Marketing Intelligence Weekly: Video Testing Gets Sharper
Date: 25th September 2026
TL;DR
Google is moving Search campaigns that use its campaign-level broad-match setting into AI Max. Check which campaigns are affected and how keyword priority will change.
YouTube has announced testing for up to three different cuts of a video. It is coming soon, not available to every channel today.
Instagram's AI labels are worth checking on your own creative; a claimed reach penalty from one marketer is not proof of a platform-wide effect.
Cheap Meta leads can become expensive if nobody can contact or qualify them. Measure the next step after the form.
Apple's coming EU tracking-prompt changes could alter reported attribution. Compare any movement in platform ROAS with actual sales.
Search, Social & Platform Updates
Google's AI Max migration changes more than a campaign label.
This does not mean every Search campaign is being moved or that AI Max is inherently better or worse. It means a defined group of advertisers should know when their campaigns changed and what they are comparing afterwards. A jump in traffic, lead volume or cost could reflect a different mix of searches rather than a sudden improvement in the ads themselves.
YouTube wants creators to test the opening, not only the packaging.
At its 23 September event, YouTube said it would soon let creators compare up to three cuts of a video to see which hook holds attention best. Title and thumbnail testing is already familiar; testing different opening edits could give brand and creator teams a more direct answer to the question, “Where do viewers lose interest?” The video-cut feature is still forthcoming, so there is no basis yet for promising a particular lift.
Why it matters
Both updates move a marketing decision closer to something observable. Google requires a clean before-and-after account record; YouTube is offering a way to test the creative itself. Neither tool replaces the basic discipline of knowing what changed and what outcome matters.
What you can do
In Google Ads, identify campaigns using the campaign-level broad-match setting. Record the migration date, search-term mix and conversion quality before deciding whether the new setup helps.
For YouTube, prepare genuinely different first moments for a future video test: change the opening idea or pacing, not merely a word on screen. Compare retention alongside your actual campaign goal once the feature reaches your account.
Paid Media & Acquisition
Meta lead campaigns need a second scoreboard.
A marketer running lead generation across several niches described a familiar-looking mismatch: acceptable cost per lead on the platform, but contacts who do not answer, provide unusable details or do not recall submitting a form. This is one advertiser's account, not evidence of a new Meta-wide failure. The useful question is whether the team is optimising for the first conversion that is easy to count instead of the first conversion that is valuable.
For a service business, two campaigns with the same reported cost per lead can have very different economics if one produces appointments and the other produces dead numbers. Higher-intent forms, a landing page or better qualification might help, but each adds friction. The right test is not “Did the form count fall?” It is “Did contactable prospects and real opportunities improve for the same spend?”
Why it matters
Lead quality is where ad-platform reporting meets the sales team's working day. Optimising only for submissions can shift the cost downstream, particularly for small teams that cannot call hundreds of weak leads.
What you can do
Join campaign and form data to contact attempts, qualified conversations and appointments. Review cost per qualified lead alongside platform cost per lead.
Test one quality lever at a time—for example, a qualifying question or a landing page—and keep the downstream definition of a good lead unchanged while comparing results.
Content, Brand & Communications
Check how AI-assisted creative appears after it is published.
An Instagram marketer reported that edited Reels received “AI Info” labels and believed their audience skipped them more often. Meta has explained that its AI labels can rely on technical signals in content, but that account's
claimed skip-rate increase and reach penalty have not been established more broadly. For a brand, the actionable point is simpler: the published post, not the editing timeline, is what viewers see.
Before treating a label as the cause of weaker reach, compare similar posts, audiences and publishing conditions. A poorly performing Reel may have a weaker opening, less relevant subject matter or simply different distribution. Trying to hide AI use by stripping metadata would also solve the wrong problem. Better to understand when labels appear, keep creative provenance clear and judge the audience response honestly.
Automation can save writing time while losing the voice.
In a separate B2B account, a social media manager said a client replaced their captions with generated copy for a month and saw comments fall sharply. It is one unverified account, not a general verdict on AI writing. The lesson is to test the finished message, not the speed of producing it. Copy that sounds interchangeable may meet an output target while doing little to start a useful conversation.
Why it matters
AI tools are part of ordinary creative production now. The quality check has to move from “Was it generated?” to “Does the audience understand, trust and respond to it?” Disclosure, clarity and the first few seconds of attention are all part of that check.
What you can do
Audit a sample of published Instagram creative for labels and compare like-for-like retention before drawing conclusions about reach.
If you use AI to draft captions, keep the brand's concrete examples and point of view. Compare replies, saves or qualified enquiries with your previous baseline—not just the number of posts shipped.
Measurement & Martech
Apple's EU privacy change is a reporting watchpoint, not a sales forecast.
Apple says that from iOS and iPadOS 27.2, apps in the EU can use an alternative App Tracking Transparency prompt. In Germany, France, Italy, Poland and Romania, that alternative will be the available version. Apple also says an app may ask again a year after a user's earlier choice, even if the user previously declined, unless tracking requests are disabled in device settings.
One interpretation circulating among advertisers is that more consent could make Meta's reported return on ad spend rise without more purchases. That is plausible as a measurement scenario, but it is not an observed outcome and users still control their choices. Indian teams running campaigns into Europe should therefore note the change without building next quarter's forecast around it.
Why it matters
An attribution chart can change because the measurement system changed. The business question - did we gain customers or revenue?—needs an independent answer.
What you can do
Mark the relevant iOS rollout dates in European campaign reports and segment the five named countries rather than averaging them into all markets.
Compare platform-reported conversions with orders, first-party analytics and revenue. Treat a higher reported ROAS as a prompt to investigate, not automatic proof of stronger demand.
Comments