Demand Gen

This page is updated every two months with current best practices for Google Ads Demand Gen campaigns. Demand Gen brings social-style visual advertising to Google inventory across YouTube, Discover and Gmail, and rewards strong creative and well-built audience signals. Each update draws on our own experience plus authoritative industry sources and verified real-time research. Bookmark this page and check back for the latest Demand Gen best practices. Each update includes worked examples with the arithmetic shown.

Last updated: 6 August 2026

In This Guide

  1. Executive Summary
  2. Benchmarks & Numbers at a Glance
  3. What Demand Gen Is
  4. Campaign Setup & Goals
  5. Creative & Ad Formats
  6. Audiences & Lookalikes
  7. Bidding Strategies
  8. Placements
  9. Measurement & View-Through
  10. Demand Gen vs PMax vs Video
  11. Common Mistakes to Avoid
  12. What Changed Recently
  13. References

Reading this to build your own campaign? Skip the learning curve — we’ll build the whole thing for you for $129, delivered within 24 hours as a spreadsheet you own. See how it works →

1. Executive Summary: Five Principles for Demand Gen in 2026

Google Ads Demand Gen has matured significantly since its introduction as a replacement for Discovery campaigns. As of August 2026, it is a full-funnel, AI-optimised placement spanning YouTube, Discover, Gmail, and the Google Display Network. The following five principles summarise everything that follows in this article.

  • First-party data is the foundation. Every audience strategy — Lookalike segments, Custom segments, and remarketing — works best when seeded with high-quality first-party lists such as Customer Match, site visitors, and converters. Without strong seeds, the AI has nothing to learn from.[3]
  • Creative diversity drives performance. Google’s AI cannot optimise what it cannot test. Campaigns need at least three unique assets per aspect ratio (square, landscape, vertical), at least one video, and enough headline and description variation to let the system find the best combination per placement and audience.[9]
  • Bidding follows data maturity. Launch with Maximise Conversions, then migrate to target CPA (tCPA) or target ROAS (tROAS) only after accumulating at least 50 conversions. Running a CPA target before that threshold produces erratic performance.[5]
  • View-through conversions are a signal, not a success metric. Demand Gen’s multi-placement reach means many users convert after seeing — but not clicking — an ad. Pair view-through data with Conversion Lift studies, Brand Lift studies, or geo experiments before drawing conclusions about incremental value.[4]
  • Structure enables learning. Separate Lookalike, Custom segment, and remarketing audiences into distinct ad groups with tailored creative. This prevents audience cross-contamination, makes performance readable, and lets you scale what is working without disrupting what is not.[8]

2. Benchmarks and Numbers at a Glance

All figures below are expressed in USD as published; apply an approximate AUD conversion of ×1.55 for budget planning in Australian markets. All vendor claims are from 2026 publications unless noted. Where ranges overlap between sources, the conservative (wider) range is shown.

Metric Typical range or threshold Applies when Source
CTR — Demand Gen overall 0.50%–2.00% Vendor claim; ecommerce accounts, all placements blended [59]
CTR — first-launch / weak performance 0.35%–0.75% Vendor claim; new campaigns in learning phase [60]
CTR — average benchmark 0.75%–1.20% Vendor claim; established campaigns with optimised creative [60]
CTR — optimal target 1.50%–2.50% Vendor claim; high-quality creative with strong audience match [60]
CPC — Demand Gen overall USD $0.30–$1.50 Vendor claim; ecommerce accounts, all placements blended [59]
CPC — first-launch / weak USD $2.00–$2.75 Vendor claim; new campaigns before creative or audience optimisation [60]
CPC — optimal target USD $0.40–$1.20 Vendor claim; mature campaigns with strong Ad Strength [60]
CPM — Demand Gen overall USD $2.80–$5.20 Vendor claim; blended placement average [62]
CPM — YouTube only USD $4.50–$8.00 Vendor claim; YouTube in-stream and in-feed within Demand Gen [62]
CPM — Discover only USD $3.20–$6.50 Vendor claim; Discover feed placements within Demand Gen [62]
CPA — Demand Gen ecommerce USD $15–$60 Vendor claim; ecommerce, wide range by vertical and funnel stage [59]
Video view rate — YouTube/Demand Gen 31.9% Study across thousands of accounts; blended YouTube/Demand Gen video inventory [65]
CPV — YouTube/Demand Gen average USD $0.026 Study across thousands of accounts; blended YouTube/Demand Gen video inventory [65]
CTR — YouTube/Video overall average 0.65% Study; YouTube/video campaign average, aligns with Demand Gen video inventory [73]
Minimum conversions before switching to tCPA 50 conversions Google Help Centre guidance; applies to all Demand Gen campaigns moving from Maximise Conversions [5]
Minimum daily budget for tCPA campaigns 10× target CPA per day Google Demand Gen performance guide; required for Smart Bidding learning headroom [7]
Minimum seed list size for Lookalike segments 100 users Vendor claim (Google Ads Help); minimum to generate a similar segment [12]
View-through conversion window — default 1 day Vendor claim (Google Ads Help); default setting for Demand Gen view-through attribution [66]
Recommended images per asset group 4 images Vendor claim (Google Ads Help); recommended minimum per Demand Gen asset group [12]
Recommended YouTube videos per asset group 5 videos Vendor claim (Google Ads Help); recommended minimum per Demand Gen asset group [12]

3. What Demand Gen Is and When to Use It

Demand Gen is Google’s AI-powered campaign type designed to reach audiences who are not yet actively searching — people browsing YouTube, reading the Discover feed, or checking Gmail. Unlike Search, which captures declared intent, Demand Gen creates intent by surfacing relevant creative in high-attention, visually rich environments. It replaced Discovery campaigns in 2024 and has been expanded significantly through a series of “Demand Gen Drops” culminating in the June 2026 release.[4]

Demand Gen sits in the middle and upper funnel. Its primary role is to move audiences from unawareness or passive interest toward consideration and first interaction. It is not a replacement for Search when demand already exists, and it is not a replacement for Performance Max when conversion efficiency across all Google inventory is the sole objective. It is, however, the best Google-native tool for social-style prospecting, lookalike expansion, and creative storytelling at scale.[5]

When Demand Gen is the right choice

  • You need to reach people who do not yet know they want your product — classic demand creation, not demand capture.
  • You have strong creative assets (images and video) and want control over where they serve, including placement-level channel controls across YouTube, Discover, Gmail, and the Google Display Network.[7]
  • You want to run Lookalike-style prospecting using first-party data without committing to the full black-box nature of Performance Max.[3]
  • You are running an upper-funnel brand campaign alongside a lower-funnel Search or Shopping campaign and need a clearly separated prospecting layer.
  • You want to test creative in a controlled experiment environment using Demand Gen experiments before rolling assets into Performance Max.[4]

When Demand Gen is not the right choice

  • Your sole objective is maximising conversion volume across all Google inventory — Performance Max is better suited.
  • You have no video or high-quality image assets — the campaign will underperform without the creative diversity the AI needs.
  • Your monthly budget is below the minimum needed for the learning phase (see Section 7 for the 10× CPA rule).

