How does YouTube advertising actually work in 2026?
YouTube ads are bought through Google Ads. The platform routes paid inventory through several campaign types, but the one that matters most for reach and conversion in 2026 is Demand Gen, which replaced Video Action campaigns in 2024 and is now the default path for advertisers who want YouTube alongside Gmail and Discover in a single buying structure.
The active ad formats on YouTube itself, as documented in Google's help centre, are:
- Skippable in-stream ads: play before or during videos; viewer can skip after 5 seconds. Available on desktop, mobile, TV, and consoles.
- Non-skippable in-stream ads: 15 or 20 seconds (regional variation); must complete before the content plays. Also available across all devices, with 30-second versions on TV.
- Bumper ads: up to 6 seconds, non-skippable. Enabled automatically when skippable or non-skippable formats are on.
- Shorts ads: vertical-format skippable ads served in the Shorts feed.
Overlay ads, which appeared on the lower portion of the video player on desktop, were retired in April 2023. Source: YouTube Help Centre, ad formats documentation. YouTube's advertising formats reference is the primary source for current format specifications and device availability; the YouTube for Advertisers blog covers platform announcements and format changes.
Google Display Network campaigns are also migrating into Demand Gen through 2026, a change documented in the Google Ads Help Centre on Demand Gen (answer 13695777). If you are still running standalone Video or Display campaigns from a pre-2024 structure, that infrastructure is becoming a legacy path.
Platform note
Demand Gen is not a YouTube-only campaign type. It allocates budget across YouTube (including Shorts), Gmail, and Google Discover. Your YouTube share of spend within a Demand Gen campaign is algorithm-controlled, not directly settable. This is the most common source of misaligned expectations between buyers and their platform partners.
What does Demand Gen actually do, and where does the AI stop?
Demand Gen (short for Demand Generation) is Google's answer to the social video ad units that Meta and TikTok popularised: a campaign type that runs creative across immersive, feed-like surfaces and lets the algorithm decide placement mix, audience targeting, and creative variant rotation in real time.
The AI in Demand Gen, powered by Gemini, tests creative combinations across your asset group and optimises toward your conversion goal. It can rotate headline and description variants, select between video lengths, and decide whether a given impression goes to YouTube or Gmail based on predicted likelihood of conversion. These are the functions the AI performs well when you supply enough input material.
What Demand Gen's AI cannot do:
- Fix a broken conversion architecture. If your landing page has no clear conversion event, or if your Google Ads conversion tag fires on the wrong signal, the AI optimises toward the wrong outcome.
- Create the creative for you. Demand Gen works well when you supply at least five asset variants: a short hook video (under 30 seconds), a longer explanation video, at least one vertical Shorts-spec asset, and multiple headline and description combinations. Feeding a single creative asset into Demand Gen produces a single-variant campaign with no AI leverage.
- Provide placement-level YouTube reporting. Demand Gen aggregates results across YouTube, Gmail, and Discover. If you need to know specifically what YouTube is contributing independently of Gmail, you need supplementary measurement, such as brand lift studies, Marketing Mix Modelling (econometric models measuring incremental channel contribution), or tagged UTM parameters read through your own analytics platform.
- Guarantee Shorts coverage. Shorts placements are eligible inside Demand Gen, but you need at least one vertical creative asset in your asset group for the AI to use that placement.
This is not a criticism of the platform. It is a structural description. The gap between what Demand Gen offers and what a buyer assumes it does is where most YouTube advertising underperformance originates. A competent partner closes that gap before the campaign launches, not after the first report.
What are the three YouTube advertising partner models?
Most buyers evaluating YouTube advertising support end up choosing between three structural options. Each has honest advantages and honest constraints.
1. Traditional YouTube advertising agency
An agency takes a retainer and manages your campaigns under a managed-account structure, often meaning your campaigns run from their Google Ads Manager account rather than yours. The agency handles campaign build, creative trafficking, reporting, and sometimes creative production.
