Remember the October 30 deadline I flagged for Meta's new business terms? It moved. The new date is November 20, and the clause that caused the fight is mostly still there. Today's briefing is short on launches and long on one thing you can act on this month: what you agree to when you keep using Meta for your shop.
Meta pushes its business terms to November 20, and the crawling clause stays
On October 23, Meta moved the effective date of its updated Commercial Terms and Self-Serve Ad Terms from October 30 to November 20. Its stated reason: more time to explain the changes to partners. Publishers and advertisers had already complained to EU and UK competition authorities, so I read the delay as a pause, not a retreat.
Here's where things stand. In the digest where Meta backed off its AI-training wording, I told you the word "train" had been cut. That's still true. The standalone sentence is gone, and the permitted purpose now reads "advertising and business products." But trade press that compared the documents found the crawling permission itself still in both agreements, and reports say it makes no promise to honour your robots.txt file.
Two other lines deserve your ten minutes:
- Meta's AI features "may make mistakes," and you carry the risk of what they produce.
- You indemnify Meta against claims, while its own liability is capped at the greater of $100 or what you paid in the last twelve months. Carrying on after notice counts as accepting.
Why it matters for you: if your Instagram or WhatsApp business account links to a shop website, assume Meta's systems can read that site. Check what's public on it: supplier prices, draft pages, customer photos you never meant to show. And if AI-generated ad edits are switched on, look at what they publish under your name. One caveat: I couldn't confirm whether you're liable for AI features you never turned on, so don't take that on faith either way. The preview pages also didn't match the announced date for a while, so check which version your own account shows.
Four AI labs agree on one "don't train on my site" switch
Quieter news, but it fits. On October 30, OpenAI, Google, Microsoft and Anthropic said they'd honour a single robots.txt token, __usage_ai-training, so a site can opt out of AI training across all their crawlers with one line. It was proposed by Anthropic's Kevin Kelley as an individual IETF draft. It's not a finished standard.
The fine print is where the story lives:
- It covers training only. Bots can still fetch your pages.
- It doesn't pull your content out of models already trained.
- Timing is uneven. Anthropic says it has started; OpenAI says "coming months"; Microsoft targets February 1, 2027; Google says 2027. Reports differ on who's live, so treat the dates as promises, not facts.
Why it matters: it's the first time the big labs agreed on one switch instead of making you hunt for each bot's name. Notice who's missing, though. Meta's crawler uses a single token for both indexing and AI training, so a site can't accept one purpose and refuse the other. If you own your shop's website, ask whoever built it to add the new line. It costs nothing, and you'll have decided on purpose rather than by default.
Meta's Business Agent gets a price list
This one is from September, but it's the cleanest way to read today's terms story. Meta's AI customer-reply agent, which answers questions, recommends products from your catalog and books appointments on WhatsApp, Messenger and Instagram, launched in June as free to start with paid plans promised. Meta One then put numbers on it in September, according to TechCrunch:
- Essential: from $14.99 a month, with expanded Business Agent access and a verified badge on the WhatsApp Business App.
- Advanced: from $49.99, with more agent responses, scheduling, and team access.
- Expert: from $149. Max: from $499.
Pricing and availability vary by region, app and account, so Iranian sellers shouldn't assume any of this reaches them. My take: "free" was the trial. The agent's capacity is now a line item, and your customer conversations are the thing it meters. Before you build your shop's replies around any platform's built-in agent, ask what happens to your costs, and your data, when the free tier ends.
Quick takes
- TikTok: on October 5 it launched a conversational Shopping Assistant and a one-click checkout called Buy Direct, in eligibility-based advertiser testing. TechCrunch says it was built with partners including Salesforce, Shopify and Stripe. Buying is moving inside the feed.
- Meta ads: a roundup of its Advertising Week announcements lists direct checkout, AI-generated product videos, an Ads MCP for outside AI tools, and chat-based campaign building. That's one roundup, not Meta's own post, so verify before you plan around it.
- Same pattern everywhere: the platforms want the sale to finish inside their walls. Your customer relationship is what you keep if you answer in the DM yourself.
The week's theme is control: who reads your site, who answers your customers, who gets the sale. If you want those replies handled by an assistant that works from your own catalog and rules in Instagram, WhatsApp and Telegram, see how Vardast does it. For the basics, start with our guide to the five jobs a WhatsApp AI bot has to do and what actually works in AI customer service for online stores.