Four data points from the AI-in-marketing stack this week. One of them should make you check your ad account settings before you do anything else.
Meta’s new terms make your AI-edited ads your problem
Meta revised its commercial terms to add an “AI Features” section. Under the new language, advertisers are solely responsible for output from Meta’s AI tools, including features Meta switches on automatically after an account update.
Bookshop.org discovered a generative AI feature had added an illustration to a human-made ad after an account update opted it in. The company has since disabled all of Meta’s AI ad features. Exverus pulled live ads after AI-generated changes, including sitelinks, translated copy, and AI-generated images and carousels, passed through several review checkpoints. AdCellerant now turns off Meta’s default AI automation on every campaign it manages.
Meta says advertisers can disable individual Advantage+ features in Ads Manager. The terms also let Meta crawl domains tied to a business’s products and use that content in its own products. Meta pushed the effective date from 30 October to 20 November after publisher and regulator pushback.
The same week, OpenAI said it will turn on “creative enhancements” by default for ChatGPT self-serve ads from 30 October, letting its system rewrite headlines and change landing pages.
What to do: Audit every Advantage+ and creative enhancement toggle in your ad accounts before the terms take effect. An AI edit you never approved is now your liability.

⏱️ Twilio’s CMO went from two weeks to one day per campaign
In a Fortune piece published 23 October, Twilio CMO Chris Koehler reported that account research and personalised content for a single campaign used to take over two weeks. It now takes under a day. The team went from three account-based campaigns per quarter to as many as 15.
His three-phase model runs: personal productivity first, then whole-team shared workflows and custom agents, then autonomous end-to-end workflows. Critically, the team mapped the manual process before automating anything. One early localisation attempt put berets on every person in a French campaign.
Koehler cites Deloitte data showing only 25% of organisations have moved 40% or more of their AI pilots into production. His argument is that Twilio’s gains came from the groundwork: brand consistency, data hygiene, governance, and integration, not from buying more tools.
What to do: Write down the manual process before you automate it. That is where the 5x output came from.
Shoppers ask ChatGPT, then go check everything else
A Feedvisor survey found 16% of shoppers start product searches in ChatGPT, but only 9% go there to buy, and only 5% purchase immediately after an AI recommendation.
PartnerCentric tracked what shoppers do after getting an AI suggestion: 33% check reviews, 22.3% compare prices, 18.8% search Amazon or another retailer, 14.8% visit the brand’s website, and 7.6% look for creator content. Another 33% say shipping and returns policies often or always change what they end up buying.
Prosper Insights found people will delegate low-stakes tasks to AI agents before purchases: 12.1% would have an agent book a restaurant, 11.0% buy groceries, and 7.3% buy apparel. Note that all three are vendor surveys, so treat the percentages as directional rather than precise.
What to do: Look at what a shopper sees after an AI recommends your product, specifically reviews, price against Amazon, and your returns policy. That is where the sale is won or lost, and AI visibility work does not touch it.
ChatGPT sent 0.62% of Google’s traffic. Google’s AI Overviews are on 88% of commercial searches.
Marie Haynes analysed 12 months of GA4 data across 100 sites and more than 145 million sessions. ChatGPT sent 0.62% as many sessions as Google organic search, and referral growth from ChatGPT stopped in August 2025. Google organic drove 27 times as many conversions as all LLMs combined.
ChatGPT visitors do convert at a higher rate: 13.8% against 3.02% for Google organic, across the 55 sites that tracked conversions. The sample has no stated industry mix or conversion definition, so the comparison is context-dependent.
Separately, Semrush found AI Overviews appearing on 88% of 400 tracked US commercial keywords, up from 45.75% in September 2025. For “reviews” queries specifically, AI Overviews appeared on 89% of searches, up from 30%. Semrush measures appearance, not clicks.
What to do: Track AI Overviews presence on your commercial terms this month. That is where organic shelf space is being lost, and where the visits still actually come from.
Quick Hits
- Cuttable raised A$20m for an AI ad platform that tells brands which creative to make next.
- A Snapchat/Ipsos survey found 70% of consumers find hyper-personalised AI ads intrusive.
- Sprinklr added eight customer-insight agents to Gemini Enterprise’s agent marketplace for existing Sprinklr customers.
- Peec AI launched an agent that ranks AI search fixes by impact and asks for approval before acting.
- Kochava added token budgets and model routing to StationOne so marketing teams can cap AI spend.
- Google added brand tools to Gemini, letting users generate on-brand visuals from uploaded fonts, colours, and images.
- A YouTube creator study found 72% of uploaders use AI and 61% want to put themselves into AI video.
- Target’s “Ask Target” AI shopping assistant is in beta on its site and app for holiday gifting.

