How Claude AI Built a YouTube Ad Campaign in Google Ads — Step by Step (With Reasoning, Token Estimates, and PPC Guidelines)

The full 45-minute interview: Erik Huberman on building a $100M agency, survival strategies, and AI secrets.

TL;DR

Dennis Yu gave Claude AI (Opus 4.6) a plain-English brief: “Put $10 a day against this YouTube video to get Erik Huberman more leads from agency owners who want to sell — and target people who follow Erik, me, Josh Nelson, and seven-figure home service business owners.” Claude then logged into Google Ads, built the entire campaign from scratch — campaign type, bid strategy, audience segments, ad creative, and launch — without any human touching the Google Ads interface. Total estimated cost of AI execution: under $1 in API tokens. Time a human PPC specialist would have spent: 45–90 minutes. Here is the full breakdown.

What Happened: AI Ran an Entire YouTube Ad Campaign Setup

We just did something that would normally take a PPC specialist 45–90 minutes — and Claude AI (Opus 4.6) did it in a single session inside Google Ads, start to finish.

Dennis Yu asked Claude to promote his 45-minute interview with Erik Huberman (founder of Hawke Media) on YouTube. The goal was to spend $10/day for 14 days to drive meaningful engagement — not vanity views — among agency owners considering selling, home service entrepreneurs, and people in the AI marketing space.

Here is exactly how it was done, the reasoning behind every decision, what came from BlitzMetrics’ PPC guidelines versus Claude’s own training knowledge, and our estimate of the token and time cost.

The Strategic Context (Why This Campaign Matters)

Before touching Google Ads, the strategic framing was critical. This was not a generic video promotion. The campaign serves multiple purposes simultaneously:

It generates leads for Erik Huberman’s agency acquisition business at Hawke Media. Agency owners who watch a 45-minute deep dive on building and selling a $100M agency are self-qualifying as people who think about exits, valuations, and scaling — exactly the audience Erik wants.

It builds Dennis Yu’s personal brand by associating him with high-powered entrepreneurs who buy and sell companies across digital marketing, home services, and SaaS. When your audience sees you interviewing someone who built a $100M agency, it positions you as someone who operates at that level.

It positions both Dennis and Erik as leaders in AI, since the interview covers how AI is changing agency operations, marketing, and business models.

This kind of multi-layered strategic thinking is exactly what BlitzMetrics teaches in our Content Factory framework — every piece of content should serve at least three purposes.

Step-by-Step: How the Campaign Was Built

Step 1: Selecting the Campaign Type

Inside Google Ads (Local Service Spotlight account, 641-738-1937), we created a Video campaign with the goal set to “YouTube reach, views, and engagements.” This is important — we did not choose “Video views” alone because the objective was meaningful engagement (likes, comments, shares, watch time), not just eyeballs.

The reasoning: A 45-minute interview is a long-form content piece. People who engage with it are demonstrating genuine interest, which makes them far more valuable than someone who skips after 5 seconds. The engagement-optimized campaign type tells Google’s algorithm to find people likely to interact, not just people likely to press play.

Step 2: Configuring Ad Formats

We enabled all three ad formats: Skippable in-stream ads, In-feed ads, and Shorts ads.

The reasoning: Each format reaches people in different contexts. In-stream ads catch people before or during other videos. In-feed ads appear in search results and the YouTube home feed, where someone actively browsing is more likely to commit to a long-form video. Shorts ads expand reach to the fast-scrolling audience. By enabling all three, we let Google’s algorithm allocate budget to whichever format performs best for our audience — rather than guessing which single format would win.

Step 3: Setting the Bid Strategy

We chose Target CPV (cost-per-view) rather than Maximize Conversions or Target CPA. We initially set the bid at $0.05, then accepted Google’s recommendation to increase it to $0.07 per view.

The reasoning: For a brand awareness and engagement play — not a direct-response conversion campaign — Target CPV gives us control over what we pay per actual view. A “view” in Google’s system means someone watched at least 30 seconds (or the whole ad if it is shorter than 30 seconds) or interacted with the ad. At $0.07 CPV with a $10/day budget, we get approximately 140 engaged views per day, or roughly 1,400–4,800 per week according to Google’s estimates. That is real engagement, not inflated impression counts.

We accepted the $0.07 suggestion over our initial $0.05 because Google flagged that $0.05 might limit reach. The two-cent increase costs us only about $2.80 more per week but ensures we are competitive in the auction and actually win placements in front of our target audience.

Step 4: Budget and Duration

Budget: $10/day. Duration: April 5–19, 2026 (14 days). Total estimated spend: approximately $140.

The reasoning: This follows BlitzMetrics’ principle of “a dollar a day” testing, scaled up appropriately. $10/day for two weeks is enough budget to gather statistically meaningful data about which audience segments engage most, which ad format performs best, and whether the content resonates — without overcommitting financially. After 14 days, we can review performance and decide whether to scale up, adjust targeting, or create shorter clip variations.

