AI Briefing: Pinterest, Microsoft and Google bring new generative ad features for social and search

This week has brought another flurry of AI-related ad news as major players like Microsoft and Google, and even social platform Pinterest roll out more ways for advertisers to create and buy ads across social, search and chatbots.

On Tuesday, Pinterest debuted new ways for advertisers to use generative AI to generate ad creative and automate media buying through its recently unveiled Pinterest Performance+ platform. One new tool lets advertisers turn blank backgrounds into “lifestyle imagery” to enhance a brand’s products. Alpha program advertiser Walgreens tested the tool for 50 products — including a perfume bottle ad that appears to sit on a showroom table — to drive 55% higher click-through rates.

First announced in summer 2024, the feature is now rolling out to more advertisers that want to test new ad formats powered by Pinterest’s generative AI foundation model. By training its model on Pinterest’s visual data set, the company aims to give advertisers ways to create ads that feel native to the platform and personalized for users’ interests and tastes. The updates also build on Pinterest’s multi-year strategy to use AI for improving both user personalization and advertiser performance, said Matt Crystal, Pinterest’s vp of product. 

“The ability to say we know [one person] is a mid century modern guy and [another person] is a maximalist is because you’ve expressed that to us via your behavior on the platform,” Crystal told Digiday, describing it as a special sauce of sorts. “The things you say, the boards you’ve created, etc. For me, that’s what’s really exciting as we think about this personalization vision.”

Pinterest’s updates are just the latest in the steady stream of ad-related news from tech giants. In recent months, online platforms have rolled out a number of new advertising features for using AI machine learning and generative AI models. For example, Microsoft this week released new ad tools for Copilot, Bing and Edge. A day later, Google expanded ads for its AI Overviews search summaries.

Naturally, advertisers see the benefits of AI-powered tools for ad creation and automated media-buying. However, some also think a lack of control could be a major obstacle to adoption. Calla Murphy, senior vice president of digital strategy and integrated media at Belardi Wong, said platforms should give advertisers more control over where dollars are going related to customer acquisition, site visits and existing customers. While brands are open to testing ad-creation tools, she’s seeing clients limit tests to feed-based ads like dynamic product ads. 

“I actually think some of our larger brand clients will probably not be first adopters, because they have that aspect of brand control,” Murphy said. “And it will really impact smaller advertisers who maybe will give up brand control for rapid creative generation.”

With so many recent updates, here’s a lay-of-the-land with highlights from Pinterest, Microsoft, Google, Meta, TikTok, Snap and Amazon.

Pinterest

Along with its generative AI updates, Pinterest also officially introduced its automated ad-buying Pinterest Performance+ platform. The company hopes to drive more lower-funnel performance while decreasing campaign creation time and costs. Rather than optimizing bids based on traditional metrics like conversion and clicks, Pinterest will let advertisers optimize bids based on return on ad spend (ROAS).

When asked how AI-generated ads perform versus ads created by agencies and brands, Crystal said generative AI helps advertisers scale their ad campaigns based on their product catalogs rather than relying just on humans.

“Creating thousands of that former image is really expensive and really time consuming,” he said. “Versus [saying] ‘I already have a million SKUs in my product catalog, I already have photos of each of these, how can you help me make these assets I already have more engaging?’”

Microsoft 

On Wednesday, Microsoft added more AI tools for advertisers creating ad campaigns while also overhauling how Copilot users will engage with ads within a chatbot conversation. Unveiling a new “streamlined” ad experience, Microsoft plans to serve ads in a separate section below Copilot’s organic answers, while a new “ad voice” feature will aim to make ads feel more natural. Ad targets within a Copilot chat will also be based on an entire conversation rather than only based on the user’s previous prompt.

As for advertiser tools, Microsoft introduced new ways for marketers to use Copilot for creating images, videos, banners and text marketing messages. A new “Diagnostics” tool will help advertisers inspect campaigns during set-up before they go live while a new “Performance Snapshot” tool will help show key insights and trends. Microsoft has also expanded the new AI advertising features beyond Copilot to Bing search and its Edge browser.

Meta

Meta’s already rolled out features powered by its Llama family of AI models, including tools for image variation, text overlay, image expansion, text generation and background generation. Smaller advertisers have used Meta’s AI for expanding images. Larger advertisers including e-commerce companies have been testing Meta’s ad background generator to help highlight products. Other AI tools have focused on automation in areas like its Advantage+ platform.

Although its recent Meta Connect conference didn’t focus on advertising, Meta said at least 1 million advertisers have used various generative AI tools, and that more than 15 million ads were created with generative AI in the past month. Having a lot of assets can help improve AI tools to create on-brand content. However, Meta vp of Product Management Tak Yan said the number of assets an advertiser needs is “quite minimal,” which lets even smaller advertisers leverage what they have.

In an interview last month, Yan said the company is working to incorporate brand tone, voice and other brand guidelines beyond existing features for color schemes and text overlay. Meta’s also looking at ways to create ads based on info on a company’s website and expanding tools for video ads and other ad formats.

