Labs

Two AI Ad
Rewrites Flip On
This Week. Small
Advertisers Have Hours
to Opt Out.

Meta switches on in-app product browsing for existing ads on Oct. 28; OpenAI enables ChatGPT 'creative enhancements' by default on Oct. 29 or 30. Neither requires per-version approval.

Within 72 hours this week, the two largest AI ad surfaces reachable by self-serve small advertisers will begin rewriting live campaigns by default. On Oct. 28, Meta turns on in-app product browsing inside existing Facebook and Instagram ads. On Oct. 29 or Oct. 30 (OpenAI’s help page and its Oct. 20 email to self-serve advertisers disagree), ChatGPT enables what it labels "creative enhancements" across new and existing campaigns in OpenAI Ads Manager. Both defaults land roughly 30 days before Black Friday.

Neither requires per-version human approval. OpenAI’s own language, in the Oct. 20 email reproduced by PPC Land, is unusually direct: “when enabled, these ads can run automatically without requiring you to review or approve each version.” AdExchanger’s Oct. 23 roundup called it “spooky” and noted the single toggle covers four change types that “can change anything from the headline to the destination link.” Headline, description, image, destination URL. One switch.

Meta’s parallel move is narrower in mechanics but broader in reach. The new in-app page carries what Meta describes as “product info and personalized suggestions” and activates inside ads already running under Advantage+ Creative and Advantage+ Destination. Meta’s justification is a 16-day experiment from Aug. 6 to Aug. 21 across roughly 117,700 advertisers, which the company says produced a 3% average lift in conversions at greater than 99% confidence. The footnote is doing load-bearing work: a 3% lift, platform-measured, is the entire evidentiary basis for a default change that affects every eligible advertiser.

The context is a Meta that has already been forced to walk back AI auto-edits once this fall. Adweek documented advertisers, including Bookshop.org and the agency Exverus, pulling live creative to strip out auto-applied sitelinks, translations, and AI-generated carousels they hadn’t approved. Media buyer David Herrmann is among those raising the alarm on the product-browsing rollout. Meta’s updated Commercial Terms, effective Oct. 30, make advertisers “solely responsible” for AI-created or AI-changed content shipped through its systems. The liability moves down the stack at the same hour the AI moves up it.

For operators between five and 30 employees running paid acquisition into BFCM, the practical question is narrow. The opt-outs sit inside each platform’s campaign settings, and the window to touch them closes within hours of this piece’s publication. Owners who built landing-page funnels and tested headlines through September are, by default, handing both to a model this week. The platforms have decided the test is worth running on live spend. The 3% lift is Meta’s number; whose number governs the downside is now, per the Commercial Terms, the advertiser’s.

Google’s AI Mode shift earlier this month rhymed with this one in exactly the way platform rollouts usually rhyme: the surface changes first, the consent language updates second, the liability clause lands third.

Sources