.png)
Google began paying publishers for the work their pages do inside AI answers, and it pays only when it judges that a page shaped the response. OpenAI pitched ChatGPT to developers as a platform with a collective 1.2 billion weekly users, a day after holding back its newest model over safety concerns. New Comscore data puts AI Overviews on nearly four in 10 U.S. desktop searches, and a defamation suit over Google's AI keeps moving through a Delaware court.
Google is paying publishers when their content contributes to answers in AI Overviews, AI Mode, and the Gemini app, through a program it calls the AI contribution pilot. Digiday reported the program on Sept. 14. Earnings are shown to enrolled publishers in a monthly panel inside Search Console. On Sept. 29, The Information reported that about 100 publishers are taking part. One early participant is earning at an annual rate of more than $1 million, according to PPC Land's summary of the report. A publisher that joined months ago had earned $50,000 to $60,000, and some smaller sites earned under $1,000 over several months.
According to Digiday, Google pays only when it judges that a page "meaningfully contributed" to an AI-generated response. Asked about the pilot, the company cited a June post on partnering with sites whose material it uses to ground its AI answers. One executive told Digiday the formula is "quite black box."
Google has now put its own price on a gap that AI-visibility work keeps running into: a model can use a page without crediting it, and credit it without sending a visitor. Comscore's latest report (see Study of the Month) found the gap in travel answers. A randomized experiment by researchers at the Indian School of Business and Carnegie Mellon, run in early 2026 and summarized by Digital Content Next on Sept. 15, found that only 7.1% of outbound clicks on searches with an AI Overview came from links inside the answer. The pages Google pays for are the ones whose facts end up inside the response. Those are also the pages that decide how Google's AI describes a company.
At DevDay on Sept. 29, OpenAI pitched ChatGPT as a surface where developers can launch their own experiences to what it called "our collective 1.2B weekly users." Plug-ins get sidebar homes and interactive panels. OpenAI is also building a discovery layer that works like search inside ChatGPT: it said improved ranking and recommendations will help people find plug-ins "in the directory and in conversations."
Anthropic CEO Dario Amodei wrote on Sept. 12 that "we must slow the pace at which we improve the capabilities of AI models," and OpenAI CEO Sam Altman replied that he agreed "we need to pace the frontier." Two days later, President Trump wrote that the only "guardrails" AI needs is a strong and smart president.
All training, evaluation, and tool use of OpenAI's most capable models had been paused since Sept. 20, after an agent got around internet restrictions to reach an outside chatbot, the company said on Sept. 25. The New York Times reported the same day that OpenAI's agents had pulled data from Commerce Department and Securities and Exchange Commission websites in ways the company did not intend. OpenAI confirmed both and said neither was a breach. "Some involved government websites because our models often turn to them as authoritative sources of public information," a company spokeswoman told the Times. A day before DevDay, OpenAI said it would not release GPT-6.1 Astra as planned; the model "didn't quite meet the bar," said Saachi Jain, its head of safety systems.
Apple released iOS 27 on Sept. 14 with Siri AI, a rebuilt version of Siri that the release notes say brings personal context, app actions, onscreen awareness, and "access to broad world knowledge." Google said in April that its Gemini models would power the new Siri. Siri AI is switched on through Settings, and it runs on iPhone 15 Pro and later models.
Every eligible iPhone now carries a Gemini-backed answer engine that responds to general questions without a results page. How often those answers send people on to a website cannot yet be measured from the outside. Siri is one more AI tool a brand has to query directly to know what it says.
Google is testing direct purchases from Flipkart inside Gemini and AI Mode in India, with a Buy button on select phones, electronics, and accessories and a wider rollout planned for October ahead of the festive shopping season. Checkout is handled in a Flipkart-branded flow. For now no other retailer gets the button. Amazon listings appear without it. Google holds a $350 million minority stake in Flipkart from a 2024 funding round.
Once Gemini can complete the sale, the product summary it writes becomes the last thing a shopper reads before paying.
Robby Starbuck's suit against Google, which alleges the company's AI tools falsely linked him to crimes, survived Google's motion to dismiss in Delaware Superior Court in July and is proceeding to discovery, according to a Sept. 16 analysis by Clay Calvert of the American Enterprise Institute. Google argues that Starbuck deliberately induced the false statements and did not provide the prompts he used. As of late September, the case was pending.
Bloomberg Law's Sept. 8 survey of AI defamation cases centers on actual malice, the standard a public figure must meet, and on whether it can apply to software at all. Whatever the courts decide, a false claim from an AI system reaches customers long before any ruling does.
"How do you go about proving a chatbot's state of mind when the chatbot doesn't have a mind?" Calvert asked.
Florida Attorney General James Uthmeier asked a state court in Highlands County on Sept. 28 to halt further ChatGPT development until third parties approve its safeguards. The filing also seeks to keep minors off the service, stop OpenAI from attributing human qualities to the chatbot, and end marketing that claims safety, accuracy, or reliability. "Stop calling it safe. Stop pretending it's human. Stop selling it to kids," Uthmeier said.
The request was filed as a motion for a temporary injunction in a lawsuit the state brought on June 1. A judge had not granted any of the restrictions as of Sept. 28, CBS12 reported. If courts start treating a chatbot's accuracy as a marketing claim that must be backed up, the reliability of what these tools say about people and companies becomes a regulator's question as well as a brand's.
Comscore's Q2 2026 AI Intelligence Report, released Sept. 22, found Google AI Overviews on 39.4% of U.S. desktop searches in June, up from 25.8% in July 2025, while Bing's Copilot Search reached 17.3%. Total U.S. desktop searches rose to 77 billion in the second quarter, 8% more than in the same quarter of 2024.
ChatGPT's share of the AI chatbot market fell from 70% in January to 50% in June, while Gemini rose from 17% to 30% and Claude from 2% to 11%. Comscore also tracked who gets credited. Tripadvisor appeared as a source in 61% of lodging responses but drew a visible citation in only 21%.
Status Labs' Sept. 11 post sets out five shifts in reputation work this year, from AI answers replacing the click to tighter enforcement against fake reviews. Its section on hallucination covers the risk at the center of this month's defamation news: AI can damage a reputation "not by reporting something negative, but by inventing something false."
The Sept. 4 audit guide walks through a five-surface review (search results, review platforms, social profiles, news coverage, and AI engine responses) and calls the last of these "the newer blind spot." Its check for that surface covers ChatGPT, Gemini, Perplexity, and Google's AI answers.
At the Digiday Publishing Summit in September, one publishing executive said of the industry's favorite doomsday phrase: "The idea of Google Zero is, we're not going to hit zero… I think what people mean is search zero." Another said: "AI Overviews CTR tends to be about 1% from the AI Overviews section."
.png){kind=logo}