Worked example

Choosing Demand Gen over Performance Max for a new product launch

  • Setup: A Melbourne homewares retailer spending $8,000 AUD per month on Google Ads wants to launch a new furniture range in September 2026. They have an existing Search campaign capturing branded and category queries. They have four lifestyle video assets (16:9, 60 seconds each) and twelve product images across three aspect ratios. Their historical CPA for furniture is $45 USD (~$70 AUD).
  • Numbers: $8,000 AUD/month ÷ 30 days = $267 AUD/day. At a $70 AUD CPA target, the 10× rule requires $700 AUD/day minimum — the total budget is insufficient for a tCPA Demand Gen campaign at launch. They instead allocate $4,500 AUD/month to Demand Gen on Maximise Conversions, targeting 64 clicks/day at a $1.00 USD CPC (~$1.55 AUD CPC) blended benchmark.[60] The remaining $3,500 AUD stays in Search.
  • Decision: Launch Demand Gen with Maximise Conversions bidding, no tCPA target set, prospecting-only audiences (Lookalike from top 500 purchasers, Custom segment from competitor research terms), and channel controls limiting YouTube and Discover only (Gmail excluded for image-led furniture creative).
  • Why: Google’s guidance states that campaigns without 50 conversions of history should start with Maximise Conversions rather than tCPA, and the 10× CPA budget floor is not met at launch.[5][7]

4. Campaign Setup and Goals

A well-structured Demand Gen campaign from the outset saves weeks of remediation later. Google’s February 2026 best-practices update identified four setup priorities: optimised targeting, Lookalike audiences, new customer acquisition goals, and robust data infrastructure.[1]

Conversion goal configuration

Set your conversion goal before you set your bid strategy. For most accounts, the primary conversion action should be the same action used in Search and Shopping campaigns — purchase, lead form submission, or booking — so that the AI is optimising toward the same downstream value. If you are running Demand Gen for upper-funnel purposes, configure a micro-conversion (video view, page engagement, scroll depth) as a secondary action and measure brand lift separately.[3]

New Customer Acquisition goal

If growth from new users is the primary objective, enable the New Customer Acquisition (NCA) goal and upload your existing customer list for exclusion in prospecting campaigns. This prevents the algorithm from burning prospecting budget on users who have already converted. Google explicitly recommends this configuration in its February 2026 guidance.[1]

Data infrastructure

Before launching, confirm that sitewide tagging is implemented via the Google tag gateway, that conversion actions are firing correctly, and that any offline conversion data (CRM leads, in-store purchases) is connected through Google’s Data Manager.[1] Demand Gen’s AI optimisation is only as good as the conversion signals it receives — poor tagging is the single most common cause of learning phase failure.

Campaign structure

Google recommends separating audiences with different strategies or creative angles into distinct ad groups.[8] A practical three-ad-group structure is:

  • Ad group 1 — Prospecting (Lookalike): Narrow Lookalike built from your highest-value converters. Creative should focus on awareness and brand story.
  • Ad group 2 — Intent (Custom segment): Custom segment built from high-intent search terms and competitor/solution research behaviour. Creative should address the consideration stage.
  • Ad group 3 — Remarketing: First-party site visitors and Customer Match warm re-engagement. Creative should focus on conversion-oriented messaging and offers.

Worked example

Configuring a New Customer Acquisition goal with existing customer exclusion

  • Setup: A Brisbane software-as-a-service (SaaS) account spending $12,000 AUD/month on Google Ads. Their CRM contains 3,200 active customers. Their historical Demand Gen CPA for trial sign-ups is $38 USD (~$59 AUD). They want September 2026 Demand Gen spend to target only net-new prospects.
  • Numbers: 3,200 customers uploaded as a Customer Match exclusion list. Seed list for Lookalike: top 800 trial-to-paid converters (25% of the customer base). 800 users exceeds the 100-user minimum for Lookalike generation.[12] Budget allocated to prospecting ad groups only: $9,000 AUD/month ($300/day). At a $59 AUD CPA target, the 10× rule requires $590 AUD/day — budget is below threshold, so Maximise Conversions is used at launch, not tCPA.
  • Decision: Enable New Customer Acquisition goal; upload 3,200-user exclusion list; set Lookalike seed from 800 top converters; launch on Maximise Conversions. Review after 50 conversions to switch to tCPA at $59 AUD.
  • Why: Excluding existing customers from prospecting campaigns prevents budget waste on already-converted users, as recommended in Google’s February 2026 guidance.[1]

Performance Planner

Use Performance Planner before setting budgets, particularly if you are running Demand Gen alongside Search or Performance Max. The tool models expected conversion volume at different spend levels and highlights the point of diminishing returns — critical for justifying budget allocation to stakeholders.[1]

5. Creative and Ad Formats

Creative is the primary performance lever in Demand Gen. Because the campaign serves across visually diverse environments — YouTube pre-roll, Shorts, the Discover feed, Gmail promotions — no single creative executes well everywhere. Google’s guidance is explicit: use a diverse set of assets rather than a single creative angle, and target “Excellent” Ad Strength.[1][9]

Required asset formats

Asset type Recommended formats Primary placement Minimum recommended count
Image — landscape 1.91:1 Discover feed, Gmail At least 1 per ad group
Image — square 1:1 Discover feed, YouTube in-feed At least 1 per ad group
Image — portrait/vertical 4:5 or 9:16 YouTube Shorts At least 1 per ad group
Video — horizontal 16:9 YouTube in-stream, in-feed Included in 5-video target
Video — square 1:1 Discover, YouTube in-feed Included in 5-video target
Video — vertical 9:16 YouTube Shorts Included in 5-video target