“Another aspect of this would be how we can put Gen AI and consumer interactivity together,” Yan told Digiday. “That might include giving consumers control on the experience that they have with ads. And then [with] AI, we can produce interactive, personalized experiences, like with format changes, new affordances for them to control the [ad] creative they ultimately see.”

Reddit 

Over the past few years, Reddit also has acquired several startups — including MeaningCloud, Spiketrap, Spell, and Oterlu — to help expand capabilities for its ads manager and other business tools. Another recent acquisition is Memorable, an AI startup that helps advertisers see which creative performs the best on Reddit.  

Along with building tools to help advertisers automate bidding, Reddit has rolled out AI tools to help forecasting, improved keyword targeting and other campaign analysis features. Another AI tool helps advertisers generate headlines when marketing the same product to various audiences. Other tools include prompt templates to help them find the right audiences using generative AI, gauge a brand’s share of voice, and understand brand sentiment across the platform.

“Maybe you and I are both interested in a credit card,” posed Roelof van Zwol, Reddit’s vp of ads engineering, in a Digiday interview last month. “Maybe you’re looking because you want to get extra points or benefits in another way, like the concierge services. And I might be looking at it because I’m not willing to pay any fees. If you want to generate headlines, then it would be great if you can personalize the message to different intents that we’ve identified people coming to Reddit.”

Along with using its data set to build its own generative AI tools, Reddit also is partnering with companies like Google and OpenAI. While Reddit licenses its data to Google to help improve model training, a Reddit deal with OpenAI includes bringing AI tools to Reddit and bringing Reddit content to ChatGPT.

Other platforms with recently added AI tools

  • Google on Thursday announced a broader expansion of ads within AI-generated search summaries (AI Overviews) along with some changes to how they appear within the new real estate. The updates follow the debut of ads for for its AI Overviews earlier this year. This week’s updates follow other ad product rollouts over the past year powered by Google’s Gemini family of AI models across Search, YouTube, Gmail and other areas. Other features previously added include ways to generate new audience segments, a chat-based buying tool for search ads, visual tools for Demand Gen campaigns, and features for shopping ads.
  • TikTok‘s suite of AI tools include its creative AI suite called Symphony, which it introduced this spring. Other tools include ways to make avatar influencers and a partnership with Adobe to let advertisers create AI-generated content using TikTok’s library of licensed music.
  • Snap has developed ways to use generative AI to create augmented reality ads with custom AR lenses. Snap also has another feature to help AR creators make custom machine learning models and assets for their lenses, which Snap said can speed up development time by weeks or months. Other tools include a GenAI Copy Generator for writing headlines and translating text into other languages for global campaign 
  • Amazon also has a number of new AI tools for advertisers. Last month, the e-commerce giant debuted a new ad-only video generator. It’s also released tools for creating AI-generated images and has recently started bringing ads to Rufus, its chat-based shopping assistant.

Prompts and Products — Other AI news and announcements from this week

  • Microsoft debuted major updates for its Copilot platform including new AI voice and vision capabilities.
  • In the process of molting out of its nonprofit status, OpenAI announced another major $6.6 billion fundraising round, new developer tools and a new feature called Canvas to help users write and code. Accenture and Nvidia announced a new deal to help drive enterprise adoption of AI.
  • Twilio has an integration with OpenAI’s real-time API to help build voice-based AI applications.
  • Soul Machines debuted more features for its AI avatars including object recognition and the ability to remember conversations.
  • Politico announced a new partnership with Y Combinator to provide Politico Pro subscribers with AI-generated summaries about policy news, legislative updates and more.
  • After signing more than a dozen new AI laws, California Governor Gavin Newsom vetoed an AI safety bill that he said could hinder innovation. Meanwhile, a newly passed law banning election-related AI deep fakes was blocked by a California judge who said the law would be unconstitutional by threatening free speech. 

Other AI-related stories and reports from Digiday and beyond

  • AppliedXL and Associated Press team up to use AI to sift through federal regulations (Fast Company)
  • Meta confirms it may train its AI on any image you ask Ray-Ban Meta AI to analyze (Techcrunch)
  • How The New York Times Incorporates Editorial Judgement in Algorithms to Curate Home Screen Content (Medium)
  • Americans’ Views Mixed on Tech’s Role in Politics (Substack)
  • People are skeptical of headlines labeled as AI-generated, even if true or human-made, because they assume full AI automation (Nexus)
  • College students used Meta’s smart glasses to dox people in real time (Verge)
  • Media Briefing: How Dow Jones is developing an AI model to help its planning team respond to advertisers’ RFPs (Digiday)
  • What happens in the Google ad tech antitrust trial now that testimony is done? (Digiday)
https://digiday.com/?p=557168

More in Media Buying

Monks’ head of social sees this as the biggest challenge for marketers in the creator economy

Monks’ Amy Luca shares insights on balancing timely and timeless marketing in the creator economy.

How influencer shops and agencies are adding content studios to boost production speed, revenue

Influencer shops expand content studios to boost production speed and drive additional revenue streams. Depending on who you talk to, they get there in different ways.

AI Briefing: Index Exchange and Cognitiv integrate to use generative AI for programmatic curation

The companies are the latest to bring deep learning and large language models to media buying tools.