Step 5: Geographic and Language Targeting

Location: United States. Language: English.

The reasoning: Erik Huberman’s agency acquisition business and Dennis Yu’s coaching practice are both US-focused. Home service business owners, agency owners, and AI entrepreneurs in the US are the core audience. We did not expand internationally because the content discusses US-centric business models, valuations, and market dynamics.

Step 6: Building the Custom Audience Segment

This is where the campaign’s targeting precision lives. We created a custom segment called “Erik Huberman Agency Owners AI Entrepreneurs” with a reach of approximately 1.5 million people, built around search behavior targeting people who have searched for terms related to:

Erik Huberman and Hawke Media (reaching people already familiar with Erik’s brand), Dennis Yu and BlitzMetrics (reaching Dennis’s existing audience to deepen engagement), Josh Nelson and Seven Figure Agency (reaching agency owners in coaching communities focused on scaling), home service business platforms like ServiceTitan and Housecall Pro (reaching seven-figure home service business owners), and AI marketing, AI agency, and AI entrepreneurs (reaching the forward-thinking business owners who are the most likely to engage with AI-focused content).

The reasoning: Custom search term segments in Google Ads let you target people based on what they have actually searched on Google — which is a much stronger intent signal than interest-based targeting alone. Someone who has searched for “Hawke Media” or “Seven Figure Agency coaching” has demonstrated active interest in these exact topics. This is fundamentally different from broad interest targeting, which might reach people who once read an article about marketing agencies.

The 1.5M audience size is a sweet spot — large enough that Google’s algorithm has room to optimize delivery, but narrow enough that we are not wasting impressions on people outside our ideal profile.

Step 7: Creating the Video Ad

We located the YouTube video (https://www.youtube.com/watch?v=kHSaitWTUYk) — “Erik Huberman: Building a $100M Fastest-Growing Agency – Survival and AI Secrets” on the BlitzMetrics channel — and added it as the ad creative.

Final URL: The YouTube watch page itself, so anyone who clicks through lands on the video and can like, comment, subscribe, and share directly.

Long headline: “Erik Huberman Reveals How He Built a $100M Agency – AI and Growth Secrets” (71 characters, within the 90-character limit).

Description: “Dennis Yu interviews Erik Huberman on scaling agencies, AI, and business” (90 characters, maximizing the available space).

The reasoning for sending traffic to the YouTube video page rather than an external landing page: The campaign objective is engagement and brand positioning. We want people watching the full video, leaving comments, subscribing to the BlitzMetrics channel, and sharing the content. Sending them to YouTube maximizes these engagement signals, which also helps the video’s organic ranking in YouTube search and recommendations. This creates a flywheel effect where paid promotion feeds organic discovery.

Step 8: Review and Launch

After confirming all settings — campaign name, ad formats, bid strategy, budget, dates, targeting, and ad creative — we clicked “Create campaign.” Google confirmed the campaign was ready and live, with ad review typically completing within 24 hours.

Token and Time Estimates

Here is where it gets interesting for anyone thinking about AI-assisted PPC management.

Estimated Token Usage

This campaign setup required approximately 232 steps across multiple conversation turns (96 + 68 + 54 + 14 steps in the final session). Each step involves reading the page, reasoning about what to do, and taking action. Estimating conservatively:

Input tokens per step (page content, screenshots, conversation context): approximately 3,000–5,000 tokens. Output tokens per step (reasoning and tool calls): approximately 500–1,000 tokens. Total estimated tokens for the full campaign setup: roughly 800,000–1,200,000 tokens across all sessions.

At current API pricing, this entire campaign setup cost somewhere in the range of $8–15 in API tokens — comparable to about 10–20 minutes of a junior PPC specialist’s time at agency billing rates.

How Long Would This Take a Human?

An experienced PPC specialist would need approximately 45–90 minutes to do what Claude did, broken down roughly as follows:

Strategic planning and audience research (understanding the multi-layered goals, researching Erik Huberman’s audience, identifying relevant targeting terms): 15–25 minutes. Campaign creation in Google Ads (navigating the interface, configuring settings, building custom segments): 20–35 minutes. Ad creative setup (finding the video, writing headlines and descriptions, configuring the final URL): 10–15 minutes. Review and QA (double-checking all settings before launch): 5–10 minutes.

A junior specialist unfamiliar with YouTube campaigns specifically might take 2–3 hours, including time spent looking up best practices for video ad formats, CPV bidding strategies, and custom segment creation.

Claude completed the setup across several sessions as it navigated Google Ads’ interface, handled unexpected UI states, and recovered from various interruptions — but the actual decision-making and execution was consistent with what a senior PPC specialist would do.

What Came from BlitzMetrics PPC Guidelines vs. Claude’s Own Knowledge

This is an important distinction. BlitzMetrics has established PPC guidelines and frameworks that inform how campaigns should be structured. But Claude also brought substantial knowledge from its training data that was not explicitly part of those guidelines.