Google recommends at least three unique assets per aspect ratio and a minimum of five YouTube videos per asset group.[12] The June 2026 Drop introduced broader video aspect-ratio transformations — vertical to square, vertical to landscape, and square to landscape — reducing the production burden when only one video format is available.[4]

Ad Strength and ABCD creative guidance

Ad Strength in Demand Gen reflects asset diversity, not a single ad’s quality. To reach “Excellent,” you typically need all recommended asset counts filled with genuinely different creative angles — not slight variations of the same image with a different colour filter. For video, follow Google’s ABCD framework: Attention in the first five seconds, Branding visible early, Connection with the audience’s problem or aspiration, and Direction — a clear CTA repeated in both voiceover and on-screen overlay.[9]

Image quality standards

Google’s Demand Gen help page specifically recommends brand and inspirational imagery with minimal text and effective lighting.[12] In practice, this means avoiding assets repurposed from static display banners with heavy text overlays — these consistently underperform in the Discover and YouTube in-feed environments, which reward editorial-quality photography and video.

Gemini-powered creative insights

The June 2026 Drop added Gemini-powered creative insights that surface recommendations for optimising image and video assets specifically for YouTube within Demand Gen.[4] This tool is now available in the interface and should be reviewed after the first 14 days of a campaign’s life, once it has enough impression data to generate meaningful recommendations.

Worked example

Building a Demand Gen asset set from a limited creative library

  • Setup: A Sydney fitness equipment retailer has one 30-second brand video (16:9 horizontal) and six product photos (all landscape). They want to launch a Demand Gen campaign targeting home gym buyers in October 2026 with a $5,500 AUD/month budget.
  • Numbers: Current asset set: 1 video format, 1 image aspect ratio = Ad Strength will not reach “Excellent.” Google recommends 5 videos and 4 images across at least 3 aspect ratios.[12] Gap: 4 additional video formats, 3 additional image formats. Using the June 2026 video aspect-ratio transformation tool, the single 16:9 video is transformed into a 1:1 square version and a 9:16 vertical version — adding 2 formats at zero production cost.[4] Remaining gap: 2 additional video concepts needed. For images: 6 existing landscape images are cropped to 1:1 (square) and 4:5 (portrait) using Google’s asset cropping tool — adding 12 additional image assets.
  • Decision: Use the aspect-ratio transformation tool to generate square and vertical video from the existing 16:9 asset. Crop existing landscape images to 1:1 and 4:5. Commission 2 additional 15-second video concepts (lifestyle and testimonial) to reach 5 distinct video assets. Upload all assets targeting “Excellent” Ad Strength before launch.
  • Why: Google’s guidance requires at least 3 unique assets per aspect ratio and 5 videos per asset group to reach “Excellent” Ad Strength, which is the threshold for full AI optimisation capability.[9][12]

6. Audiences and Lookalike Segments

Audience strategy is where Demand Gen separates from generic display advertising. The campaign type’s AI can only work with the audience inputs it is given — vague or broad targeting produces vague and expensive results. The recommended 2026 structure uses three distinct audience layers, each in its own ad group, built outward from first-party data.[3][8]

Lookalike segments

Lookalike segments in Demand Gen are built from first-party seed audiences: Customer Match lists, site visitors, app users, or conversion-based lists. The minimum seed size is 100 users, though in practice larger seeds (500 users or more) produce more stable lookalike models.[12] Start with a narrow expansion setting before widening, so the early conversion data reflects a tighter audience match and gives the bidding algorithm cleaner signals to learn from.[4][9]

Seed your Lookalikes with your highest-quality converters — not all site visitors. A Lookalike built from people who purchased above a $200 AUD threshold will behave very differently from one built from all site visitors who bounced after ten seconds. Multiple 2026 practitioner guides consistently make this point.[4][11]

Custom segments

Custom segments let you reach people based on recent search behaviour and the websites or apps they have interacted with. This makes them the closest intent-based targeting option available inside Demand Gen, filling the gap between Lookalikes (similarity-based) and Search (declared intent).[9][11] Use Custom segments to capture active researchers — people searching competitor brand names, category terms, or solution-category queries — and keep them in a dedicated ad group so their performance is readable independently of your Lookalike and remarketing groups.[6][9]

First-party remarketing

Customer Match and site visitor remarketing lists form the warm re-engagement layer. These audiences already know your brand and typically convert at a lower CPA than cold prospecting. Keep them in a separate ad group with creative that assumes brand familiarity — promotional offers, product-specific messaging, and urgency-based CTAs — rather than the brand-story creative used for prospecting.[4][9]

Optimised targeting

Google’s February 2026 guidance explicitly lists optimised targeting as a best practice for conversion-focused Demand Gen campaigns.[3] Some practitioners recommend disabling it for tighter audience control, but this conflicts with Google’s current recommendation. The conservative approach, given the conflicting advice, is to leave optimised targeting enabled but monitor impression share by audience segment weekly to detect any unwanted audience drift.[3][16][17]

Worked example

Three-ad-group audience structure for a B2C skincare brand

  • Setup: A Perth skincare brand spending $7,200 AUD/month on Demand Gen. Their Customer Match list contains 2,100 past purchasers. Their site receives 18,000 unique visitors per month. Their Demand Gen campaign targets women aged 28–50 interested in skincare, running November–December 2026.
  • Numbers: Ad group 1 (Lookalike): seed = top 600 purchasers who spent more than $80 AUD (28% of 2,100 list). 600 users well exceeds the 100-user minimum.[12] Narrow expansion setting. Budget: $3,000 AUD/month. Ad group 2 (Custom segment): search terms including competitor brand names + “natural skincare” + “anti-ageing serum” — estimated audience 45,000 users in-market. Budget: $2,400 AUD/month. Ad group 3 (Remarketing): 18,000 site visitors (last 30 days) minus 2,100 existing customers = ~15,900 net remarketing pool. Budget: $1,800 AUD/month. Total: $7,200 AUD/month.
  • Decision: Create three separate ad groups with the allocations above. Ad group 1 uses awareness/brand creative (16:9 video + 1.91:1 images). Ad group 2 uses consideration creative (comparison messaging, ingredient callouts). Ad group 3 uses conversion creative ($15 AUD off first reorder, free shipping).
  • Why: Google’s 2026 guidance specifies separating each distinct audience strategy into its own ad group to prevent cross-contamination and enable clean performance comparison.[8]

