Meta wants your ad clicks on its pages and your site for AI training

Meta made advertisers three offers this month.

You can refuse two, but the third is pure Godfather.

Taking the scenic route: Product browsing is a new Meta ads setting that sends your ad clicks to a Meta-hosted catalog of your products instead of your site.

Meta switches it on by default for existing ads after October 28. It says an August test with 117,000 advertisers lifted conversions 3% on average.

You can switch it off any time in Advantage+ Creative and Destination settings.

Your pixel wasn’t invited: Meta is also testing AI-generated landing pages. Its AI builds them from your ad and site content and shows them instead of your website.

The test started October 14 on under 5% of impressions, Common Thread Collective reports. Your pixel and GA4 don’t fire there, so those visits vanish from your analytics. Not nice.

To opt out, switch off “test new AI generation features” in Ad Account Settings.

The one offer you can’t refuse: From October 30, Meta’s updated Commercial Terms, the rules every business account accepts, let Meta crawl your linked website.

As per Ars Technica, Meta can use that content to “train” its products. The only opt-out is leaving Meta.

Its old ad terms only allowed crawling for your ads. Meta says nothing changes for its crawlers, and robots.txt still works, so we’d check yours.

Lights, camera, reference: Instagram’s new “Use as reference” mode plays a Reel under 90 seconds in your camera view, so you can sync a response (iOS only).

Edits, Meta’s Reels editing app, also added creator tools:

• Tappable links to the songs and sounds you use
• Kits that save style combinations for reuse
• Trending topics among your audience (over 10,000 followers only)

Decide which offers to refuse before October 28.

First BFCM signals point to a $1 trillion holiday and pricier clicks

Black Friday is next month. May the odds and the CPCs be ever in your favor.

But will they?

The prize pool is growing: NRF’s annual holiday forecast predicts US retail sales in November and December will grow 4.2% to $1.013 trillion, faster than last year’s 3.1%.

Shoppers will still mind their budgets, though. NRF expects inflation and rising gas prices to hit income groups differently, so plan offers at every price point.

Flat CPCs had a good run: Google Search CPC growth jumped from 1% in Q2 to 4% in Q3, per Tinuiti’s Digital Ads Benchmark Report.

Amazon’s retreat from US Google Shopping auctions in July 2025 kept CPCs flat for a year. Then other ecommerce giants moved in.

Now everyone else bids higher to hold impression share, so budget for pricier clicks this BFCM.

Social platforms expect a bonanza: TikTok’s new holiday shopping data shows TikTok Shop, its in-app store, passed $500 million in sales over BFCM 2025.

LIVE Shopping, TikTok’s livestream selling format, more than doubled its US sales year over year in the first half of 2026.

TikTok expects it to play an even bigger role this season. Selling there? Plan your BFCM livestreams now.

In a convenient timing, bots get a checkbox, too: Shopify’s Checkout MCP and Cart MCP let AI agents fill carts and check out on Shopify stores.

Now shoppers buying through an agent can:

• Opt in to your email or SMS marketing
• Make the same privacy choices as in your own checkout

You get those choices with the order, so AI agent sales can grow your lists too. The catch: only agents that adopt the update pass them along.

Let the bidding games begin.

Your custom conversion belongs in a reporting column

Selling one blue shirt? Then telling Meta to find only blue shirt buyers sounds like the obvious move.

But Jon Loomer makes the case that optimizing for that custom conversion usually costs you more than it gets you.

Three options, two with the same first name: A standard event—like Purchase or Lead—is Meta’s default.

A custom event covers an action the standard list doesn’t, like Jon’s PodcastPlay.

A custom conversion uses rules, no code needed, to carve a slice out of an existing event. Such as purchases of that one blue shirt.

Narrower target, thinner data: Sales campaigns already fight for conversion volume, and a custom conversion that counts one product shrinks the pool further.

When Meta sees very few conversions, it assumes your ads are failing.

It also narrows your reporting. If someone clicked your ad and bought something else, you and your client both want to know.

Inflated results are smaller than they look: Default reporting mixes click, view-through, and engage-through conversions, first vs. all conversions, and sometimes sales of products you weren’t promoting. Never take it at face value.

If you send traffic to one product page, most reported conversions will be for that product anyway. The attribution breakdowns show what sits behind the total.

Main gist: Optimize for the standard event. Then add the custom conversion as a column in Ads Manager to see exactly how many of that product you sold.

Meta gets its volume, and you still get your product-level number.

When the custom conversion earns the slot: First, it needs enough volume on its own to exit learning.

Second, you need a real problem beyond inflated numbers, like Meta leaning hard into a product that doesn’t make you money.

That rarely happens unless the page makes it easy to buy something else.

The Crew’s tip: Meta’s rough bar for exiting learning is about 50 optimization events per ad set in a week. Check your custom conversion column against it before you switch.

Let Meta chase the broad signal and do the narrow counting yourself.

Everyone wants unified measurement, but almost nobody has it

Marketing measurement works like a group project: every channel claims credit for the A.

That’s according to a recent report we found.

Unified cross-channel measurement (40%) is the clear winner, followed by higher-quality data (33%), and better incrementality testing (31%).

A quarter would prioritize AI for faster insights, tied with building in-house expertise.

The surprise is at the bottom of the list. Only 16% picked using insights to plan future spending, which is the whole point of measuring.

What we think: Every platform grades its own homework, so teams end up with five reports and no single answer. Leaders want one source of truth, but they need better data first.

Before you buy a unified tool, run one holdout test on your biggest channel.

Then tie every report to one budget decision.