The U.S. Federal Trade Commission and 22 state attorneys general sued Amazon yesterday, alleging the company secretly manipulated its ad auctions to make advertisers pay higher prices. The Wall Street Journal first reported the planned action earlier in the day.
At the center of the case is what Amazon called a “soft reserve”:
- Amazon told advertisers it used a second-price auction, in which the winner generally pays an amount based on the next-highest bid. But beginning in 2019, the FTC alleges, Amazon secretly introduced a soft reserve that could push the winning price higher.
- An internal Amazon document cited by the FTC described the mechanism as an “invented auction participant” that could raise the price even when no competing advertiser had bid that amount.
- The FTC says that by 2024, Sponsored Products advertisers were paying their own bid about 80% of the time rather than a price based on the next-highest bid. An Amazon employee cited in the complaint said the mechanism produced prices “beyond what [can] be achieved through advertiser competition.”
Read:
- FTC to File Lawsuit Alleging Amazon Deceived Advertisers (August 31) – The Wall Street Journal (subscription)
- FTC, States Sue Amazon Over Secret Ad Surcharge Scheme (August 31) – FTC[dot]gov
- Amazon’s response to the FTC’s lawsuit regarding Sponsored Ads (August 31) – Amazon
From tipsheet: What happens when AI optimizes both sides of an auction whose rules are set by the seller? The auction could start to look less like price discovery and more like a negotiation between machines over who captures the value in the transaction. Or are those the same thing in an autonomous advertising market?
SELL-SIDE
GAM publisher data co-op emerges
Google Ad Manager has introduced a publisher data co-op that lets publishers contribute aggregated, anonymized network and audience data to help build machine-learning models predicting broader user interests and behaviors.
In return, participating publishers can access additional monetization features, agency demand and, eventually, audience-extension lookalikes to find off-property users with behaviors similar to their first-party audience segments.
Spotted in the wild: New Google publisher data co-op (August 31) – Consultant Nathan Thomas on LinkedIn
More:
- Participate in the Google publisher data co-op – Google Ad Manager Help
- What’s new in Google Ad Manager (August 24) – Google Ad Manager Help
From tipsheet: Google is turning Ad Manager into a learning layer across publishers. The co-op gives GAM a new role: converting pooled publisher signal into models, demand and audience extension. That is notable for a product that could still face structural remedies in Google’s ad tech antitrust case.
LLMs & CHATBOTS
OpenAI confirms $1 billion advertising run rate
OpenAI confirmed in a blog post yesterday that ChatGPT Ads has achieved an annualized revenue run rate of $1 billion.
OpenAI CTO of Applications Vijaye Raji said on LinkedIn yesterday:
“0->$1b in less than 200 days.
Our fastest product ever to reach this milestone, and one of the fastest-growing ads businesses in the world. It was just a few months ago here I announced the launch of self-serve and the roll out of CPC bidding.
20K+ advertisers. 40+ countries. And we are just getting started. Don’t blink…”
Read more on LinkedIn. (August 31)
OpenAI said the self-serve business launched in May now represents a “material share” of the business and that there are more than 50 technology and measurement partners.
Read: A milestone in expanding access to AI (August 31) – OpenAI
Given the momentum, we recently postulated that ChatGPT Ads revenue has been growing approximately 25% every two weeks since launch. If it continues on that trajectory, OpenAI could generate roughly $1.6 billion in full-year advertising revenue this year.
“Show me the money!”
In preparation for an IPO, OpenAI must prove the opportunity ahead is real. The challenge may be explaining to investors why that trajectory remains short of the $2.5 billion reportedly projected for 2026, let alone the $100 billion goal for 2030 that was part of the same report.
From tipsheet: If OpenAI wants to accelerate advertising even more, it will eventually need an ad product that can reach its highly valuable professional and business users. Right now, Google can reach this cohort through Search, including AI Overviews and AI Mode. OpenAI cannot.
SELL-SIDE
Google’s AI Overviews and the publisher bargain
Digital Content Next CEO Jason Kint reported last week that Judge Amit Mehta had sharp words for Google during a motion-to-dismiss hearing in the Penske Media v. Google case over Google’s AI Overviews.
Kint wrote on X:
“The core issue: can Google use its adjudicated search monopoly to force publishers to provide content for AI uses at zero cost? Some notable exchanges today….