Worked example

Diagnosing a weak Lookalike segment from a low-quality seed

  • Setup: A Gold Coast travel agency running Demand Gen with a Lookalike built from 14,000 website visitors (all visitors, no quality filter). After 21 days, the campaign has a CTR of 0.38% against an average benchmark of 0.75%–1.20%[60] and a CPC of $2.40 USD against the average benchmark of $1.40–$2.00 USD.[60] Both metrics are in the “first-launch/weak” range despite the campaign being three weeks old.
  • Numbers: CTR 0.38% vs. 0.75% benchmark = 49% below average.[60] CPC $2.40 USD vs. $2.00 USD upper benchmark = 20% above weak-performance ceiling.[60] Lookalike seed: 14,000 all-visitors vs. recommended high-quality converters. Converters in the account: 420 past bookers. 420 users exceeds the 100-user minimum.[12]
  • Decision: Rebuild the Lookalike seed from the 420 confirmed bookers only. Pause the existing all-visitor Lookalike ad group. Relaunch with narrow expansion setting. Review CTR and CPC after 14 days.
  • Why: Seeding a Lookalike with low-intent all-visitors dilutes the model; the 2026 best practice is to use only highest-value converters as the seed source.[4][11]

7. Bidding Strategies

Demand Gen supports four bid strategies: Maximise Clicks, Maximise Conversions, target CPA (tCPA), and target ROAS (tROAS). The correct strategy depends on where the campaign sits in its data lifecycle, not on what the advertiser wishes to achieve. Applying tCPA to a campaign with 12 conversions is a common and costly mistake — the algorithm lacks sufficient data to bid efficiently and will either under-deliver or overspend to force volume.[5]

Bidding strategy progression

Stage Recommended strategy When to move to next stage Key condition
Launch (0–50 conversions) Maximise Conversions After 50 conversions recorded in the campaign No tCPA target set; let the algorithm learn[5]
Efficiency (50+ conversions) tCPA After tCPA is stable for 2–3 weeks; conversion values are known Set tCPA at or slightly above observed CPA to avoid under-delivery[5]
Value optimisation (stable tCPA) tROAS When conversion values are reliable and consistent Requires accurate conversion value tracking; not suitable for lead-only goals[1]
Reach/awareness Maximise Clicks or CPM-based N/A — use when upper-funnel reach, not conversion volume, is the KPI Pair with Brand Lift measurement; do not evaluate on CPA[4]

The 10× budget rule

Google’s Demand Gen performance guide states that the daily budget should be at least 10 times the target CPA when running tCPA bidding.[7] This is not a recommendation — it is a functional threshold. Below this level, the Smart Bidding algorithm does not have enough budget headroom to explore the auction landscape, resulting in erratic CPAs and frequent under-delivery.

Worked example

Validating budget adequacy before switching from Maximise Conversions to tCPA

  • Setup: An Adelaide B2B software account running Demand Gen since 1 August 2026. By 28 August 2026 the campaign has recorded 53 conversions (demo requests) at an average CPA of $47 USD (~$73 AUD). The campaign’s daily budget is $200 AUD ($129 USD).
  • Numbers: 53 conversions exceeds the 50-conversion threshold for switching to tCPA.[5] 10× tCPA rule: $47 USD × 10 = $470 USD/day (~$729 AUD/day) required. Current daily budget: $200 AUD ($129 USD). $200 AUD is 27% of the $729 AUD required minimum — budget is critically insufficient for tCPA at this target.
  • Decision: Do not switch to tCPA at $47 USD. Either (a) increase budget to $729 AUD/day before switching, or (b) set tCPA at $12 USD (USD $12 × 10 = $120 USD/day, within the $129 USD available) and accept a lower-volume, lower-CPA efficiency target. Option (b) is selected: tCPA set at $18 AUD ($12 USD), budget maintained at $200 AUD/day, with a planned budget review after 30 days.
  • Why: Google’s guidance requires a daily budget of at least 10× the tCPA target; operating below this threshold causes Smart Bidding to under-deliver or overshoot the CPA target unpredictably.[7]

8. Placements: YouTube, Discover, and Gmail

Demand Gen serves across four surfaces: YouTube (in-stream, in-feed, and Shorts), the Discover feed, Gmail (Promotions and Social tabs), and the Google Display Network — the last of these was added progressively through 2025 and was confirmed in Google’s December 2025 Drop.[7] Each surface has different creative requirements, audience behaviour, and performance benchmarks.

Placement-level benchmarks

Placement CPM (USD) CPC (USD) CTR Creative priority
YouTube (all formats) $4.50–$8.00 $0.55–$1.20 1.2%–2.1% Video-first; 16:9 + 9:16 Shorts
Discover feed $3.20–$6.50 $0.45–$0.95 1.5%–2.8% High-quality editorial images; 1.91:1 and 1:1
Gmail Part of blended $2.80–$5.20 Part of blended $0.30–$1.50 Part of blended 0.5%–2.0% Subject-line equivalent headline; image secondary
Blended (all placements) $2.80–$5.20 $0.30–$1.50 0.5%–2.0% Diverse asset set across all formats

Sources for table: [59][62]

Channel controls

Channel controls allow advertisers to restrict which surfaces a Demand Gen campaign serves on. This is particularly useful when creative is only available in one format (for example, only video assets) or when a specific placement is consistently delivering below-benchmark CPAs that distort overall performance reporting.[1][7]

In practice, the most common channel control decisions are:

  • Exclude Gmail when the campaign is video-only and there are no high-quality image assets — Gmail renders primarily as an image-led email-style unit.
  • Exclude YouTube Shorts if no 9:16 vertical video assets are available, because horizontal video served in Shorts format produces poor viewability.
  • Limit to YouTube + Discover for brand awareness campaigns where Gmail’s lower-attention environment is less relevant to the awareness objective.

