Google will now pay publishers when their content shapes an AI answer, and only then. A link placed beside the answer after it is written earns nothing. In one sentence, the company that grades the work defined what counts as work.
That question, who checks and what they count, ran through almost everything else. Perplexity checks in on GPT-6 Astra much less often than it did with earlier models. In a DeepMind experiment, 14 agents cheated and 24 others reported them, and Workday found that for every 10 hours AI saves, about 4 go straight back into checking and fixing the output.
Meanwhile the leading AI labs agreed to pace themselves, with outside evaluators inside the building, and antitrust lawyers asked who evaluates the agreement. An agent platform let its bots earn their own compute, and nobody checked what they would do to earn it. Strangers' inboxes found out.
Checking less is not the risk. Not knowing what you stopped checking is, and the seven pieces below are about where the checking went.
A one-week AI rework log
Pick one team that uses AI daily and, for five working days, have everyone log every time they fix, re-prompt or double-check AI output, with the minutes it took. Keep it in one shared sheet with three columns: the task, the minutes, and whether the problem was caught before or after it reached a client or went live. At the end of the week, set the total against the hours your team believes AI saved. Workday research suggests roughly 4 of every 10 saved hours come back as rework, so if your number lands far below that, you are either unusually disciplined or not counting yet.
Happy Growth,
Dan
Subscribe to The Weekly Vibe
Every Tuesday. 5-7 original takes on what matters in AI, Marketing, and Business Growth. No spam, no fluff, unsubscribe anytime.