AI Is Moving Closer to Purchase Moments. So Are Your Ads.
Meta's Muse Wants to Do the Shopping for You
An AI agent that acts, not just answers, could become advertising's next front door.
What's the News:
Meta launched Muse on September 8, and it’s already the No. 1 free app on Apple’s U.S. App Store, with more than 2.5 million downloads since launch. Unlike a traditional chatbot, Muse can take action for users: researching options, booking travel, filling out forms, negotiating bills, and making purchases.
Commerce is already becoming an important part of that experience. Muse supports checkout through Stripe’s Link, Shopify is integrating Shop Pay, and Amazon recently blocked Muse from autonomously purchasing on its site. That standoff raises an interesting possibility: rather than letting outside agents freely transact on its platform, Amazon and Meta could eventually connect through advertising or another commercial partnership.
Why It Matters:
The agent becomes the new point of influence: If people increasingly ask agents to find products, compare options, and make decisions, the agent is where brands win or lose consideration.
Meta still has to decide how Muse makes money: Today, Meta says Muse conversations and data are not shared with its advertising systems. But if adoption keeps growing, Meta will need to decide how the product fits into its broader business model. Advertising is one possible answer.
Intent moves closer to the sale: Instead of interrupting a feed, an advertiser could potentially show up when someone is actively asking an agent for help solving a problem or making a purchase.
Muse is worth watching for two reasons: people are actually using it, and it sits much closer to action than traditional social media. Meta already built one of the world’s biggest advertising businesses around understanding intent. Agents could give it another place to do that—closer to the decision itself.
We are not there yet. But the brands best positioned for agent-era ad formats will be the ones AI can already understand and confidently recommend.
Marketers Should:
Watch agent adoption, not just AI usage. The important shift is from people asking AI questions to trusting it to take action.
Get your business information in shape. Clear products, services, pricing, FAQs, differentiators, and proof points give agents better material to work with.
Prepare for a new kind of intent. An agent helping someone choose may be much closer to conversion than someone casually browsing a feed.
Expect advertising models to evolve. Agent ads are still emerging, but discovery and paid media are moving closer together.
Keep testing the fundamentals. New surfaces won’t change the basics: strong creative inputs and measurable business outcomes still win.
Google Just Put a Number on Smart Bidding's Learning Period
"Give it two weeks" is out. "Give it 50 conversions" is in.
What's the News:
Google updated its Smart Bidding guidance to say calibration can take up to roughly 50 conversion events or three conversion cycles after a change, a shift from framing the learning period mostly in elapsed time (three weeks, or one to two conversion cycles). A campaign enters “Learning” after a bid strategy is created, reactivated, or has its settings or composition changed. How long it takes depends on conversion volume, cycle length, and the strategy itself.
The conversion-goal guidance now explicitly covers Search, Shopping, and Performance Max, and Google notes PMax can take longer when most traffic comes from outside Search and Shopping.
Why It Matters:
A yardstick you can actually defend: This gives a concrete, volume-based way to judge whether a bid strategy has had enough data to work, rather than just counting weeks.
Low-volume accounts feel it most: Hitting 50 conversions can stretch over a much longer calendar window than a client or internal stakeholder might expect.
Patience is part of the strategy: Google is explicit that 50 isn’t a hard requirement and learning can go faster with strong historical data. But it also warns against frequent changes to budget, bidding, or status during that window.
Google’s explicit language shifting from time to number of conversions reinforces the data-driven and statistically sound methodologies Silverback uses with clients, aiding in conversations about marketing experiments.
Marketers Should:
Reframe learning around volume. “Give it roughly 50 conversions” is an easier number to defend than “give it two weeks” when a stakeholder wants to react early to a bidding or goal change.
Translate conversions into calendar time. An account driving 10 conversions a week needs about five weeks to hit 50. Say that up front, before the change goes live.
Set expectations early on lower-volume accounts. These are the ones most likely to get pulled out of learning prematurely by an anxious stakeholder.
Build learning resets into QA. Bid-strategy or conversion-goal changes made mid-cycle reset the clock in a way that’s easy to lose track of. Log every change and its date.
Your Multi-Location SEO Report Might Be Lying to You
New openings and closures can make SEO look better, or worse, than it really is.
What's the News:
Search Engine Land contributor Dayna Lucio argues that multi-location brands should stop judging SEO by top-line totals. She says new openings, closures, and a few standout markets can make overall performance look better or worse than it really is. Her main fix is a “TrueView”: a year-over-year comparison that only includes locations open in both periods.