YouTube Shorts

Google’s February 2026 guidance specifically calls out vertical images and video for YouTube Shorts.[2] Shorts placements require 9:16 vertical assets; the June 2026 video aspect-ratio transformation tool now automatically generates 9:16 from existing 16:9 or 1:1 assets, but the output should be reviewed manually — automated cropping of wide-format videos can cut important visual elements.[4]

Worked example

Using channel controls to exclude Gmail when running a video-only Demand Gen campaign

  • Setup: A Canberra financial planning firm running a Demand Gen campaign in October 2026 with a $6,000 AUD/month budget. Their asset set consists of three 16:9 videos (30 seconds each) and two 9:16 vertical videos. They have no image assets. The campaign has been live for 10 days and Gmail is receiving 22% of impressions but generating a CTR of 0.28% against the blended benchmark of 0.50%–2.00%.[59]
  • Numbers: Gmail share: 22% of total impressions = 0.22 × estimated 180,000 impressions/month = 39,600 Gmail impressions. Gmail CTR: 0.28% vs. 0.50% floor = 44% below the low end of the acceptable range.[59] Estimated Gmail clicks: 39,600 × 0.0028 = 111 clicks. At blended CPC of $0.85 AUD, Gmail spend = ~$94 AUD/month for 111 clicks — low-quality traffic at higher proportional cost.
  • Decision: Apply channel control to exclude Gmail. Reallocate 22% of budget (~$1,320 AUD/month) to YouTube and Discover only. Revisit Gmail exclusion when image assets are produced.
  • Why: Gmail renders primarily as an image-led format; video-only asset sets underperform there, and channel controls are specifically provided to address this mismatch.[1][7]

9. Measurement and View-Through Attribution

Measurement is the area where Demand Gen is most frequently misunderstood. Because the campaign reaches users across multiple surfaces in a visual, non-search context, many conversions occur after an ad impression rather than after a click. This produces view-through conversions — a signal that requires careful interpretation, not uncritical acceptance.[4][11]

View-through conversions

Google Ads records a view-through conversion when a user sees (but does not click) a Demand Gen ad and then converts within the view-through attribution window. The default view-through window is one day.[66] View-through conversions are not the same as incremental conversions — some portion of those users would have converted anyway through other channels. Treating view-through counts as equivalent to click-through conversions in CPA calculations will systematically overstate Demand Gen’s contribution.[4][11]

Incrementality measurement

The correct approach to measuring Demand Gen’s true value is incrementality testing. Google provides three native incrementality tools:[4]

  • Conversion Lift Study: Compares conversion rates between an exposed group and a holdout group that did not see Demand Gen ads. Provides the cleanest read on incremental conversions.
  • Brand Lift Study: Surveys exposed and non-exposed users on brand awareness, consideration, and purchase intent. Appropriate for upper-funnel campaigns where conversion is not the immediate objective.
  • Geo lift testing: Compares matched geographic regions with and without Demand Gen spend over a defined test period. Useful when the campaign is too small for a statistically significant in-platform lift study.

Reporting columns and social parity

Google added new reporting columns that include view-through conversions to make Demand Gen reporting comparable to paid social (Meta, TikTok) reporting conventions.[1] The June 2026 Drop also introduced Web to App Acquisition Measurement for app-install attribution, which connects Demand Gen impressions to downstream app installs across surfaces.[4]

Worked example

Distinguishing view-through volume from incremental conversions using a geo lift test

  • Setup: A Melbourne subscription box brand running Demand Gen from 1–31 October 2026 with a $10,000 AUD/month budget. In-platform reporting shows 320 view-through conversions and 85 click-through conversions in the month. The account manager wants to understand how many of the 320 view-through conversions are incremental.
  • Numbers: Total reported conversions: 320 view-through + 85 click-through = 405. A geo lift test is designed using two matched metro regions: Melbourne (Demand Gen active, $10,000 AUD/month) and Brisbane (Demand Gen paused, $0 spend). Test period: 1–31 October 2026. Baseline Brisbane conversion rate from Search + organic: 1.8 conversions per 1,000 sessions. Melbourne conversion rate during test: 2.4 per 1,000 sessions. Incremental lift: (2.4 − 1.8) ÷ 1.8 = 33% incremental lift attributable to Demand Gen. Applied to 320 view-throughs: 320 × 0.33 = ~106 incremental view-through conversions. Actual incremental total: 106 + 85 = 191 conversions vs. 405 reported — a 53% overstatement in raw reporting.
  • Decision: Report 191 as the conservative incremental conversion count for Demand Gen in October 2026. Adjust CPA calculation: $10,000 AUD ÷ 191 = $52.36 AUD incremental CPA, not the $24.69 AUD implied by raw view-through counting.
  • Why: View-through conversions overstate impact when used alone; 2026 best practice requires pairing them with incrementality testing before reporting CPA to stakeholders.[4][11]

10. Demand Gen vs Performance Max vs Video Campaigns

Senior marketers frequently ask how Demand Gen relates to Performance Max (PMax) and standalone Video campaigns. The three campaign types are not interchangeable — each serves a distinct strategic function and operates with different levels of advertiser control.[5]

Structural comparison

Attribute Demand Gen Performance Max Video campaigns
Primary funnel stage Upper to mid funnel; awareness and consideration Full funnel; conversion maximisation Upper funnel; awareness and reach
Placement control Channel controls across YouTube, Discover, Gmail, GDN Minimal; Google allocates across all inventory YouTube only; format-level control
Audience control High; Lookalike, Custom, remarketing, NCA goal Low; audience signals only, not targeting High; full YouTube audience options
Creative control High; asset-group structure with per-format assets Medium; asset groups but Google assembles final ads High; specific video formats selected by advertiser
Bid strategies available Max Clicks, Max Conversions, tCPA, tROAS Max Conversions, Max Conversion Value, tCPA, tROAS CPV, CPM, Max Conversions, tCPA
Avg CPC benchmark (USD) $0.30–$1.50[59] Blended; closer to Search at $2.96 for search inventory[73] $0.18 avg YouTube/Video[73]
Avg CTR benchmark 0.5%–2.0%[59] Varies by inventory mix; Search component at 3.52%[73] 0.65% avg YouTube/Video[73]
Incrementality measurement Conversion Lift, Brand Lift, geo tests Conversion Lift; limited placement-level visibility Brand Lift, Reach and Frequency reporting