Google attorneys repeatedly characterized AI Overviews as a ‘product improvement’ and called the historic traffic-for-crawling bargain with publishers a ‘vague historical course of dealing.’!!! GTFO. Mehta pushed back: the situation ‘all seems really unfair.’
Mehta noted publishers lack control over how Google uses their content and described Google’s ‘product improvement’ as being built ‘on the backs of the publishers.’ Importantly, he said product improvements ‘are not immune from antitrust scrutiny.’”
Read the entire thread. (August 25)
Commenting on the proceedings as described by Kint, sell-side advocate Matthew Scott Goldstein said on LinkedIn: “Google is the largest and most effective scraper in the world, and its advantage was never better technology. It was a bargain publishers signed when the trade was traffic, being enforced now that the trade is gone.” Read more. (August 31)
For the latest: Penske Media Corporation v. Google LLC – CourtListener
From tipsheet: The question is shifting from whether AI Overviews reduce publisher traffic to whether Google can unilaterally rewrite the underlying bargain: use publisher content to answer the query while sending less traffic back.
Google has already collided with the UK’s Competition and Markets Authority over a similar issue. In the UK, Google introduced a publisher control option that lets sites decide whether their content can appear in and ground AI Overviews and AI Mode. As of yesterday, Google said it has rolled out that control globally. Read tipsheet. (June 4)
AGENCIES
WPP resurgence?
Digiday’s senior reporter Sam Bradley writes:
“Two thirds of the way through the year, per COMvergence data shared with Digiday, WPP Media has won just over $3 billion in fresh accounts, by estimated media value, while accounts lost totaled $2.81 billion, with a retention rate of 43%. That’s a favorable contrast with last year’s 16% retention and lost accounts equalling $6.98 billion.”
Read: As WPP’s Cindy Rose marks a year as CEO, turnaround is underway but far from certain (August 31) – Digiday
From tipsheet: WPP has been pitching WPP Open as the AI operating layer behind its turnaround. The new business numbers may be the first evidence that the agency holding company’s pitch is gaining traction with clients.
LLMs & CHATBOTS
Anthropomorphizing the chatbot
The chatbot gets a job. And apparently gets hungry.
- Workflow: AI Bots Are Now Your Colleagues, Thanks To Optimizely’s Latest Launch (August 31) – AdExchanger
- Prompt: Snickers Says Even Your AI Chatbot Isn’t Itself When It’s Hungry (August 31) – Adweek
MARKETERS
Marketing the AI opportunity
Ad Age senior editor of technology and AI Garett Sloane explores the challenge of marketing AI in his latest article.
Previewing it on LinkedIn, Sloane writes: “While consumers have their reservations, there is another battlefront for enterprise hearts that Anthropic and OpenAI are contending for. And the enterprise-consumer balance is one of the toughest to get right.”
Read more. (August 31)
Read: Why AI CMOs have a problem that goes beyond advertising (August 31) – Ad Age (subscription)
From tipsheet: Anthropic’s Claude has roughly 6% of the monthly active app users that ChatGPT does, according to Sensor Tower estimates published in June. Arguably, Anthropic has the easier marketing challenge because nearly all of its revenue is enterprise-related. It doesn’t have to make Claude mean one thing to consumers and another to businesses at ChatGPT’s scale.
LLMs & CHATBOTS
Developments
- Apple’s John Ternus Takes the Reins as CEO, With AI as Job No. 1 (August 31) – Mark Gurman in Bloomberg (subscription)
- Commission designates ChatGPT, Reddit, Roblox under Digital Services Act (August 30) – European Commission
- “muse code is out of beta. launching with an sdk in developer preview to build your own agents on top of it and rolling out monthly subscription plans.” (August 31) – Alexander Wang, Chief AI Officer, Meta on X
MORE
- New Netflix paper: RecSys, content concentration, and promoting the “middle tail” (August 31) – Analyst Eric Seufert on Mobile Dev Memo (subscription)
- Google does GEO: Google Search Console AI performance reports and Search generative AI control rolling out globally (August 31) – Search Engine Land
- Agentic: LG CNS Targets U.S. Ad Market with Agentic AI-Powered Marketing Solution (August 31) – press release