18 Updates - Instagram 90-Min Videos, ChatGPT Ads
Happy Wednesday.
Bank of America just emailed its customers a polite request: please stop handing your banking password to your AI agent. Yes, people are really doing that. And if the agent goes on a spending spree, the bank says you are responsible.
Here is what caught my eye today:
- Instagram is testing 90-minute videos, and the money is in brand deals, not views.
- ChatGPT ads are showing up in the wild through Amazon DSP, and AI Overviews now sit on over 80% of searches.
- Influencer measurement is still a mess, so brands are buying their way to clearer numbers.
SOCIAL MEDIA
Instagram goes long, Pinterest goes programmatic, and Meta names its number
Scorsese needed almost three and a half hours for Killers of the Flower Moon. Instagram is giving creators 90 minutes. Paycheck not included.
This section is about formats and math. Instagram wants your long-form content, Pinterest wants your programmatic budget, and Meta wants you to stop splitting your budget into ad sets that never learn.
Coming soon to a feed near you. Instagram is testing Features, a format for videos up to 90 minutes, like podcasts, vlogs and interviews. Select US creators have it now.
VidIQ says YouTube pays creators $1 to $10 per 1,000 long-form views. On Instagram, views alone pay close to nothing. So brand deals are the main reason to go long there.
My take: expect creators to pitch you sponsored long-form content soon. If a creator can hold attention for 20 minutes on Instagram, that is a different kind of inventory than a 30-second Reel, and I would price it like one.
Pinterest plugs into your media plan. Programmatic Deals let you buy Pinterest display and video ads through the ad-buying platforms (DSPs) you already use.
You can buy through Amazon Ads, Google DV360, Yahoo DSP or DeepIntent in the Americas and Asia Pacific. Europe is not included yet.
My take: this removes the last excuse for leaving Pinterest out of a programmatic plan. If your audience is already planning purchases there, buying through your existing DSP makes the test cheap to run.
And Meta's magic number is 50. Meta's peak-season guide says each ad set needs about 50 conversions a week to learn.
Below that, ad sets cost about 28% more per purchase. To find how many ad sets your budget supports, use this formula: weekly budget ÷ (cost per result × 50) = your max ad sets.
Running more? Merge them before peak spend hits.
My take: this is the simplest budget math in the whole edition, and most advertisers ignore it. Count your ad sets, run the formula, and merge before peak season, not during it.
My take on this whole section: long video on Instagram only pays if a brand is behind it, Pinterest just made itself buyable the way you already buy, and Meta is telling you plainly that thin ad sets cost you 28% more per purchase. Three different platforms, one message: consolidate.
AI MARKETING
ChatGPT ads start arriving through Amazon DSP as AI answers swallow search
AI companies built the answer machines. Now the ads are coming, and so is the ad money.
This is the section to read twice. Ads are appearing inside ChatGPT, information-seeking has overtaken writing as the top ChatGPT use, Google's AI summaries now cover most searches, and the AI companies themselves are spending hundreds of millions on ads.
DSP-served ChatGPT ads are showing up in the wild. Users are spotting them because their click URLs route through Amazon's click tracking.
The report says it expects DSP to handle buying and likely targeting, while the creative stays standard ChatGPT. Details are still thin.
OpenAI is also testing a text ad with a dotted-underlined "Learn more" link and an upward arrow. Not every ad gets it yet.
My take: the plumbing is the story here. Amazon DSP handling the buying means ChatGPT ads will slot into the same workflows you already use, and that is when the money really starts moving.
The lookup tool. OpenAI's Signals data, a monthly sample of 1M consumer messages sorted by topic, shows that looking for information now edges writing in ChatGPT.
Lookups include product questions, so make sure ChatGPT can find accurate answers about yours.
My take: people are asking ChatGPT about products before they buy. If your product pages are thin or wrong, ChatGPT will answer with whatever it can find, and that answer might be your competitor's.
Google answers first, too. AI Overviews, Google's AI summaries above the results, now appear on over 80% of searches, per Semrush. That is up from 50% in June last year.
SE Ranking puts them on 82.5% of US brand searches. Google your own brand and check what it says about you.
My take: over 80% is not a test anymore, it is the default search experience. If you have never checked what AI Overviews say about your brand, do it today, because your customers are reading that summary before they ever see your site.
And the AI brands are buying. MediaPost projects Anthropic will spend over $655M on ads this year, with total sales and marketing topping $9B.
If you sell ad space, AI brands just became a category worth pitching.
My take: they built it, now they are paying to advertise it. When a company with a $9B sales and marketing budget starts buying, I pay attention to where that money lands, because it tells me which channels work.
My take on this whole section: the answer engines are becoming ad platforms, and the ad platforms are becoming answer engines. The practical move is boring but urgent: check what ChatGPT and Google's AI Overviews say about your brand, because that is now the first impression.
