19 Updates - Meta Ad Prices, PMax, ChatGPT Ads
Happy Friday.
Meta dominates this edition. Pricier ads, an AI agent that passed 10 million users, and Meta's AI quietly editing ads nobody asked it to touch. Google wants your PMax images and videos on by default, and Google AI Mode just became my favorite free content gap tool.
Here is what caught my eye today:
- Meta's ad prices are outpacing its ad volume, so holiday auctions will cost even more.
- Performance Max is switching image and video grabs to opt-out.
- ChatGPT Ads crossed a $1B annualized run rate, and AI Mode can audit your content for free.
AI just found a critical flaw in PHP. AI helped researchers find a critical flaw in PHP, the programming language behind roughly 70% of websites, WordPress included.
Until a fix arrived, some site owners chose the safest setting available: off. Because you cannot hack a site that is not there.
If your site runs on WordPress, ask your developer if they have updated PHP. Gently. My take: AI is now finding vulnerabilities faster than teams patch them. The defense playbook is boring and effective. Update everything, keep backups, and never leave a known flaw sitting around.
META
Meta books $64.6B, Muse crosses 10M users, and the AI edits your ads unasked
Remember when a billion dollars was the cool number in The Social Network? Meta just booked $64.6B in one quarter.
This section is all Meta. Prices are up, Muse is growing, creative controls are improving, and there are a few reminders that AI is running your ads whether you approve or not.
Meta ad prices jump 14%. Meta's Q3 results show ad prices outpacing ad volume, year over year. Revenue hit $64.6B, up 26%. Average price per ad rose 14%, while ad impressions grew only 10%. Daily active people reached 3.64B, up 3%. Meta guided Q4 revenue at $73B to $76B.
My take: demand is growing faster than supply, and holiday auctions will cost even more. I would budget for pricier inventory through December, and tighten creative testing so every dollar works harder.
Meta's AI agent Muse passes 10M users. Muse, Meta's AI agent app, browses stores and buys things once users approve. Meta told investors it passed 10M users, with shopping as the top use. Over a quarter of engaged shoppers average $200+ per order.
My take: agentic shopping is moving from demo to checkout. If your product feed is a mess, an AI agent browsing it will not save you. Clean feeds, accurate prices, and clear availability matter more than ever.
Meta's Marketing API now tags creatives by age and gender. Meta's Marketing API now lets ad tools tag each creative with the age range and gender it suits best. One ad set can aim a retirement-planning video at ages 50 to 65 and a savings video at all ages.
I think this is genuinely useful for age-skewed products. If you have been hacking around this with separate campaigns, this simplifies the build and keeps everything in one place.
Meta's AI edits your ads without asking. Buyers told Adweek that Meta's AI edits their ads without approval, adding sitelinks and AI-generated images.
And from Oct 30, Meta's updated terms make you "solely responsible" for whatever its AI features create or change, even ones Meta switches on. My recommendation: audit the Advantage+ settings behind these edits, collapsed menus included. If you do not check, Meta's AI will spend your money redesigning your ads.
A New Mexico judge orders Meta to pay $25B. A jury found Facebook misled users about control over their data, and a New Mexico judge ordered Meta to pay $25B. Meta says it will appeal.
Separately, Q3 costs include a $9.89B charge for Meta's teen mental health settlement with 52 attorneys general. My take: $64.6B a quarter buys plenty of lawyers. These fines are real money, but they are not slowing the ad machine.
My take on this whole section: Meta is the most expensive it has ever been, the most automated it has ever been, and the most legally exposed it has ever been, all at once. Budget for the first, audit for the second, and keep an eye on the third.
GOOGLE ADS
Lifecycle bidding expands while PMax grabs your images and videos by default
Advertisers have asked Google for more PMax control for years. Be careful what you wish for. This month you get sharper control over who you bid on, and PMax helps itself to your images and videos.
Customer lifecycle goals expand this month. Customer lifecycle goals, which tell Google Ads to bid differently for new and existing customers, are expanding. Shopping gets New Customer High Value Mode, which bids up for new customers likely to spend more.
PMax, Search, and Shopping get three retention modes: Re-engagement for lapsed buyers, Loyalty for program members, and Retention Only for segments you name. Bid boosts in Search, Shopping, PMax, and Demand Gen can now be percentages, so a 20% boost scales with cart size.
The catch: your customer data has to actually tell these groups apart. My take: this is real control, but only if your data is clean. If your customer lists are a mess, fix the data before you trust the bidding.
PMax grabs your images and videos by default. Two features that pull your images and videos into Performance Max ads are switching to opt-out. Sourced videos take clips from approved sources like your socials, site, or YouTube channel. They are on by default for new campaigns, and possibly some existing ones.
Landing page images use pictures from your landing pages. From December 4, they are on for every new PMax campaign, even if you switched dynamic images off account-wide.
My recommendation: check Asset optimization in every PMax campaign and confirm you own the rights to what Google can grab. You do not want a licensed photo or a client's logo appearing in an asset you never approved.
