Meta’s AI holiday ad pitch

John Ebbert

Holiday plans

LLMs & CHATBOTS

Advantage+ for the holidays

Meta released its Holiday Planning Playbooks for the October to January time periods.

With the fully-automated, AI ad future predicted by Meta CEO Mark Zuckerberg only a year away, Meta wants clients to get with the AI “program” and is pitching its AI ad system, Advantage+.

Here’s a screenshot of part of the pitch—or download it at the bottom right of the page:

Within the marketing literature, Meta touted its unique “Opportunity score” optimization capability which is defined as:

“On a 0-100 point scale, opportunity score represents how optimized your campaigns, ad sets and ads are. Your score is based on how many Meta Ads Manager recommendations you apply, and those recommendations are ranked based on estimated performance impact..”

Read more about Opportunity Score here which has been around for about a year. Meta appears to be trying to deliver every bit of signal—including recommendations—it can to the marketer. In this case, how good is the opportunity to sell a particular product – outcomes-based pricing is getting closer.

There’s also a bit of evaluation of the marketer potentially going on with Opportunity Score as a Redditor noted. (Google has something similar called “Optimization Score.”) Platforms are looking to optimize the client.

Overall, these “AI updates” don’t seem huge, rather more of an evolution that may have been partially announced at Cannes in June.

Search Engine Land breaks down the 4 categories of AI-powered tools in the holiday package:

“Creator commerce: ‘Advertisers can now merge creator content with catalog ads and even feature testimonials within partnership ads.’

Reels upgrades: ‘Brands can turn single media ads into shoppable formats, add catalog video carousels on Instagram Reels, and surface more product info’

Creative diversification: ‘Flexible media lets AI repurpose assets across placements, boosting conversions (…) New creative insights and related media tools give advertisers recommendations to fight fatigue and expand reach.’

Omnichannel and Shops: ‘Omnichannel ads, now globally available, optimize for purchases online or in-store’“

Read more from Search Engine Land. (August 21)

LLMs & CHATBOTS

AI Mode reservations, but no ads

Robby Stein, VP of Product, Google Search, returned to the Google “The Keyword” blog to make a two-part announcement:

  • New agentic capabilities and personalized responses to AI Mode in Search
  • AI Mode expands globally to more than 180 countries.

The key announcement is that AI Mode now provides the long-sought ‘Holy Grail’ agentic use case: making dinner reservations. AI Mode will coordinate the many moving parts of a reservation for you. Initially, it’s only available through Google AI Ultra which will run you $125 per month. And/or, you’ll need to be connected to the “‘Agentic capabilities in AI Mode’ experiment” in Google Labs.

Mr. Stein explained the “under-the-hood” elements of reservations with AI Mode:

“AI Mode uses the live web browsing capabilities of Project Mariner, direct partner integrations on Search, and the power of our Knowledge Graph and Google Maps to help users take action on the web.”

Sharing and personalization is also part of the AI Mode mix, said Stein.

There was no mention of any advertising opportunities – yet.

Read more on The Keyword. (August 21)

Related: Our approach to energy innovation and AI’s environmental footprint – Google The Keyword blog (August 21)

BRANDS

AI data-readiness checklist

“If I were in the marketer’s hot seat, here’s how I’d prepare an AI data-readiness checklist:

Audit your data: Identify what first-party data you already collect across web, apps, CRM, surveys, and transactions. Identify gaps in accuracy, consistency, and completeness. This is not an easy task, and you will need buy-in and help from teams outside your own.

Break down silos: Fragmented data limits AI performance. Use tools or platforms that connect IDs across devices, domains, and environments (for example, web + CTV + mobile)

Do some cleaning: Standardise formats, remove duplicates, and validate entries. No AI system can compensate for poor data hygiene. If it’s messy going in, it’s messy coming out…”

Christopher Hogg, CRO, Lotame (August 20)

Read the other four on the checklist:

AGENCIES

AI, data and the holding company

Though not AI-explicitly, sell-side platform Magnite announced a partnership with data platform Acxiom which has implications for AI-enabled strategies, where first-party data is the key differentiator. Read the press release.

Stepping back a bit, Acxiom is owned by ad agency holding company Interpublic Group, which is about to merge with ad agency holding company Omnicom.

So, this is a deal between Magnite and Omnicom ultimately.

But, as often happens in Acxiom partnerships (like this), Acxiom keeps IPG at arm’s length. In the sensitive area of data management, no one on the Acxiom-IPG side wants to give any sense of Acxiom sharing its clients’ data within the holding company.

SELL-SIDE

Fastly challenges LLMs, Cloudflare

Content delivery network Fastly is taking a page from Cloudflare’s book (from early July) and calling out AI crawlers that are straining the web and its publishers.

Read the press release. (August 19)

Key takeaways from the Fastly report included:

“1. AI bots can place significant strain on unprotected web infrastructure, with peak traffic reaching up to 39,000 requests per minute.