When to run each campaign type

The most effective accounts in 2026 run Demand Gen and Performance Max in parallel, not as substitutes. Demand Gen handles the prospecting and consideration layers — building audiences and warming interest — while PMax handles conversion harvesting across all inventory. Standalone Video campaigns are appropriate when the objective is pure brand awareness at scale with CPM or CPV bidding, and creative control at the video-format level is required.[5]

Worked example

Allocating budget between Demand Gen and Performance Max for a mid-market retailer

  • Setup: A Sydney outdoor furniture retailer with a $25,000 AUD/month total Google Ads budget. They run Search, Shopping, Performance Max, and are considering adding Demand Gen for the November–December 2026 gifting season. Their current PMax CPA is $38 AUD and their Search CPA is $29 AUD. They want to grow new customer acquisition by 30% over the period.
  • Numbers: Current monthly split: Search $8,000 AUD, Shopping/PMax $14,000 AUD, remaining $3,000 AUD unallocated. PMax average CTR: blended, estimated from $0.44 USD Display CPC benchmark[73] and $2.96 USD Search CPC[73] — PMax is conversion-focused, not prospecting. Demand Gen prospecting CPC benchmark: $0.75–$1.50 AUD (converted from USD $0.30–$1.50 × 1.55 AUD/USD = $0.47–$2.33 AUD, mid-point ~$1.40 AUD).[59] Allocate $6,000 AUD/month to Demand Gen (drawn from PMax reduction of $3,000 AUD + unallocated $3,000 AUD). Demand Gen budget: $200 AUD/day. At $38 AUD target CPA, 10× rule requires $380 AUD/day — insufficient for tCPA, so launch on Maximise Conversions.[7]
  • Decision: Launch Demand Gen at $6,000 AUD/month on Maximise Conversions targeting new customers only (NCA goal enabled). Reduce PMax to $11,000 AUD/month. Maintain Search at $8,000 AUD/month. Evaluate after 50 Demand Gen conversions before setting tCPA.
  • Why: Demand Gen and PMax serve distinct funnel roles; running them in parallel with NCA goal on Demand Gen prevents audience overlap and maximises new customer reach without cannibalising PMax’s conversion harvesting role.[1][5]

11. Common Mistakes to Avoid

The following mistakes are the most frequently documented in 2026 practitioner resources and Google’s own guidance. Each is specific and actionable — the goal is not a generic caution but an identifiable failure mode you can audit for in any active Demand Gen account.[30]

1. Setting tCPA before 50 conversions

Switching to tCPA with fewer than 50 conversions forces Smart Bidding to operate without enough data. The result is typically erratic CPAs — either severe under-delivery (the algorithm bids too conservatively) or CPA spikes (it bids aggressively to force volume). Always begin with Maximise Conversions and monitor for the 50-conversion threshold before switching.[5]

2. Running a single creative angle

Uploading one hero image and one video across all ad groups eliminates the AI’s ability to optimise. Google explicitly states that creative diversity is required for full optimisation capability.[9] A campaign with three images in one aspect ratio and one video will produce a lower Ad Strength score and lose auction eligibility for format-specific placements like Shorts.

3. Seeding Lookalikes with all-site-visitor lists

A Lookalike built from all site visitors will include high-bounce, zero-intent traffic that dilutes the model. Seed only from converters or high-value engagement cohorts.[4][11]

4. Mixing audience strategies in a single ad group

Combining a cold Lookalike and a warm remarketing audience in the same ad group makes it impossible to read which audience is driving performance or apply different creative angles. Separate every distinct audience strategy into its own ad group.[8]

5. Treating view-through conversions as equivalent to click-through conversions

Raw view-through counts significantly overstate Demand Gen’s contribution in most accounts. The default view-through window is one day,[66] but many of those conversions are attributable to other channels. Always run an incrementality test before reporting view-through numbers to stakeholders.[4]

6. Running below the 10× CPA daily budget floor on tCPA

A tCPA campaign running at 3× or 4× the daily CPA target will under-deliver consistently. This is not an optimisation problem — it is a budget constraint. Resolve it by increasing budget, reducing the tCPA target, or reverting to Maximise Conversions until budget increases are approved.[7]

7. Ignoring placement-level creative requirements

Serving horizontal 16:9 video in YouTube Shorts, or text-heavy display banners in the Discover feed, consistently underperforms. Use channel controls to exclude placements where you do not have format-appropriate assets, rather than serving mismatched creative to all surfaces.[1][7]

8. Not excluding existing customers from prospecting campaigns

Without an explicit Customer Match exclusion list, Demand Gen will serve prospecting creative to people who have already purchased. This wastes prospecting budget and inflates the apparent conversion rate of the campaign with easy re-converts.[1][4]

Worked example

Diagnosing and correcting a tCPA set below the 10× budget floor

  • Setup: A Hobart e-commerce account running Demand Gen since 15 September 2026, with tCPA set at $55 AUD ($35 USD) and a daily budget of $120 AUD ($77 USD). After 14 days, the campaign has delivered only 6 conversions and is spending an average of $91 AUD/day — 24% under the $120 AUD budget cap.
  • Numbers: 10× tCPA rule: $55 AUD × 10 = $550 AUD/day required.[7] Current budget: $120 AUD/day = 22% of the required minimum. Conversions in 14 days: 6 — well below the 50-conversion threshold for tCPA operation.[5] Two violations identified: (1) tCPA set before 50 conversions; (2) budget is 78% below the 10× floor.
  • Decision: Remove tCPA target; switch to Maximise Conversions. Maintain $120 AUD/day budget. Set a calendar reminder to review at 50 conversions. When switching back to tCPA, set tCPA at $12 AUD ($7.74 USD) — the maximum tCPA that the $120 AUD budget can support under the 10× rule ($120 AUD ÷ 10 = $12 AUD).
  • Why: Both the 50-conversion minimum and the 10× budget floor must be satisfied before tCPA bidding can operate effectively; violating either condition causes consistent under-delivery.[5][7]

12. What Changed Recently: June–August 2026

Google has maintained a monthly “Demand Gen Drop” release cadence through 2026. The most recent confirmed updates, from the June 2026 Drop, introduced three significant changes.[4] Earlier 2026 updates that remain part of the current product are also noted for completeness.