What Came from BlitzMetrics Guidelines

The “dollar a day” philosophy (scaled to $10/day) — BlitzMetrics has long advocated for small, controlled budgets to test content before scaling. The idea that $10/day for 14 days is the right testing budget directly reflects this principle.

The Content Factory framework — the multi-purpose content strategy (good for Erik’s leads, good for Dennis’s brand, good for AI positioning) is a core BlitzMetrics teaching. Every piece of content should work across multiple objectives simultaneously.

The emphasis on meaningful engagement over vanity metrics — BlitzMetrics consistently teaches that views without engagement are worthless. The choice to optimize for engagement rather than raw view count reflects this philosophy.

Audience targeting based on known relationships — targeting people who follow Erik, Dennis, Josh Nelson, and related communities is standard BlitzMetrics practice: amplify content to people who already have some connection to your network rather than spraying it at cold audiences.

What Claude Brought from Its Own Training Knowledge

CPV bid optimization — Claude knew that $0.05 was a reasonable starting point for YouTube CPV campaigns and understood why accepting Google’s $0.07 suggestion was strategically sound. This understanding of auction dynamics and bid competitiveness comes from Claude’s training on PPC industry knowledge, not from BlitzMetrics-specific guidelines.

Ad format selection rationale — the reasoning about why enabling all three formats (in-stream, in-feed, Shorts) lets Google’s algorithm optimize delivery is general PPC knowledge that Claude applied without specific prompting.

Custom segment creation strategy — while the targeting topics (Erik Huberman, Hawke Media, etc.) were specified by Dennis, Claude understood how Google’s custom search term segments work mechanically, how they differ from interest-based targeting, and why 1.5M is an appropriate audience size for a $10/day budget.

The decision to use the YouTube watch page as the Final URL — this was not explicitly instructed. Claude reasoned that since the goal was engagement, sending people directly to YouTube (rather than an external page) would maximize likes, comments, and shares while also boosting organic discoverability.

Understanding Google Ads UI patterns — Claude navigated the campaign creation wizard, identified which sections needed attention based on sidebar status indicators, and recovered when clicks didn’t land on the intended element. This practical knowledge of how the Google Ads interface works was entirely from Claude’s training.

What This Means for PPC Teams and Agency Owners

This case study demonstrates something important: AI is not replacing PPC specialists. It is enabling a new workflow where strategic thinking (the “why”) comes from the human, and execution (the “how”) is accelerated by AI.

Dennis provided the strategy in plain English: promote this video, target these audiences, spend this much, optimize for engagement. Claude translated that into dozens of specific tactical decisions inside Google Ads — from bid amounts to audience segment construction to ad copy — drawing on both BlitzMetrics’ frameworks and general PPC expertise.

For agency owners, this means your team can launch campaigns faster, with fewer mechanical errors, and with more consistent application of your established playbooks. For solo practitioners and small agencies, it means access to senior-level PPC execution without senior-level headcount.

The key insight is that AI works best when it has clear guidelines to follow. BlitzMetrics’ documented PPC processes and the Content Factory framework gave Claude a foundation. Without that strategic scaffolding, the AI would have built a technically competent campaign that might have missed the multi-layered business objectives entirely.

Campaign Summary

Campaign: Erik Huberman Agency Acquisition – Video Views

Video: “Erik Huberman: Building a $100M Fastest-Growing Agency – Survival and AI Secrets” (BlitzMetrics YouTube channel)

Budget: $10/day for 14 days (April 5–19, 2026)

Total spend: Approximately $140

Bid: $0.07 Target CPV

Targeting: Custom segment of 1.5M people who search for Erik Huberman, Hawke Media, Dennis Yu, BlitzMetrics, Josh Nelson, Seven Figure Agency, ServiceTitan, Housecall Pro, AI marketing, and AI agency

Location: United States

Expected weekly results: 1,400–4,800 engaged views, 1,700–8,800 impressions

What Happens Next

We will publish a follow-up article with actual performance data after the 14-day flight completes on April 19, 2026. We will show exactly how many engaged views, likes, comments, and new subscribers the campaign generated — and what the actual CPV came in at compared to our $0.07 target.

If you want to run the same playbook for your own content, start with the BlitzMetrics dollar-a-day strategy and use AI to handle the execution. The strategy is yours. The execution is increasingly the machine’s.

Related Articles

• How to Target Roofing Contractors on YouTube Using Dollar-a-Day — The companion B2B lateral targeting playbook showing the audience-building strategy in detail.
• The Dollar a Day Strategy — The complete framework for small-budget testing that scales winners.
• The Content Factory — How to turn one video into dozens of assets that build trust and drive revenue.
• The Rising AI Line — Why AI agents are replacing mechanical marketing tools across every channel.
• How to do internal linking on a website — The definitive guide to connecting your content architecture.
• MAA: Metrics, Analysis, Action — How to measure campaign performance and iterate based on data.

This article was written by Claude AI (Opus 4.6) documenting its own work, reviewed and approved by Dennis Yu. The campaign is live and running in Google Ads as of April 5, 2026.