- Marketing Machine
- Posts
- Meta's Opportunity Score 🤓 HubSpot's ChatGPT Connector 😬 An Inflection Point For The Creator Economy 📈
Meta's Opportunity Score 🤓 HubSpot's ChatGPT Connector 😬 An Inflection Point For The Creator Economy 📈
Also: Scientists Discover That Feeding AI Models 10% 4Chan Trash Actually Makes Them Better Behaved
Meta's Opportunity Score 🤓 HubSpot's ChatGPT Connector 😬 An Inflection Point For The Creator Economy 📈
Also: Scientists Discover That Feeding AI Models 10% 4Chan Trash Actually Makes Them Better Behaved

Happy Friday Marketing Engineers!
Today:
🗺️ UK Government Harnesses Gemini To Support Faster Planning Decisions
🤖 Introducing The V-JEPA 2 World Model And New Benchmarks For Physical Reasoning
🚦 5 Reasons Direct Traffic Can Drop Suddenly
📧 What Is Email Marketing?
🤑 How To Use Landing Page Sales Funnels To Convert More Visitors
🦾 Scaling An AI Native Company Part 2
🚀 Ada{m}
P.S. Were you forwarded this email? Click here to subscribe
Portfolio worries? Not anymore
Stop stressing over lost work, outdated portfolios, and missing links.
Authory automates your portfolio updates and backs up everything you write forever. No more anxiety about losing your content when a site shuts down.
Everything is organized, searchable, and ready to share. With Authory, you can write confidently knowing your portfolio is always in top shape.
Thousands of writers already have peace of mind — it’s your turn.
🗞 Marketing News
Meta Boosts Campaigns With Opportunity Score [2 min read]
Meta is rolling out Opportunity Score globally in Ads Manager, a new tool that provides a 0-100 optimization score along with real-time, personalized recommendations that helped advertisers achieve a 12% median decrease in cost per result during testing. This represents an evolution from static frameworks like Performance 5 to a dynamic, machine learning-powered approach that makes it easier for advertisers to find and implement the right recommendations for their specific accounts without guesswork. Also, all sales and app campaigns now start fully enabled with AI-driven optimizations integrated with the opportunity score, delivering 7-9% CPA improvements while maintaining access to manual controls when needed.
HubSpot's announcement of a deep research connector to ChatGPT has generated significant excitement among users and martech observers, with many viewing it as a potential paradigm shift for SaaS applications and AI integration. However, sharing customer data with any AI technology carries inherent security risks that require careful evaluation based on each organization's risk appetite and existing data-sharing policies, despite HubSpot's assurance that ChatGPT cannot use the data for training purposes. Beyond security considerations, the connector's effectiveness will ultimately depend on the quality and structure of CRM data, since poor data quality will limit ChatGPT's ability to provide useful insights, following the principle of "garbage in, garbage out.
An Inflection Point For The Creator Economy [6 min read]
Ad revenue from user-generated content is set to surpass traditional media revenue for the first time, with creator revenue expected to increase 20% this year and more than double to $376.6 billion by 2030. As creators gain access to larger budgets for content production, creator-produced media will become increasingly indistinguishable from traditional media in quality and scope. Eventually, the term "Creator Economy" will feel outdated as the boundaries between creator content and traditional media blur into one unified media landscape.
🤖 AI
Scientists Discover That Feeding AI Models 10% 4Chan Trash Actually Makes Them Better Behaved [3 min read]
Including a controlled amount of toxic content from 4chan during AI model training actually makes models better behaved and easier to detoxify later, contrary to the typical practice of filtering out all harmful content beforehand. When models are trained with some toxic data, harmful concepts become more distinct and separated within the model's internal structure rather than being entangled with other concepts, making these toxic elements much easier to identify and suppress without affecting overall performance. Models trained with 10% toxic content not only generated the least harmful output when detoxification techniques were applied, but also proved more resilient against jailbreak attacks and outperformed other detoxification methods.
The UK Government developed Extract, an AI system built with Google's Gemini that converts paper planning documents into digital data in 40 seconds versus the current 2-hour manual process. Extract uses Gemini's multimodal capabilities to read complex documents, extract information from blurry maps and handwritten notes, and convert historical maps to modern digital formats with precise geographical coordinates. Rolling out to all councils by Spring 2026, the system will help process England's 350,000 annual planning applications faster while demonstrating AI's potential for transforming other public services.
Meta released V-JEPA 2, an AI model that can watch videos to understand how the physical world works and use that knowledge to control robots in new environments it has never seen before. The model can successfully complete basic robot tasks like picking up and placing objects with 65-80% success rates by planning actions based on what it learned from watching over 1 million hours of video. However, new tests show that while humans easily understand physics from videos with 85-95% accuracy, current AI models including V-JEPA 2 still struggle significantly with basic physical reasoning tasks.
🧠 Hacks & Tactics
5 Reasons Direct Traffic Can Drop Suddenly [3 min read]
Direct website traffic can suddenly drop due to technical issues like analytics outages and website downtime, which affect all traffic channels simultaneously. Changes in user behavior patterns, such as regional holidays or social media sharing through private channels ("dark social"), can also cause significant drops in direct traffic numbers. Session stitching problems and improper tracking code implementation can lead to misattributed traffic sources, making direct traffic appear lower than it actually is.
