Google is testing payments for content contributing to AI answers, Mueller says AI position data is hard to report, and Cloudflare splits training from search.

Zero-click search is a query that ends without a click to any result. How panels measure the rate, why estimates run from 22% to 68%, and who disputes them.

Sales teams lost embedded CRM profiles on February 1, 2025 unless they bought a higher tier, the complaint says. Treble damages sought; LinkedIn yet to reply.

89% of surveyed investors want marketing capitalised, yet IAS 38 forces it into expenses. Why do finance chiefs still treat brand as the easiest line to trim?

Qwen2.5-72B lost 6.9 points on average across 11 benchmarks, a cost for teams feeding customer text into LLMs. Why does redaction score even worse than tags?

Frankfurt judges held that Meta's ad auction and feed ranking amount to control over content, removing hosting protection. Which ranked platforms come next?

Deal drops land at midnight, 8 a.m. and 1 p.m. PDT over 48 hours, days after sellers locked in stock. How could Alexa price alerts change ad auctions this fall?

Average hourly pay nears $24 for US core operations staff, who also get a 10% grocery discount from October 1. When does the banking perk arrive, and for whom?

Judge George Wu's 16-page tentative ruling keeps TikTok US under reporting duties through 2029. A Monday hearing will test the last quarter of a $400m deal.

Frankfurt ruled Meta's ad auction ends its hosting defence at 250,000 euros per breach, LinkedIn faces antitrust class action, and streamers built a D.C. lobby.

Sixteen terms, four groups of four. Four working vocabularies that rarely meet, and one group that looks settled until the fourth tile refuses.

Programmers lose a DMCA theory worth up to $25,000 per violation as judges find AI code tools make new works. What survives for publishers suing over training?

Microsoft staff called AI training an astonishing theft while OpenAI trained on 3.9 million Times copies. Can fair use survive the September 17, 2026 record?

Onsite auditors inside the largest AI labs and a May 2027 rulebook for verifiers are on the table. Can an order with no legal force change what labs build?

Buyers in the US, UK and Canada upload one-column CSVs of up to 100,000 sites or app IDs, with no spend data. What do vendor-built AU benchmarks leave out?

Few-shot prompting places a handful of worked examples inside the input to a language model so it infers a task without retraining. Mechanics, limits, disputes.

Pseudonymisation swaps identifiers for substitute values reversible only with separately held data. How it works in ad tech and why it remains personal data.

Hallucination is AI output that sounds confident but is false or unsupported by any source. How models produce it, how it is measured and where liability lands.

Sponsored Products: cost-per-click ads that promote a single catalogue item inside retail search results. How the auction, bids, placements and disputes work.

A reference price is the benchmark a shopper uses to judge an offer, usually the struck-through "was" figure displayed beside a discount in ads and listings.