Next time, wish for fewer defaults. I would rather Google ask permission than make me hunt for opt-outs after the fact.
AI MARKETING
Google AI Mode's follow-up questions can be your free content gap audit
One of my favorite tricks in this edition comes from Claudia Tomina, who turned Google AI Mode into a free content strategist. It is simple, it is free, and it works for any service business.
AI Mode asks the questions your page does not. Claudia's client was one of more than a dozen Michigan law firms running nearly the same pitch, promising to fight for clients and charge nothing unless they win.
So she typed a real buyer query, "Do I need a lawyer after a car accident in Michigan?", into AI Mode and clicked through its follow-ups. None of them were about calling an attorney. They dug into Michigan's pain-and-suffering threshold, how comparative fault cuts a settlement, and when vehicle damage goes to small claims court.
My take: every follow-up AI Mode keeps returning to is a question your page should answer. Run your buyer's question through AI Mode before your next service page. Whatever your page cannot answer yet goes in the brief. Whatever every competitor already says can probably go.
UNKNOWN is a perfectly good answer. Her 10-phase prompt tags every finding as OBSERVED, VERIFIED, CLIENT-SUPPLIED, INFERRED, or UNKNOWN, which keeps the AI from passing off a plausible guess as fact. Settlement ranges and years of experience came back UNKNOWN, each with a question for the attorney.
Then the attorney had answered most of those questions on podcasts, so Claudia fed 13 hours of YouTube transcripts into Claude. They filled almost every gap, from the fee structure to real verdicts and settlements. Timestamps let the page link each claim to a 30-second clip.
The Crew's tip in the newsletter: podcast answers age, so get the expert to sign off on every fee and figure before the page goes live. No podcast? Record a one-hour call with the client instead. My take: your client has already given the interview somewhere. Dig it up before you pay for new content.
Each AI model cites something different. Yext studied 17.2 million AI citations. Gemini favored official websites. Claude cited reviews and social content two to four times as often as other models. Perplexity pulled from business-owned sites roughly 37% to 50% of the time.
My take: a page that sounds like every competitor gives the models nothing to pull out. If your page is unique, specific, and well-structured, you get cited. If it is generic, you are invisible.
My take on this whole section: AI Mode is a free content gap audit sitting in your browser. The brands that win AI search will be the ones whose pages actually answer the follow-up questions.
DATA STORIES
Half of America still hasn't tried an AI chatbot
AI feels like it is everywhere, at least if you are in a marketing circle. Pew Research Center just asked US adults a simpler question, and the answer is humbling.
Half of America has never used a chatbot. Pew Research Center asked US adults whether they have ever used a chatbot like ChatGPT, Gemini, or Copilot, and the country splits almost exactly in half.
The biggest drop comes at age 50. Usage falls 19 points from the 30 to 49 group to the 50 to 64 group, and young adults are nearly 3x as likely as seniors to use a chatbot. Younger adults run into AI at school and at work, so using it becomes a habit. Older adults have fewer reasons to try it, and for them the search bar still gets the job done.
My take: check your audience's age before you move budget to AI search. If most of your buyers are over 50, most of them do not use chatbots, so keep investing in the classics. For now. Still, stay on alert. It is only a matter of time before AI usage catches up everywhere.
QUICK UPDATES
OpenAI: ChatGPT Ads hit a $1B annualized run rate less than 200 days after launch. OpenAI's engineering post shows how your ad gets in: a custom AI model reads the whole conversation, then shortlists ads by relevance, predicted engagement, and past performance for the auction. Nothing fits? No ad. And OpenAI calls that a success.
ChatGPT Ads: Got customers in Argentina, Nigeria, or Pakistan? You can now reach them in ChatGPT. OpenAI added 26 countries, including 14 more in Latin America, for 89 in total. Buy through tech partners or self-serve Ads Manager.
Spotify: More ears, softer prices. Spotify's ad-supported audience grew 12% to 499 million in Q3, and ad revenue rose 10% to €482 million. It sold more music ad impressions at lower prices, and automated channels drove the most growth. Your audio budget might rise.
Reddit: A new Reddit Ads Manager planner predicts reach, impressions, and costs before launch, and an Optimization Score rates each campaign from 0 to 100% with suggested fixes. Reddit says advertisers scoring 80%+ saw 19.6% lower blended CPA, though stronger advertisers may simply score higher.
Walmart: Your next Walmart customer might be a bot with a spending limit. The company's new Agent Gateway lets approved AI agents search its inventory and buy for shoppers. Next month, Agent Profiles let shoppers approve every order or cap each agent's spend, and see its orders separately.
Gemini: Bye, pasting hex codes into prompts. Gemini's new small-business hub saves your website, brand kit, fonts, and colors, so later images from Nano Banana 2.1 stay on-brand. It also turns receipts and spreadsheets into financial models and connects to Google Ads, Canva, and Shopify.