June 2026 Drop — confirmed new features

  • Broader video aspect-ratio transformations: Demand Gen now supports automated conversion of video assets across vertical → square, vertical → landscape, and square → landscape ratios. This reduces the production cost of covering all placement formats and is particularly useful for accounts with limited video production budgets.[4]
  • Gemini-powered creative insights: A new in-platform tool that analyses image and video assets and recommends specific optimisations for YouTube placements within Demand Gen. Recommendations include pacing, branding timing, CTA visibility, and aspect-ratio suitability.[4]
  • Web to App Acquisition Measurement: Advertisers with app presence can now attribute Demand Gen exposure to downstream app installs, providing cross-surface conversion visibility that was previously unavailable for app-install journeys originating from Demand Gen impressions.[4]

Earlier 2026 updates still in effect

  • Channel controls across all surfaces (February 2026): Advertisers can now specify which of YouTube, Discover, Gmail, and the Google Display Network their Demand Gen campaign serves on. This replaced the previous all-or-nothing placement model and is now a standard setup step.[1]
  • 9:16 vertical image ads on YouTube Shorts (February 2026): Vertical image assets are now eligible to serve as Shorts ads within Demand Gen, not just vertical video — significantly broadening Shorts eligibility for accounts without vertical video production.[1][2]
  • Video enhancements and Veo-based video variations (March 2026): Google’s Veo AI model can generate additional versions of an original video asset, creating format or pacing variations without requiring additional production. Asset Uplift A/B Experiments were also introduced to test AI-generated asset variations against originals in a controlled environment.[51]
  • View-through conversion reporting columns (February 2026): New reporting columns that include view-through conversions were added to Demand Gen campaign reporting to bring it into parity with paid social reporting conventions — useful for cross-channel comparison but requiring the incrementality caution described in Section 9.[1]

Strategic implication of the June 2026 direction

Reading across the recent update cadence, Google is clearly investing in three directions simultaneously: more advertiser control over where ads serve (channel controls, format-level controls), more AI assistance for creative production (Gemini insights, Veo variations, aspect-ratio transformations), and better cross-channel measurement (Web to App, view-through columns, lift studies).[1][4] Accounts that adapt their workflows to use these tools — particularly Gemini creative insights and asset-ratio transformation — will reduce production cost while maintaining the creative diversity that the AI requires to optimise effectively.

Worked example

Using Gemini-powered creative insights to prioritise video asset improvements

  • Setup: A Newcastle fashion retailer running a Demand Gen campaign from 1 July 2026 with three 16:9 videos (15 seconds, 30 seconds, and 60 seconds) and eight images across three aspect ratios. After 21 days, the 60-second video has a view rate of 18.4% — below the 31.9% average benchmark.[65] The 15-second and 30-second videos are at 34.2% and 29.7% respectively.
  • Numbers: 60-second video view rate: 18.4% vs. 31.9% benchmark = 42% below average.[65] Gemini creative insights (available June 2026+) flags the 60-second video for: (1) brand logo not visible until 22 seconds; (2) CTA appears only once in the final 5 seconds; (3) no voiceover CTA. 15-second video: above benchmark at 34.2%, 7% above average. 30-second video: within 7% of benchmark.
  • Decision: Pause the 60-second video. Use the Gemini recommendations as a brief for an edited version: move logo appearance to seconds 0–3, add CTA overlay at 5 seconds and 25 seconds, record voiceover CTA. Use the 15-second video as the primary asset while the edit is completed. Retest the edited 60-second version using a Demand Gen Asset Uplift A/B Experiment.[51]
  • Why: Gemini-powered creative insights provide specific, in-platform recommendations tied to YouTube performance benchmarks; acting on them is faster and more precise than manual creative review.[4]