Where this model works well: Brands that do not want any media operations in-house and are comfortable with the agency owning both strategy and access. High-volume, high-creative-output advertisers who need consistent production and trafficking throughput.
Where it fails: Managed accounts mean the advertiser loses access to historical data and audience lists when the relationship ends. Retainer structures can disincentivise the agency from simplifying the account, because a more complex account is harder to replace. Demand Gen requires creative production cadence; agencies that were built on static display trafficking often cannot sustain it.
2. In-house YouTube advertising team
An internal hire or team owns the Google Ads account, builds and manages campaigns, and reports internally. This is the highest-control option and, when resourced correctly, the lowest marginal cost at scale.
Where this works well: Businesses with consistent YouTube spend above levels that justify a full-time hire, a stable creative pipeline, and clear internal measurement capability. Markets like the US and Australia, where digital marketing talent is available, make this more viable than Singapore or Malaysia where the specialist labour pool is smaller.
Where it fails: Demand Gen is a system that rewards platform-level expertise and cross-advertiser pattern recognition. An in-house team working one account does not accumulate the comparative signal that an external team working twenty accounts accumulates. It also fails when the team is under-resourced for creative production, which is one of the most common in-house failure modes on video channels.
3. AI-native consultancy
A consultancy works inside your account (not a managed account), architects the campaign structure and measurement model, and either runs campaigns on an advisory basis or manages a setup-and-handover engagement. The ongoing role is more about decision support and structural review than daily bid management, which Demand Gen's AI handles.
Where this works well: Businesses that want platform strategy and measurement architecture without surrendering account access. Advertisers who have Demand Gen running but are not confident about what it is optimising toward or why. Businesses entering YouTube for the first time who need to set the system up correctly before deciding whether to run it in-house or with ongoing support.
Where it fails: If you need consistent high-volume creative production and trafficking, a consultancy is not the right primary production partner. The model is better suited to structural and strategic decisions than ongoing creative throughput.
The distinction between the second and third model is partly about seniority level. A well-resourced in-house team with someone who has built Demand Gen campaigns across multiple accounts is a strong setup. A junior in-house hire managing YouTube alongside ten other channels is the worst of all three options.
For a parallel analysis of the same three-model question on Google Search, see our guide on Google Ads agency vs AI consultancy. The considerations overlap significantly.
Partner model comparison: who handles what
| Capability | Traditional agency | In-house team | AI-native consultancy |
|---|---|---|---|
| Account ownership | Often agency-held MCC; data risk on exit | Always advertiser-held | Always advertiser-held; consultancy works as guest access |
| Demand Gen architecture | Variable; depends on team vintage | Variable; depends on hire experience | Core competency; structural setup is primary deliverable |
| Creative production | Often included or subcontracted | Depends on resourcing | Brief and direction; not production throughput |
| Shorts creative (9:16) | Depends on creative capability | Depends on internal creative | Spec and brief; client or separate creative partner produces |
| Measurement beyond platform | Often platform-only reporting | Depends on analytics capability | MMM (Marketing Mix Modelling) and independent measurement setup included |
| Cross-advertiser pattern recognition | High if agency runs many YouTube accounts | Limited to own account history | High across client base |
| CTV (Connected TV) YouTube coverage | Standard YouTube Select available | Available via Google Ads device targeting | Informed by broader CTV strategy; see CTV closed-loop guide |
| Contract structure | Monthly retainer; often 3-month minimum | Salary + benefits; fixed cost | Project-based or quarterly advisory; no long lock-in |
Which YouTube advertising partner model fits your situation?
Work through the flowchart below. Each decision node is a real structural question, not a preference question.
Read left-to-right. Each decision is a structural question.
How does YouTube advertising vary across Singapore, Malaysia, Australia, the US, and Canada?