Why It Matters:
Growth may just be more storefronts: If your brand opened 20 stores last year, your YoY organic growth is probably partly just more storefronts, not better SEO. That’s a hard thing to have a CFO spot before you do.
Decline can be a false alarm: One closed market or one weak metro can make a healthy program look like it’s failing.
Segmentation shows where the next dollar should go: Breaking results out by geography, brand vs. non-brand, and Google Business Profile vs. traditional organic shows where revenue is actually being won or lost.
Marketers Should:
Build a same-store TrueView before your next YoY report goes out. Show it next to the total, so you can explain growth rather than just report it.
Put UTMs on every Google Business Profile website link starting today. Without them, GBP traffic blends into organic and you can’t tell which source is converting.
Track new locations in a separate view with their own benchmarks. Franchise owners and regional leaders need realistic ramp expectations, not comparisons to stores that have been open for 10 years.
Separate branded from non-branded search at the market level. Brand demand shows awareness you already have. Non-branded “near me” visibility is where you grow.
When conversions drop, check traffic before anything else. If traffic dropped, look at search visibility. If traffic held but users took fewer actions, look at CRO, reviews, or the location page itself.
Take what’s working in top markets and test it in the laggards. Segmentation earns its keep when it changes where you spend effort, not when it just produces more charts.
Google Tests "Dive Deeper" in Discover, and Publishers Are Right to Be Worried
Google is inserting another layer between the user and the click.
What's the News:
Google is testing a new “Dive deeper” button inside the Google Discover feed, announced by Google Search VP Robby Stein. When tapped, the button doesn’t send users to a publisher’s article. It opens a Google-generated AI topic overview with links to related stories pushed below it. The test starts with video cards, but code found by independent researchers suggests article cards are next. Google has not shared which markets are in the test, what share of users see it, or whether publishers can opt out.
The stakes are real: Discover now drives 68% of Google traffic to nearly 2,000 news and media sites, according to Chartbeat data.
Why It Matters:
Discover-dependent brands are exposed: For publishers, media brands, and content-heavy B2C sites, this is another layer inserted between the user and the click.
The answer arrives before the visit: The AI overview delivers the key facts before the user ever reaches the publisher, and the outbound links are buried below it with truncated headlines.
Traffic and engagement are both at risk: Even if users eventually click through, the engagement signal and the traffic volume both take a hit.
This fits the same pattern Silverback experts have been watching across Search, AI Overviews, and now Discover: Google is systematically inserting itself between user intent and the destination.
Marketers Should:
Baseline your Discover traffic now. Pull current volume from Search Console’s Discover report so you have a clean pre-test benchmark. GA4 blends most Discover visits into Google organic.
Be ready to measure the rollout. If “Dive deeper” expands broadly, you’ll want to see the impact clearly. Look at the data by content type, not just in aggregate.
Earn the citation, not just the click. Content that gets cited as a source inside the AI overview is more durable than content that just competes for the click. Being the reference Google’s summary pulls from is the new version of ranking first.
Meta Is Taking Away Another Lever
Placement and platform exclusions are on the way out. Creative is what's left.
What's the News:
Meta is beginning to remove advertisers’ ability to manually exclude certain placements and even platforms, giving its automated system more control over where ads run. It’s another step toward Meta’s broader vision of fully automated campaign management.
Why It Matters:
Control shifts from placement to inputs: As Meta removes more manual levers, performance will depend less on precisely controlling where an ad appears and more on giving the algorithm strong creative that can perform across Feeds, Reels, Stories, and other placements.
Efficiency is already under pressure: Metricool’s 2026 Social Ads Report found average spend per Meta campaign rose 47% year over year, while ROAS on Meta Sales campaigns fell from 5.48x to 4.57x. Fewer levers plus rising costs means your inputs carry more of the weight.
The goal isn’t one “perfect” ad. It’s a portfolio of genuinely different concepts that gives Meta enough signal to find what resonates, wherever it chooses to serve them.
Marketers Should:
Shift energy from delivery to inputs. Spend less time trying to out-optimize Meta’s delivery system and more on what you can still control: creative diversity, audience insight, messaging, and measurement.
Build a creative portfolio, not a single winner. Genuinely different concepts and executions, not minor variations, give the system real signal to work with.
Design for every placement. If you can’t choose where an ad runs, every asset needs to work in Feeds, Reels, and Stories.
Verify with your own measurement. As Meta controls more of delivery, incrementality and business outcomes are how you confirm it’s actually working.
We’re reinventing performance marketing because everything changed while you were reading this.
We help you navigate this new landscape with data-driven insights, platform expertise, and creative that connects, so your brand not only keeps up, but gets ahead.
Work with Us