DoorDash, Instacart, and Uber Eats each rewrote their search stacks around LLMs, but chose wildly different architectures. The post shows how each company’s existing infrastructure dictated whether the LLM sits shallowly as a re‑ranker or deeply as a full‑text engine, offering a template for any AI‑powered product.
Nike and Lululemon are being sued in California for listing fictitious “regular” prices and striking them through to create deceptive “phantom discounts.” Plaintiffs say the practice violates the state’s False Advertising Law, and a win could force a sweeping change in how online retailers display discounts, exposing brands to costly settlements.
Amazon's ad unit generated $76 billion in the trailing twelve months, with Q2 alone delivering $20 billion, a 26% YoY jump. The surge is fueled by new AI‑powered Sponsored AI Prompts and strong performance of its core Sponsored Products. This places Amazon’s ad platform solidly alongside Google’s as a major revenue engine.
Version 2.3 of the Agentic Advertising Management Protocol adds enterprise‑grade deployment, built‑in governance, OAuth authentication, and integrations with OpenAI, Google Gemini and major ad platforms. Those upgrades let AI buying agents run at scale inside existing ad tech stacks, turning experimental bots into trusted, auditable production tools for advertisers.
Metricool’s analysis of nearly 800k long‑form videos shows average views per post rose 76% year‑over‑year, while average watch time dropped 37% and estimated ad revenue per post fell from $2.65 to $1.20. Creators are getting more eyeballs but fewer monetizable moments, reshaping how success is measured.
Facing steep AI compute token bills, ad agencies are treating tokens like media inventory, buying them in bulk ahead of demand and adding a margin. Proponents say it hedges rising costs and keeps profit margins, while critics warn it turns agencies into speculative futures markets.
ClickUp slashed 22 % of its workforce and announced cash salary bands up to $1 million for employees who deliver “100× impact” with AI agents. The move signals that senior marketing leaders, even a solo CMO, could command seven‑figure pay if they exploit the new AI‑first model.
Seven experiments and over 3,100 eBay sales show consumers are 27% more likely to buy and willing to pay up to 72% more for items with accidental flaws. The research suggests brands should own and highlight production mistakes, except when product function is critical, to turn errors into a unique selling point.
You’re caught up.
That was all 8 stories in the 31 July issue.