2. Commerce, media & entertainment, and high-tech sectors face the highest levels of scraping for training AI models.

3. AI crawlers account for 80% of AI bot traffic, with roughly half of that attributed to Meta alone. In contrast, AI fetchers make up the remaining 20% of the traffic.

4. ChatGPT generates the most real-time traffic to websites, with 98% of fetcher bot requests attributable to OpenAI’s bots.

5. Most AI models are trained predominantly on content originating from North America, shaping their alignment according to the region’s cultural and geopolitical perspectives”

The Register reported on the findings and noted, “Facebook owner Meta’s AI division accounts for more than half of those crawlers, while OpenAI accounts for the overwhelming majority of on-demand fetch requests.” Read the Register’s summary. (August 21)

Read the Fastly Threat report here (PDF).

SELL-SIDE

AI licensing deals FTW

Australian Associated Press (AAP) announced on Tuesday that it has come to agreement with Google and will provide news content to Google’s generative AI-enabled Gemini app. Read the press release. (August 19)

Australia’s The Sun reported that financial terms between the AAP and Google were not disclosed. Read that one. (August 19)

Search referral traffic on the open web and the ad revenue that publishers earned from it are being slowly replaced by content licensing deals with AI companies.

And licensing continues to be the chosen method for large publishers to do deals with AI companies despite recent discussion of usage-based pricing purportedly coming from answer engine Perplexity this week (the same company that was buying the Chrome browser last week.).

A selection of recent, public licensing deals include:

SELL-SIDE

The Reddit “double-dip”

Mobile Dev Memo analyst Eric Seufert mused about Reddit on Wednesday saying that the publisher, which is active with AI licensing deals with LLM companies, is “double-dipping.” And, that’s a good thing.

He explained:

“Reddit is double-dipping: it is monetizing its vast corpus of data through data licensing deals while also utilizing it with a native chatbot that, in success, will increase engagement, drive impressions growth, and contribute to advertising revenue growth. Reddit Answers can be seen as an engagement-boosting strategy, analogous in mission and purpose to Meta’s pivot to the open graph and short-form video and Google’s introduction of AI Overviews and AI Mode (covered in my Google’s Gambit series: parts one, two, and three). In this way, Reddit’s data serves as the engine to two high-margin revenue streams.”

Eric Seufert, Mobile Dev Memo

Mr. Seufert said Reddit’s ‘double dip’ sets it apart—in a positive way—from other publishers of similar size.

Read more on Mobile Dev Memo. (August 20 – subscription)

TECH

Charts: When AI Search ads explode

eMarketer Chief Content Officer Daniell Zia Wigder provided 5 charts on her LinkedIn profile that may be of interest to the ad industry. She wrote, “From Reddit, Inc. to YouTube, AI, ecommerce penetration by category, and a ranking of global ad spend titans, these charts give a sense of where opportunities lie.”

Here are two of the charts below beginning with AI search. Considering there are few opportunities to advertise in an AI chatbot these days, AI search is poised to explode. eMarketer’s chart confirms it:

The following chart from eMarketer’s Wigder would appear to be good guidance on what industries may hold the greatest opportunity for AI considering their lack of penetration – Food & Beverage and Autos.

See the other three charts on Ms. Wigder’s LinkedIn profile. (August 21)

TECH

Incrementality is the way

Eric Tilbury, Senior Director of Ad Operations and Solutions Engineering at AI ad tech company Inuvo, shared his frustrations on LinkedIn with CPA optimization in programmatic.

Mr. Tilbury wrote:

“Too often, you see buyers try to get that CPA as low as possible to replicate walled garden performance numbers. This de-optimizes what you’re trying to do in programmatic, and it’s really to drive incrementality. Convince someone to consider your brand that would not have done so without seeing an ad.”

Incrementality attribution continues its march in the age of AI and “last click” is only a potential last step—if there is a click. Read more on LinkedIn. (August 20)

Related: “Maximize the performance of your Google Ads conversion data” (August 20) – Google Ads Developer blog

PROMPT

How would you build a customer data platform to be used five years from now?

Response from OpenAI’s ChatGPT:

If you’re building a customer data platform (CDP) for five years from now, you’ll want to design for where customer expectations, privacy regulation, and AI-powered marketing are heading. Here’s a structured way to think about it:

  1. Core Architecture

  • Unified Data Layer: Real-time ingestion from CRM, POS, web, app, IoT, CTV, and AI agents (shopping/voice assistants). Schema-flexible to handle new interaction modes.
  • Event-Stream Foundation: Kafka-like event processing with automatic enrichment (e.g., geo, context, intent).
  • Composable Microservices: Each function (identity resolution, segmentation, consent management) is modular to swap in/out as tech evolves…

Read more on tipsheet.

MORE