G'day,
Welcome to the first This Week in AI from All Intents & Purposes. Every week I'll sift the artificial intelligence (AI) firehose and tell you what matters if you run marketing, run a business or sign off on the budget.
The format's simple: what happened, so what, and one thing to do about it. No hype, no jargon, no three-letter acronyms left unexplained.
And what a week to start. The ads have officially followed the questions, right here in Australia.
1. Google switches on ads in AI Mode for Australia
What happened: Google is scaling ads in AI Mode (its conversational search experience) to Australia, New Zealand and eight other countries, and launching ads in follow-up questions in the United States. Google says AI Mode queries are, on average, three times longer than traditional searches. Google also said lower AI Mode monetisation helped slow Search revenue growth to 12%.
Source: Google · Search Engine Journal
So what? Your Search, Shopping and Performance Max campaigns are eligible by default, and you can't target or opt out of AI Mode on its own. Results get lumped into standard reports. And some placements carry "AI explainers" that Gemini writes from the query, the answer and your own website, which you can't edit or approve. Your landing pages just became ad copy.
Do this: Audit your negative keywords and brand exclusions this week, then reread your top landing pages as if a robot were going to quote them. Because one will.
2. Google's new frontier model is here (sort of)
What happened: Google DeepMind announced Gemini 4 Carbon, its strongest model yet, focused on cybersecurity, coding and long-running tasks, just two weeks after Gemini 4 Argon. It tops the Artificial Analysis Intelligence Index (v5.0) with a score of 59. For now it's limited to selected cyber defenders, with introductory pricing of $2 per million input tokens and $10 per million output tokens.
So what? A new "best model in the world" now lands about once a fortnight. The real story for business owners is the price: frontier-grade capability keeps getting cheaper, which makes automation that didn't stack up last year worth another look.
Do this: Pick one repeatable workflow (reporting, briefs, product copy) and re-test it against the latest models once a quarter, not once a headline.
3. Meta's new ad terms put AI content risk on you
What happened: After pushback from publishers and advertisers, Meta has pushed its updated Commercial Terms and Self-Serve Ad Terms back from 30 October to 20 November. The terms say businesses using Meta's AI features are "solely responsible" for the risks and any resulting content. They also let Meta's crawlers access content on websites associated with your business. Meta says robots.txt (the file that tells crawlers what they can access) still applies.
Source: Marketing-Interactive
So what? If Meta's generative tools whip up an ad variant with a claim you can't back up, that's your problem, not theirs. The licence also covers content an agency provides on your behalf.
Do this: Before 20 November, add a human sign-off step for every AI-generated Meta creative, and check what your robots.txt actually says.
4. Australia moves to ban chatbots that hook kids
What happened: The federal digital duty of care bill passed the House of Representatives on Wednesday and heads to the Senate. It would put an overarching safety obligation on social media, games, apps and chatbots. An amendment the government accepted would require platforms to ban material or conduct that "simulates emotional intimacy" and fosters a child's dependence on it. The government wants it passed by the end of 2026, but its path through the Senate is unclear.
Source: The Guardian
So what? This is no longer just a problem for big tech. If your brand runs a customer-facing chatbot or an AI "companion" that Australians can reach, you'll want to know whether you're in scope before the Senate decides for you.
Do this: List every chatbot and AI assistant across your site and apps, who owns each one, and how it behaves if a user is under 18.
5. Canva builds its own image model
What happened: Canva released Canva Craft, the first image model built and trained from scratch by its in-house research team. It plans the composition first (layout, hierarchy, text), then renders images at up to 2K resolution in more than 30 aspect ratios. It's available to all Canva users in Dream Lab and to developers through Canva's public application programming interface (API).
Source: Canva · TechCrunch
So what? Hats off to the home team. Canva trained it on millions of human-made templates, so it's built for marketing assets: legible text, sensible layouts and outputs you can split into editable layers. That's the bottleneck in most paid social programs: not ideas, but variants.
Do this: Run one ad set's creative variants through Craft next to your current process. Compare time to launch and click-through rate.
6. Claude Duet: an AI assistant that comes to you
What happened: Anthropic opened a waitlist research preview of Claude Duet, starting with Pro and Max subscribers in the United States. It's a proactive agent that sends a morning brief, monitors your connected apps and keeps working on tasks in the background. It also asks for approval before it sends, pays, signs or deletes anything.
Source: Anthropic
So what? It's US-only for now, but the direction is clear: AI is moving from answering questions to doing the work. It'll free up your admin time, and your customers will increasingly send their own agents to research, compare and buy. If your site can't be read by a machine, you're invisible to the buyer's assistant.
Do this: Write down the three weekly admin jobs you'd hand off tomorrow. That's your agent brief, ready for when it lands here.
Quick hits
- Anthropic files for an initial public offering (IPO): it plans to list on the Nasdaq under "ANT"; share count and price range aren't set yet. Anthropic
- ChatGPT Ads hits 89 countries: OpenAI added 26 markets, says 100,000+ advertisers have run campaigns, and keeps ads to the Free and Go plans. OpenAI
- AI vs a 27-year-old computer science problem: OpenAI and Anthropic both claim AI-found proofs refuting the "exponential time hypothesis". Machine-checked, but not yet fully confirmed by humans. Science
- Alphabet's quarter: revenue up 25% to US$127.6 billion, but shares fell 8% as Search ad growth slowed. TechCrunch
Phil's take: intent now arrives in full sentences
For twenty years we bought keywords. Three or four words, a match type, a bid. This week Google confirmed that AI Mode questions are three times longer, and both Google and OpenAI are putting ads right where those conversations happen. Intent hasn't gone anywhere. It's just showing up as a paragraph instead of a phrase. The winners won't be the brands with the biggest keyword lists. They'll be the ones whose websites, reviews and content make them the obvious answer, with paid media as the backup when the AI needs a nudge. That's generative engine optimisation (GEO) and search working as one team, not two line items fighting over budget. For all intents and purposes, search is now a conversation. Make sure your brand has something worth saying.
Until next week,
Phil CEO, Intender · Intent marketer
Not sure how your brand shows up when AI answers the question? Get a second set of eyes on it. Book a Fresh Eyes Audit and we'll show you where intent is leaking.