References

  1. [1] https://blog.google/products/ads-commerce/demand-gen-drop-february-2026/ blog.google
  2. [2] https://almcorp.com/blog/google-demand-gen-best-practices-february-2026/ almcorp.com
  3. [3] https://www.youtube.com/watch?v=I-B7y-WgfDE www.youtube.com
  4. [4] https://www.youtube.com/watch?v=BLn0q16khu0 www.youtube.com
  5. [5] https://www.lunio.ai/blog/google-demand-gen-strategy-guide www.lunio.ai
  6. [6] https://www.youtube.com/watch?v=CkF-Hh_6yPE www.youtube.com
  7. [7] https://blog.google/products/ads-commerce/demand-gen-drop-december-2025/ blog.google
  8. [8] https://www.groas.com/post/google-demand-gen-campaigns-complete-2026-guide www.groas.com
  9. [9] https://support.google.com/google-ads/answer/14693848?hl=en-SE support.google.com
  10. [10] https://prospeo.io/s/demand-gen-best-practices prospeo.io
  11. [11] https://www.youtube.com/watch?v=-7CdMDUh3bI www.youtube.com
  12. [12] https://support.google.com/google-ads/answer/13695777?hl=en support.google.com
  13. [13] https://www.storegrowers.com/google-demand-gen-campaigns/ www.storegrowers.com
  14. [14] https://twominutereports.com/blog/google-ads-best-practices twominutereports.com
  15. [15] https://spideraf.com/articles/setting-up-and-optimizing-google-demand-gen-campaigns spideraf.com
  16. [16] https://www.reddit.com/r/googleads/comments/1qaoyec/its_2026_whats_one_piece_of_google_ads… www.reddit.com
  17. [17] https://support.google.com/google-ads/thread/413560562/google-ads-best-practices-for-2026?… support.google.com
  18. [18] https://www.linkedin.com/posts/menachemani_google-demand-gen-campaigns-when-to-use-activit… www.linkedin.com
  19. [19] https://www.groas.com/post/google-demand-gen-campaigns-2026-complete-guide-setup-creative-… www.groas.com
  20. [20] https://blog.google/products-and-platforms/products/youtube/youtube-demand-gen-updates/ blog.google
  21. [21] https://www.landbase.com/blog/demand-generation-2026-what-changed-what-works www.landbase.com
  22. [22] https://www.definedigitalacademy.com/blog/demand-gen-set-up-in-2026-a-step-by-step-guide www.definedigitalacademy.com
  23. [23] https://steerads.com/en/blog/google-ads-demand-gen-campaigns-2026 steerads.com
  24. [24] https://support.google.com/google-ads/answer/15594567?hl=en support.google.com
  25. [25] https://www.storegrowers.com/demand-gen-audiences/ www.storegrowers.com
  26. [26] https://www.linkedin.com/posts/tauqeerjamil_google-just-shared-key-demand-gen-tips-activit… www.linkedin.com
  27. [27] https://www.dotidot.io/post/demand-gen-campaigns-specs-and-best-practices www.dotidot.io
  28. [28] https://www.youtube.com/watch?v=iKgDAsWE600 www.youtube.com
  29. [29] https://www.youtube.com/watch?v=OxMe4PcKCjo&vl=en www.youtube.com
  30. [30] https://www.definedigitalacademy.com/blog/the-biggest-mistakes-i-see-with-demand-gen-campa… www.definedigitalacademy.com
  31. [31] https://www.growthspreeofficial.com/blogs/google-ads-demand-gen-b2b-saas-setup-audience-st… www.growthspreeofficial.com
  32. [32] https://searchengineland.com/google-clarifies-sensitive-audience-targeting-rules-for-deman… searchengineland.com
  33. [33] https://theb2bplaybook.com/b2b-demand-generation-strategy-2026 theb2bplaybook.com
  34. [34] https://support.google.com/google-ads/answer/14509385?hl=en support.google.com
  35. [35] https://almcorp.com/es/blog/google-demand-gen-best-practices-february-2026/ almcorp.com
  36. [36] https://support.google.com/google-ads/answer/16797388?hl=en support.google.com
  37. [37] https://services.google.com/fh/files/blogs/external_demandgen_implementation_guide.pdf services.google.com
  38. [38] https://business.google.com/ca-en/accelerate/resources/articles/learn-what-new-in-demand-g… business.google.com
  39. [39] https://support.google.com/google-ads/answer/13695389?hl=en support.google.com
  40. [40] https://pulse.coobx.net/en/google/mastering-demand-gen-dv360-best-practices-2026/ pulse.coobx.net
  41. [41] https://support.google.com/google-ads/answer/15209611?hl=en support.google.com
  42. [42] https://www.youtube.com/watch?v=xNkUEecQxvc www.youtube.com
  43. [43] https://directiveconsulting.com/blog/the-b2b-marketers-guide-to-google-ads-best-practices-… directiveconsulting.com
  44. [44] https://www.definedigitalacademy.com/blog/how-google-ads-will-work-in-2026 www.definedigitalacademy.com
  45. [45] https://blog.google/products/ads-commerce/new-demand-gen-features-2025/ blog.google
  46. [46] https://business.google.com/en-all/accelerate/announcements/ business.google.com
  47. [47] https://blog.google/products/ads-commerce/demand-gen-drop-january-2026/ blog.google
  48. [48] https://blog.google/products/ads-commerce/demand-gen-drop-june-2026/ blog.google
  49. [49] https://support.google.com/google-ads/answer/15110871?hl=en support.google.com
  50. [50] https://business.google.com/us/accelerate/announcement/introducing-demand-gen-drops-to-hel… business.google.com
  51. [51] https://blog.google/products/ads-commerce/demand-gen-drop-march-2026/ blog.google
  52. [52] https://business.google.com/us/accelerate/gml-announcements/ business.google.com
  53. [53] https://blog.google/products/ads-commerce/demand-gen-drop-november-2025/ blog.google
  54. [54] https://blog.google/products/ads-commerce/google-display-ads-demand-gen/ blog.google
  55. [55] https://www.groas.com/post/google-ads-demand-gen-campaigns-the-complete-2026-guide-from-di… www.groas.com
  56. [56] https://www.reddit.com/r/googleads/comments/1v38fz9/no_more_google_display_ads_display_ads… www.reddit.com
  57. [57] https://openmoves.com/blog/googles-newest-campaign-type-demand-gen-replaces-discovery-camp… openmoves.com
  58. [58] https://www.customerlabs.com/blog/heres-whats-inside-latest-google-ads-demand-gen-campaign… www.customerlabs.com
  59. [59] https://www.storegrowers.com/demand-gen-benchmarks/ www.storegrowers.com
  60. [60] https://affectgroup.com/blog/demand-gen-benchmarks-ctr-and-cpc-planning-ranges/ affectgroup.com
  61. [61] https://www.digitalapplied.com/blog/google-ads-demand-gen-cpm-billing-2026-advertiser-play… www.digitalapplied.com
  62. [62] https://www.attnagency.com/blog/google-demand-gen-campaigns www.attnagency.com
  63. [63] https://roa-marketing.com/blog/google-ads-demand-gen-campaigns-2026/ roa-marketing.com
  64. [64] https://www.youtube.com/watch?v=5L02E3zZeTs www.youtube.com
  65. [65] https://www.dataslayer.ai/blog/how-to-track-youtube-ads-performance-in-2026-metrics-benchm… www.dataslayer.ai
  66. [66] https://support.google.com/google-ads/answer/13695597?hl=en support.google.com
  67. [67] https://searchlab.nl/en/statistics/google-ads-statistics-2026 searchlab.nl
  68. [68] https://bir.ch/blog/google-ads-cost-breakdown bir.ch
  69. [69] https://metricnexus.ai/blog/google-ads-benchmarks-2026 metricnexus.ai
  70. [70] https://usermaven.com/blog/google-ads-benchmarks usermaven.com
  71. [71] https://www.storegrowers.com/youtube-ads-benchmarks/ www.storegrowers.com
  72. [72] https://www.storegrowers.com/google-ads-benchmarks/ www.storegrowers.com
  73. [73] https://www.digitalapplied.com/blog/google-ads-benchmarks-2026-cpc-ctr-cvr-industry www.digitalapplied.com

Don’t want to do this yourself?

Get a complete, ready-to-launch Google Ads campaign — keywords validated against real search data, copy written to the best practices on this page, and a proper negative-keyword list — built for your business and delivered within 24 hours.

Build my campaign — $129

One-off price · Human-reviewed · No subscription · No access to your Google Ads account

Share the Post:

Related Posts