YouTube's reach is strong in all five markets leapbuzz serves. The platform differences that matter for buyers are cost structure, talent availability, and local regulatory context for certain categories.
| Market | CPM context | Demand Gen talent | Category notes |
|---|---|---|---|
| Singapore (SG) | Higher CPMs relative to regional peers; premium inventory in a small, high-income market | Small but capable specialist pool; cross-market experience common | MAS-regulated financial and insurance categories require careful claims management across all video placements |
| Malaysia (MY) | Lower CPMs than SG; strong reach across Malay, Chinese, and English language audiences | Growing in-house capability; bilingual creative often needed | Language and cultural targeting requires local brief input; English-only creative underperforms |
| Australia (AU) | Mid-to-high CPMs; competitive in finance, retail, and travel categories | Strong in-house and independent market; Google Premier Partner presence | ACCC consumer law applies to advertising claims; financial services category requires ASIC-compliant disclosures |
| United States (US) | Highest CPMs of the five; most competitive inventory, especially Q3-Q4 | Deepest talent pool globally; widest agency choice | FTC guidelines on endorsement and AI-generated content disclosure; state-level privacy law variation |
| Canada (CA) | CPMs broadly similar to AU; French-language Quebec adds creative complexity | Strong market; significant US agency crossover | CRTC broadcast rules apply to some video formats; French-language content obligations in Quebec for regulated categories |
Cross-market setup note
Running a single Demand Gen campaign across all five markets from one budget pool typically means no market gets enough signal to optimise properly. A structured approach separates at least the US from APAC markets given the CPM differential, and treats Malaysia as a distinct creative brief from Singapore even when running on the same campaign.
The paid social and video landscape for these five markets is covered in more depth on our YouTube advertising platform page and the broader paid social services page.
What questions should you ask a YouTube advertising partner before hiring?
These are the questions that separate a YouTube advertising partner who understands Demand Gen from one who does not.
- Do you work inside my Google Ads account or your own manager account? The answer determines what happens to your data and audience history if you leave. Your own account means you own everything. A manager account means the partner controls access and, depending on the contract, your history may not be fully portable.
- How do you structure a Demand Gen asset group, and how many creative variants do you recommend before launch? A partner who says "one or two videos is fine" has not built Demand Gen with the AI variant-testing logic in mind. The honest answer involves at least five variants covering different lengths, at least one vertical asset, and multiple text combinations.
- How do you report on YouTube performance specifically when Demand Gen aggregates across placements? If the answer is "we pull the Demand Gen report and break down by placement," that is partial. The full answer involves what supplementary measurement they run beyond the platform, whether that is brand lift, Marketing Mix Modelling, or tagged attribution paths through the advertiser's own analytics.
- How do you handle Shorts creative? Can you spec and produce 9:16 vertical video? Shorts inventory is growing inside Demand Gen. A partner with no vertical creative capability is leaving Shorts placements either empty or filled with reformatted horizontal assets that underperform in the vertical feed.
- What is your approach to Connected TV (CTV) YouTube placements? CTV (Connected TV, meaning smart TV and streaming device inventory) is part of YouTube's addressable reach. Creative that works on mobile does not automatically work on the lean-back TV environment. Ask whether the partner considers device type in their creative brief or just lets the algorithm decide. See our CTV closed-loop measurement guide for the broader measurement context.
- Can you show me an example Demand Gen campaign structure, not just a results screenshot? Results screenshots prove a campaign ran. A campaign structure shows whether the partner understands Demand Gen's architecture. Ask for the asset group structure, the conversion goal configuration, and the audience layer setup.
- How do you handle conversion architecture before the campaign launches? Demand Gen's AI optimises toward the conversion event you define in Google Ads. If that event is misconfigured (fires on the wrong page, fires multiple times per user, or fires on a proxy goal rather than a real business outcome), the AI learns the wrong signal. A competent partner audits the conversion setup before touching campaign settings.
Red flag
A YouTube advertising partner who leads with view-through rate as the primary success metric has not moved their measurement model past 2019. View counts tell you about delivery. Incrementality (what lifted because of the campaign versus what would have happened anyway) is the number that matters. Ask what the incrementality testing approach is. If there is no answer, you are buying vanity metrics.
