by engagement, weight the campaign toward them. Transition the bid strategy from Maximize Conversions to Target CPA, with the target set 5 to 10% below the historical CPA.

Build a second campaign : New Customer Only mode, high-LTV-customer lookalike as the seed, lifestyle creative only. The goal is acquisition, separated from retargeting and existing-customer reach. Run it parallel to the original campaign for the next 30 days.

Days 61 to 90 : scale and split

By day 60 you have two campaigns and 60 days of data. Read the lookalike performance : strict-tier versus AI-expanded (post March 15, 2026). Decide which to keep. If the account is retail, layer in product feed as a third campaign with Target ROAS. If the account is B2B, layer in Gmail-focused Demand Gen for high-attention placement. If the account is local, push Maps allocation to 40 to 50% of total spend.

Day 90 review : compare the three campaigns on conversion rate, CPA, ROAS, and lead quality (B2B). Kill the weakest. Scale budget on the strongest. Document creative learnings : which images and videos drove the wins, which headlines outperformed. Use those learnings to seed Q2.

Common mistakes

Asset poverty

The most common Demand Gen failure : launching with 3 to 4 creative assets. The algorithm cannot optimize what it does not have. Campaigns with thin asset pools converge on the one or two assets that work best for one or two user segments, then stagnate. Volume in : volume out. 8 to 10 images, 3 to 5 videos, 10 headlines, 10 descriptions is the floor, not the ceiling.

Killing channels too early

Disabling a channel after 48 hours because the CPA looked rough is the second most common mistake. Demand Gen needs at least two weeks per channel and 50 to 100 conversions per channel before the data is reliable. The 48-hour CPA reflects the learning phase, not the steady state. Wait. If after 14 days and 100 conversions the channel is still underperforming, then disable.

Stale lookalike seeds

A 12-month-old static seed performs noticeably worse than a monthly-refreshed seed. Lookalike audiences decay : the seed customers age, their demographics shift, the lookalike model trained on them drifts. Refresh quarterly minimum, monthly preferred. Automate the refresh through the API or a CDP integration.

Product feed neglect

For retail, an unmaintained Merchant Center feed kills product feed Demand Gen. Missing images, incomplete descriptions, expired prices, disapproved GTINs all suppress feed ads. Audit the feed monthly. Fix disapprovals within 48 hours. Add custom labels for the top sellers and configure the campaign to weight toward them.

Logos and text overlays in creative

Branding in the main creative area suppresses engagement. Logos go in the logo slot. Headlines go in the headline field. Descriptions go in the description field. The main creative stays clean : product, lifestyle, demonstration. The “looks like an ad” signal is real, and Demand Gen surfaces penalize it through lower delivery and worse engagement.

Mistaking Demand Gen for PMax

Demand Gen and PMax are not interchangeable. Running Demand Gen with PMax expectations (high direct ROAS in week one, conversion-event optimization with thin upper-funnel signal) leads to disappointment and account abandonment. Demand Gen is upper-funnel : the conversions come, but they come on a different timeline and through a different attribution path. Measure incrementality, not just last-click conversion.

Conclusion : Demand Gen as the upper-funnel default for 2026

Demand Gen has earned the upper-funnel slot in the 2026 Google Ads stack. The campaign type that started as a Discovery replacement now reaches three billion users a month across YouTube, Discover, Gmail, Maps, and GDN, with channel-level controls, AI-expanded lookalikes, Customer Match foundation, product feed dynamic creative, and bid strategies that match the funnel stage. It does what paid social does, on Google’s owned and operated inventory, with first-party signals doing the targeting.

The decision for most accounts in 2026 is not whether to run Demand Gen, it is how to integrate it with Performance Max and Search. Retail runs all three, with PMax doing direct-response, Demand Gen doing demand creation, and Search closing intent. B2B leans Demand Gen plus Search, with PMax sidelined for branded only. Local businesses run Demand Gen with Maps as the primary channel and Search as the support layer. The campaign type that was experimental two years ago is now table stakes.

The work is in execution. Asset volume, creative quality, Customer Match hygiene, channel control discipline, bid strategy timing, and the patience to let the algorithm learn before intervening : those are the levers that separate Demand Gen accounts that scale from those that stall. The platform has matured. The opportunity is in running it well.