What Is Email Marketing? [20 min read]
Email marketing encompasses a variety of communication types, including cold outreach, opt-in newsletters, post-purchase interactions, platform engagement emails, automated sequences, and transactional messages, each requiring specific approaches. Successful engagement and retention rely on effective strategies such as personalization, targeted content creation, optimal design, appropriate timing, and adherence to regulatory standards. Also, managing and continuously optimizing campaigns through database maintenance, performance analytics, automation, A/B testing, and ensuring technical deliverability is essential to maximize their impact.
Landing page funnels are designed for single-session completion with traditional top, middle, and bottom stages, distinguishing them from broader marketing funnels that represent longer customer journeys. A successful funnel follows a specific structure with click-through pages at the top to initiate engagement, middle pages to maintain interest, and conversion pages at the bottom to capture information or complete sales. The implementation of these funnels requires adherence to conversion-centered design principles, proper analytics setup from the start, and continuous optimization through A/B testing rather than a "set it and forget it" approach.
🏗️ For Founders
Scaling An AI Native Company Part 2 [6 min read]
When building AI agents for companies, you need Agent Job Descriptions (AJDs) that define the outcome the agent should achieve, required context, security/privacy requirements, and budget, rather than traditional hiring elements like company culture and benefits. Agents can be built at four different levels from Level 0 (chat platforms like ChatGPT) to Level 3 (custom code), with each level offering different balances of capability, context, and complexity, where higher levels cost less per compute dollar but require more technical expertise and architectural thinking. Despite social media hype claiming AI agents can magically replace entire teams, building effective agents requires careful thought, logical understanding of when humans should be involved, and significant technical considerations, especially for agents that need precision, scale, and extensive context.
🔮 The Future
AI Max uses Google's AI to automatically expand your search campaigns beyond your chosen keywords, creating new ad copy and sending users to different pages on your site, which can increase conversions by 14% but reduces your control over targeting and messaging. It works best for advertisers using narrow keyword matching who want to discover new search opportunities, while regular Search campaigns are better for those who need predictable costs and precise control over their ads. Most successful advertisers use both approaches together, with regular Search for their most important keywords and AI Max for finding new opportunities.
12 New KPIs For The Generative AI Search Era [8 min read]
Traditional SEO metrics like clicks and rankings are becoming less useful because people now get answers directly from AI chatbots instead of clicking through search results. New metrics are needed to track how well your content performs in AI systems, such as how often it gets cited in AI answers, how easily AI can find and use your content, and whether AI bots can properly crawl your website. Marketers should start measuring these AI-focused metrics now instead of only relying on old-school SEO numbers that don't reflect how people actually search today.
YouTube Quietly Relaxes Content-Moderation Policies [2 min read]
YouTube has quietly relaxed its content moderation policies by internally coaching moderators to prioritize "freedom of expression" over strict rule enforcement for content that might be in the "public interest," including videos discussing elections, ideologies, race, gender, and other controversial topics. The platform doubled the amount of offensive material permitted in videos with public interest value from a quarter to half the video, and encouraged moderators to leave up content like City Council meetings, campaign rallies, and political conversations that would previously have been removed. While this policy shift aligns with Meta and X's move toward community-driven moderation over traditional fact-checking, YouTube has not publicly disclosed these changes, instead implementing them through internal training materials in mid-December that were later discovered by The New York Times.
👩🏫 Vocab
Cross-Selling: The practice of selling additional products or services to existing customers. It focuses on suggesting complementary items that enhance the customer's primary purchase. Cross-selling increases revenue while providing more value to customers.
Example: A bank offers a credit card to customers who open a checking account, or Amazon suggesting items "frequently bought together."
Recommendation Engines: Automated systems that analyze customer data to suggest relevant products or content. They use algorithms to predict what customers might like based on their behavior and preferences. Recommendation engines make shopping experiences more personalized and convenient.
Example: Netflix suggests shows based on your viewing history and ratings, helping you discover content you'll likely enjoy.
Bridge the gap with student loans
June’s here—are your college costs covered? With final aid letters rolling in, now’s the time to see if there’s a gap. Federal aid might not cut it, but private loans can. Schools suggest applying early, so don’t wait. Check out Money’s top picks for low-rate, hassle-free student loans.
That's it for this week! See you next week. 😎
Forwarded this email? Click here to subscribe. Don't want to miss out? Make sure to whitelist our email! 🥳😉
Something else you’d like to see here? Hit reply and let us know what you think. 🧠 Or smash a button on our feedback poll!
✌️ Team Marketing Machine
🗳️ Feedback Dept!
We want to hear from you! 😎🫵What do you think of the newsletter so far? |
🥞 Stack
Tools we use to keep our business running.
Discord - find community with text, voice, and video chat. Join us!
Beehiiv - engage newsletter audiences and grow subscribers
Clickup - project management and team collaboration
WordPress - build websites and blogs
Elementor - drag-and-drop builder for WordPress
Buffer - streamline social media scheduling and analytics
ElevenLabs - convert text into realistic voiceover
Riverside - record high-quality audio and video remotely
Transistor - all-in-one podcast hosting platform
Zapier - automate tasks by connecting web apps
Claude - AI language model focused on safe and ethical text processing
Midjourney - generate images from text prompts using AI
*contains affiliate links - let us know if you have any questions or recommendations!