INFLUENCER MARKETING
Measurement is influencer marketing's biggest headache, so brands boost their way to clearer numbers
An Instagram view, a TikTok view and a YouTube view walk into a bar. Nobody can tell if that is three people or the same person three times.
That is influencer measurement, and Aaron Baar at Marketing Dive explains how brands work around it.
Two-thirds of marketers in a new Association of National Advertisers (ANA) influencer report call measurement the hardest part. Vetting, negotiating, contracting and finding creators trail at 40% to 45%.
Same word, different math. Each platform defines views, reach, engagement and video completion its own way, and no standard accounts for audience overlap across channels. Even impressions do not add up.
Do not expect a fix soon: Aquila, the ANA's cross-media measurement service, has not started on creators and needs platforms, agencies and marketers on board.
My take: this is the core problem, and nobody is close to solving it. Until the platforms agree on what a view is, every cross-platform report you build is an estimate dressed up as a fact.
Paid media brings receipts. So more brands pay to boost creator posts, since paid campaigns report far richer data. EMARKETER expects US boosting spend to match creators' sponsorship revenue, about $14.2 billion, in 2027 and pass it in 2028.
As a starting point, the ANA suggests 30% paid and 70% organic for awareness, flipped for conversion. Then boost what performs.
My take: boosting is the workaround that actually works, because paid reporting is the only reporting you can trust. I would treat the 30/70 split as a starting point, then move money toward whatever the boosted posts prove.
My tip. Write paid usage rights and whitelisting (running ads from the creator's handle) into the first contract, so boosting a winner does not mean renegotiating.
Early IAB survey results on creator brand deals flag inconsistent usage-rights terms as a barrier to growth.
My take: this is a one-line contract clause that saves a week of emails later. If you are signing a creator without usage rights in the contract, you are buying a post, not an asset.
Goal first, creator second. Pick sales, conversions or awareness before launch, or you cannot measure a win, let alone repeat it. It matters more as creators sell on marketplaces like TikTok Shop. "Now, they are the store," says the ANA's Leah Marshall.
My take: the creator-as-store shift changes the brief completely. If the creator is the storefront, the goal is not reach, it is sales, and the brief should read like one.
One roster to rule them all. Kantar's 2026 Media Reactions survey finds more brands likely to invest in micro (66%) and nano creators (61%) than macro (41%) or mega ones (42%).
More creators, more chaos: some global brands have two units negotiating against each other for the same creator. Put one group in charge, from negotiations to measurement.
My take: small creators are winning the budgets, but a bigger roster needs one owner. If two teams inside your company are bidding against each other for the same creator, you are paying twice for the same post.
My take on this whole section: until platforms agree on what a view is, the easiest rounds to count are the ones you buy. Boost what works, lock usage rights into the first contract, and put one person in charge of the roster.
DATA STORIES
Most shoppers are not ready to chat with your store
Retailers are racing to add AI shopping assistants. Shoppers are in much less of a hurry.
Bizrate Insights and EMARKETER asked US digital shoppers how interested they are in using retail AI chatbots, and most said no thanks.
Nearly half of shoppers (47.1%) have never used a retail chatbot and have no interest in trying one. Another 11% do not even know what one is.
Only about 1 in 5 have used one at all, and most of them did not stick with it: 12% tried one but do not use it regularly, while just 7.8% are regular users.
There is still room to grow. 22.2% have not tried a retail chatbot yet but say they are interested, which is more than all current users combined.
Shoppers already know how to use search bars and filters. A chatbot adds a step, and years of frustrating customer service bots have not helped its reputation.
My take: the 22.2% who are curious but untried are the whole opportunity, and they are bigger than the entire current user base. Win them with a chatbot that helps at one clear moment, like a holiday gift finder, instead of replacing search.
My take on this whole section: do not rip out search to make room for a chatbot. Add the chatbot alongside search, show it where it helps, and let the curious 22% come to it on their own terms.
QUICK UPDATES
SEO: Rankings swinging on keywords your site has nothing to do with? Your SEO tool might be seeing things. Semrush confirmed Google sometimes serves automated requests "poisoned" results that look normal but differ from what real users see, and calls it industry-wide.
Discord: Your server's best threads could soon double as SEO content. From November 16, each public Discoverable Server gets its own web page, and from mid-December, selected messages may show up in search results. Admins pick channels and members can opt out. It is off by default in the EU, UK, Canada, Brazil, Australia and South Korea.
Microsoft Ads: Only want your B2B ads in front of accountants at KLM? In Audience campaigns, you can now set LinkedIn profile targeting (by company, industry or job function) to Target and bid, so ads show only to people who match. Search, Dynamic Search Ads and Shopping campaigns stay Bid only, which just adjusts bids.
Google Ads: Running a Black Friday promo that only touches part of your app? Google Ads is testing seasonal adjustments at ad group level for App campaigns. You enter the conversion rate lift you expect for a set period, say +50% for a three-day sale, and Smart Bidding uses it only on the ad groups you pick.