Google tests AI summaries inside business knowledge panels, and a judge rules AI Overviews can defame.

The knowledge panel for a local business shows an "AI Overview" label with a "Show more" button that opens an AI Mode style chat. Schwartz reproduced it on his own business and found the description of its services somewhat outdated. He says Google "has been testing variations of this for a while now." There is no Google statement, and this is a test, not a rollout.

An author alleges AI Overviews said he was serving a life sentence for murder. The judge partly denied Google's motion to dismiss, writing that "an ordinary user searching 'Jimmy Keene net worth' on Google is seeking factual information, not AI work product." He held actual malice plausibly pleaded because Keene "directly notified Google on three occasions" and similar statements kept appearing. One statement was dismissed as substantially true, another with leave to amend. This is a ruling on a motion, not a verdict.

After clicking a ChatGPT ad, a user can open "a clearly labeled conversation with a business-sponsored agent." OpenAI states: "The conversation with a Sponsored Agent is distinct from ChatGPT's independent answers and separate from the original conversation." Sponsored Agents are a test "with select advertisers in the United States." Separately, businesses in HubSpot can now connect a ChatGPT Ads account and "create ads, track performance, and follow up on leads", and a Shopify app goes international on 23 September.

Method: "the same thousands of questions every day through July and August 2026" on seven engines (ChatGPT, AI Mode, AI Overviews, Copilot, Claude, Perplexity, Gemini), US only, 6,095,520 citations. "Company owned pages take 53.7% of all citations." ChatGPT's company-owned share went "44.38% before the break, 55.28% after" the 8 August change, while Reddit fell from 4.42% to 0.39%. Limits: vendor data, and none of the 17 industries is coaching, professional services or local services.

Method: 904 real brands observed across ChatGPT, Gemini, AI Overviews, AI Mode and Copilot, plus a lab test of 80 matched name pairs on four models. "A company whose name scores one point higher is described 0.07 points more positively." In the lab, models sometimes described the dictionary meaning of the name instead of the company, "one description out of twelve" on GPT 5.6 Terra. For known brands, "it disappears." Limits: correlational field data, lab prompts without retrieval.