December 12th-December 18th // Estimated Reading Time: 7 minutes

Google makes Gemini 3 Flash the default AI for billions of users
Source: Axios
December 17th, 2025

Summary: Gemini 3 Flash is now the default model in the Gemini app and AI Mode in search. Gemini 3 Flash matches Gemini 3 Pro's reasoning capabilities while being faster and cheaper, and it outperforms Gemini 3 Pro on coding benchmarks. The release comes a week after OpenAI launched GPT-5.2. Gemini's mobile app downloads, monthly active users, and website visits are all growing faster than ChatGPT's, according to The Information🔒. The launch reflects the intensifying AI race between Google and OpenAI, which entered "code red" mode after Google's Gemini 3 Pro release caused ChatGPT traffic to decline.

Opinion: Google's strategy is clear: make Gemini so good, so cheap, and so ubiquitous that it becomes the default AI for millions of consumers before they even consider ChatGPT. This is significant for marketers because Gemini is most likely to introduce🔒 and succeed with ads before other AI tools, given Google’s strength in advertising. The AI ad opportunity is close…

OpenAI updates ChatGPT Images, a Google Nano Banana competitor

What It Does: GPT Image 1.5 preserves details like lighting, composition, and facial likeness across photo edits while generating images up to 4x faster than its predecessor, according to OpenAI. It's also 20% cheaper. The model excels at precise edits (adding, subtracting, combining elements), creative transformations (changing style while preserving key details), and better text rendering. This directly challenges Google's Nano Banana, which launched in November, and threatens Adobe's photo editing dominance by integrating advanced image capabilities into ChatGPT. The competition will soon heat up further with Meta expected to release its own image and video AI model, code-named Mango.🔒 

Quick Take: The 4x speed improvement and 20% cost reduction could make this particularly attractive for teams generating high volumes of product images, ad creative, and branded content. But … Nano Banana is still a better name.

New IAS Agent optimizes brand safety and performance

What It Does: IAS Agent analyzes 15+years of data and makes recommendations for brand safety and suitability settings. The tool can make recommendations and apply them within marketers’ preferred DSPs,🔒with "explainable AI" that shows the specific signals and reasoning behind each suggestion, which marketers can customize, override, or adopt. For instance, if a campaign is underperforming due to overly restrictive brand safety settings, IAS Agent might alert that you're blocking 40% of quality inventory and recommend loosening specific keyword filters.

Our Take: This is cool. But also, the proliferation of AI assistants in marketing tools overwhelming. Is a specific brand safety AI assistant necessary? Can it be built to communicate and interopate with other marketing agents?

NBCU's Contextual Targeting in LIVE uses AI to match ads with relevant moments

What It Does: The AI-powered tool continuously scans live programming in real time and automatically places ads during the most contextually relevant moments as they unfold. For example, during the upcoming live coverage of the Winter Olympics, if the announcers are talking about the beauty of the Italian Alps, the AI could automatically place a travel booking service ad. Early testing drove higher engagement and favorability. 

Quick Take: This addresses a big live TV problem—brands can't control what happens during a live event, but now they can automatically align their ads with favorable moments in real time using AI.

The Setup: Butler/Till, PubMatic, and Clubtails wanted to test whether AI agents could handle the end-to-end process of planning and executing a connected TV campaign for the alcohol brand, while complying with strict alcohol advertising standards.

The AI Solution: Butler/Till submitted a media brief using natural language through Anthropic's AI platform Claude. PubMatic's AI agents interpreted the brief, devised the media strategy, created the campaign, and now manage pacing, targeting, and performance across CTV inventory.

The Results: 

• Butler/Till's team never had to leave Claude—the entire process happened within its preferred AI platform while PubMatic's agents handled execution on the backend.

• The campaign generated a media plan in seconds rather than the hours or days typically required for manual planning. 

• Full performance results from the actual campaign will be released in Q1 2026.

Why This Matters: This represents a shift from AI-assisted advertising, where humans prompt AI for suggestions, to more autonomous advertising where AI agents handle strategy, execution, and optimization with minimal human intervention. The use of open agentic protocols like AdCP and MCP suggests the industry is moving toward standardized agent-to-agent communication.

Your Action: Assess AdCP roadmaps for your current ad tech vendors. Push them hard on it. Companies not investing in open protocols may become incompatible with agentic workflows, forcing you to manually manage campaigns while competitors automate. Interoperability is key; getting locked into closed systems will become more expensive as agentic advertising scales and the industry consolidates around open standards.

The Three Superpowers of Agentic Advertising - The group behind the Ad Context Protocol (AdCP) has officially launched AgenticAdvertising.org under the leadership of former IAB CEO Randall Rothenberg and JourneySpark Consulting CEO Matthew Egol. 

DoorDash teams up with OpenAI to offer grocery shopping in ChatGPT🔒- ChatGPT will be able to provide meal or recipe suggestions and handle the shopping in the DoorDash app.

Fluency raises $40M to automate digital ad campaigns with AI - Fluency's tech pulls relevant data, evaluates results, highlights what’s successful, and recommends changes.

Salesforce buys Qualified to add agentic marketing to Agentforce - Qualified's “always-on” AI agents turn websites into conversational experiences that autonomously identify and nurture leads.

New York passes first-in-nation laws requiring disclosure of AI performers in ads - The laws mandate that advertisers disclose when AI-generated people appear in ads and require permission from estates before using a deceased person's image or voice for commercial purposes.

AI has created a new talent paradox in programmatic agencies🔒- AI is replacing the grunt work once used to train agency talent, causing many to worry about the pipeline for future leaders.

Meta is developing a new AI image and video model code-named ‘Mango’🔒- Details are scant, but Mango is expected to debut in the first half of 2026. Image generators are hot because they drive adoption of AI tools and keep users coming back for more. 

The U of Digital AI Newsletter is intended for subscribers, but occasional forwarding is okay!

To subscribe visit Uof.Digital/Newsletters or contact us directly for group subscriptions.

And remember, U of Digital helps teams drive better outcomes through structured education on critical topics in marketing / advertising like AI, ad tech, mar tech, privacy / identity, CTV, commerce media, and more. Interested in learning more about how we can supercharge your team? Let’s talk!