Google just made you liable for whatever its AI does in your ad account

Feels like a "show your working" kind of week. Google's quietly handing you the liability for whatever its AI decides to change in your ad accounts, a data firm's finally put a real number on the ChatGPT ad spend everyone's been guessing at, and there's an actually useful checklist for why a client's site isn't showing up in ChatGPT yet (spoiler: it's usually just a broken robots.txt). Meanwhile, two entirely separate studies land on the same point: audiences can tell when nobody bothered to actually make the thing. Bit of a theme, that. Here are your five.

1. Google just moved AI ad liability onto you, starting 1 July

Google's updated Google Ads Terms of Service landed on 1 July, and the headline change is blunt: if a Gemini-powered tool generates or edits an ad, a target, or a landing page inside your account, you're on the hook for it, not Google. It arrives alongside a bigger push into "agentic" advertising unveiled at Google Marketing Live, where AI Mode increasingly behaves as a layer sitting across every campaign you run. Worth flagging to any client who's been letting automatically created assets or broad match run on autopilot: "the AI did it" stopped being a valid excuse the moment this ToS went live. Google keeps the system authority, you keep the legal liability, and that's not a typo. Worth building a proper review step into any account using AI Max or automatically generated assets, someone actually checking what's been created before it goes live, because that's now explicitly your job, not Google's. Search Engine Land

2. Someone finally has real numbers on how much money is moving through ChatGPT ads

Guideline has extended its Ad Intelligence dataset to cover verified, transaction-level ad spend on AI platforms including ChatGPT and Perplexity, representing roughly $200 billion in annual media investment across 65 countries. Unlike platform-reported figures or survey estimates, this is sourced straight from transactions, a ground-truth read on a channel that's mostly been guesswork so far. Worth knowing before the next new-business pitch that leans on "AI search is the future": you can now point to independently verified spend data instead of a platform's own rosy self-reporting, which matters when a client asks whether ChatGPT ads are worth the budget or just the hype. Worth a bookmark for benchmarking conversations, especially with any client sizing up a first AI-platform ad budget against what everyone else in their category is actually spending, not what OpenAI says they're spending. PR Newswire

3. 97% of marketers use AI daily. 78% of their customers wish they hadn't.

Canva's 2026 State of Marketing and AI report lands on an awkward number: 97% of marketing leaders use AI in their daily creative work and 99% plan to spend more on it this year, but 78% of consumers say they'd rather see an advert made by a person, even if the AI version would technically be better. 87% believe the best advertising still needs a human touch, and mentions of "AI slop", the low-effort, obviously machine-made stuff, are up ninefold in media monitoring. This one's worth sitting with longer than a headline: the tension isn't AI versus no AI, 68% of consumers don't actually mind AI in their ads if it makes them more useful. It's AI without anyone visibly steering it. The moment work stops sounding like it came from a person, trust goes with it. Worth building into any client's AI workflow: a visible human editing pass, a disclosed AI-use policy (74% say that alone would make them more comfortable), and someone willing to kill the output that reads like nobody touched it. The differentiator in 2026 isn't using AI, it's not sounding like everyone else who is. Canva Newsroom

4. The nonprofit newsroom outgrowing everyone on social, and it's not because of a trend

Rachel Karten's Link in Bio newsletter sat down with Georgia Parke, Director of Audience Engagement at More Perfect Union, one of the fastest-growing partisan-leaning YouTube channels going (nearly 3 million subscribers, over half added in the past year). Her explanation isn't a growth hack: substantive, relevant content, produced fast and left deliberately unpolished, with carousels tied to breaking news driving most of their Instagram follower growth. Worth reading past the politics for the mechanics: Parke's team treats each platform as a genuinely different audience (YouTube-first strategy, Instagram carousels for news moments, Shorts cut to stand alone rather than teasing back to a link), and deliberately avoids leaning on outrage alone to drive engagement. That's a useful corrective for any client convinced their account just needs to "go viral" once. Worth pulling into the next social strategy conversation: are a client's posts built to be relevant and complete on their own, or just polished, on-brand and forgettable? Parke's line is worth repeating back to them: "the more time we put into making a post, the less likely it would perform well." High standards, not more rules. Link in Bio

5. An actually useful checklist for getting a client's site AI-crawlable

Telepathic has published a detailed technical rundown on what actually determines whether GPTBot, ClaudeBot, PerplexityBot and Gemini can read, trust and cite a page: robots.txt rules that explicitly allow the right bots, JSON-LD schema (Article, FAQPage, HowTo) with @id references tying entities together, server-side rendering or pre-rendered snapshots for JS-heavy sites, and internal links written around entities rather than "click here". The headline stat: GPTBot traffic is up 305% year on year, PerplexityBot up over 250%. Worth an audit pass on any client site before the next "why aren't we showing up in ChatGPT" conversation: most of what blocks AI visibility is boring, fixable technical debt, a robots.txt that never got updated for newer bots, JS-rendered content bots can't read, schema that stops at Article and never gets to FAQPage or HowTo. Worth running the quick tests yourself before pitching a fix: view page source to check if content is JS-dependent, or run curl -A "GPTBot/1.0" against the URL to see exactly what the bot sees. Cheap to check, and a concrete way to show a client what's actually blocking them. Telepathic

Small business shout-out

This week's shout-out goes to Modu, founded in Copenhagen in 2018 by design engineers Theo Ginman and Jonathan Rasmussen. Modu makes life-size building toys: foam blocks, pegs and wheels that turn into rocking horses, ride-ons, balance boards or walkers, then get rebuilt into something else next year. One set, no expiry date, which is more than you can say for most toys built for a six-month attention span. Made in Europe from 100% recyclable materials and B Corp certified, Modu's whole pitch is longevity: the same blocks are meant to grow with a child from six months to six years old rather than getting binned the moment interest moves on. It started life on Kickstarter and has since built a proper following, nearly 800 Trustpilot reviews and counting. Know a small business we should feature? Reply and let us know.