Disclosure and accountability aren't the same thing, and Meta just proved it.
The company recently updated its disclosure tags for AI-generated ads on Facebook and Instagram. When you use a tool like Background Generation or Image Generation, or edit with Photoshop or Dall-E, the ad now carries a label, caught automatically through metadata.
When Business Insider asked Meta about it, a spokesperson pointed to the terms of service. Reviewing the AI's output, they said, is the advertiser's job.
So the label tells you AI touched the ad, but it says nothing about whether AI got it right. Marketers beware: disclosure and accountability are not the same thing.
If you look at the brands building trust with customers right now, the pattern is the opposite. Aerie put Pamela Anderson in a spot daring AI to capture real human beauty. It can't, and the campaign landed the same quarter Aerie crossed two billion in trailing revenue. Polaroid built a whole summer around one line on a Coney Island billboard: Go jump in some water before the data centers drink it all up.
These brands aren’t necessarily anti-AI. I’m sure they’re investing millions in an infrastructure powered by it. They’re just making sure the robots sit in the right seats.
Use AI in the system and the process, but be careful how much you lean on it in places where you're working on consumer trust.
We have a session coming up on July 30th about this very topic. Two of our creative directors, Carly Fridhandler and Mike Rubino, will discuss where AI fits in the process and how to avoid slop. Register now to join them.
Now for this week's stories.
Patrick Patterson
CEO at Level Agency
What's interesting this week
Google's new AI search guidance says the SEO basics still work Google published its first official AI search guidance in late May, and it points straight at fundamentals: crawlable sites, real expertise, and local trust signals. Its "mythbusting" section says llms.txt files, content chunking, and specialized AI schema don't move AI search visibility. Ahrefs found 97% of live llms.txt files get zero traffic. Good SEO is still good GEO.
Key takeaway: Put your budget back into crawlability, genuine expertise, and local trust signals, since Google says those are the only things actually moving AI search visibility.
Google Ads will now show shoppers which ads AI made Google is rolling out a "How this ad was made" panel inside My Ad Center, visible across Search, YouTube, and Discover. It shows whether generative AI created or edited a given ad. Ads made with Google's own AI tools get labeled automatically, and advertisers using outside AI tools now get a control to disclose it themselves. Depending on local rules, that label may also show up directly on the ad.
Key takeaway: If you're using AI tools outside Google's own suite to build ads, start using the new disclosure control now, since local rules may put that label directly on the ad whether you flag it or not.
A 50,000-view YouTube video can out-earn a half-million-view Short
SocialPilot's new breakdown admits no study directly proves long-form video builds more brand trust than Shorts. What the data does show: longer, completed viewing sessions correlate with deeper audience familiarity, and long-form naturally produces more of them. Nielsen has YouTube capturing 13.4% of total TV watch-time as of mid-2025, ahead of every traditional network and streaming service tracked.
Key takeaway: There's no data proving long-form builds more brand trust than Shorts, so don't lead with that pitch. Lead with CTV instead, it's the more defensible reason long-form deserves a place in the mix.
Virtual insider session: Can human taste be automated?
How to effectively incorporate AI into your creative process Thursday, July 30, 2026, 1 PM EDT / 10 AM PDT, plus a live Q&A.
Level's creative team is taking on the question every marketing leader keeps circling: how much of the creative process AI can realistically own.
Carly Fridhandler and Mike Rubino will cover where AI belongs in the creative process, and where it makes the work worse. They'll get into the "AI aesthetic" problem: what happens to brand perception when audiences can't tell the difference. From there, the conversation turns to what performance data is starting to show about AI-generated ads versus human-made ones.
They'll also get into the legal and disclosure questions marketing teams are already navigating, without waiting for the industry to settle them first.
Carly is an internationally recognized AI artist known as Scorpion Sorbet, named one of Forbes' top female AI artists and exhibited from Tokyo to Paris. Mike has spent the past five years connecting Level's creative vision to client business outcomes.
Stick around for the live Q&A. That's where you get real answers on which AI creative vendor claims actually hold up and which tools are worth the investment.
Marketers have more measurement tools than ever, and still can't make the call
A new report from Ebiquity and the World Federation of Advertisers surveyed 71 senior marketing leaders across 10 industries, including PepsiCo, Ford, Mondelēz, and Mars Wrigley. Just 15% say effectiveness data is a primary influence on budget decisions.
That's not for lack of tools. Eight in 10 organizations already run marketing mix modeling and brand lift studies. The report calls this a speed and alignment problem instead: 54% say insights arrive too late to act on.
Only 14% of companies say marketing and finance agree on what "effectiveness" means. Less than 3% of marketers feel confident separating short-term sales lift from long-term brand building.
PepsiCo's Sorin Patilinet put it simply: the tools, the discipline, and the data coverage already exist. What's missing is turning it into decisions that move the business.
Related signal: Level covered the same tension in May with Google's Alana Parsons, on keeping measurement decisions steady while the platforms keep changing the rules.
TikTok has labeled 3 billion videos as AI-generated, and it's about to get pickier about spam
TikTok's transparency numbers landed this week. More than 3 billion videos have been tagged as AI-generated content, using a mix of C2PA Content Credentials, creator labeling tools, and invisible watermarking. TikTok also removed over 86 million fake accounts in the first three months of 2026 alone.
The platform is now testing tighter detection for AI-generated spam accounts in politics, financial advice, and medical content. Those are the categories where fake volume does the most damage to public trust. TikTok is also joining the C2PA Steering Committee, the industry body behind the content-credential standard Google and Adobe already use.
TikTok has also committed more than $4 million to a partner program with groups like NoFiltr and the Raspberry Pi Foundation. Since launching in November 2025, that program has generated more than 200 million views of AI-literacy content.
For brands, the practical read is less about your own content and more about the environment your ads sit in. A platform actively hunting AI spam in financial and health categories is protecting the exact adjacency multi-location healthcare and fintech advertisers depend on.
Every landing page and microsite Level builds eventually runs into the same problem. An AI coding assistant reaches for defaults, and the result looks like every other AI-built page. Meta just open-sourced the fix it built for its own teams.
Astryx is a component library, 150+ accessible React components built on React and StyleX, with theming, dark mode, and templates as one system. It grew inside Meta for eight years, powering more than 13,000 internal products before this rebuild.
Astryx documents itself for AI coding assistants through a CLI, exposing the same API and examples an engineer would use. That closes the gap where an assistant fills in generic patterns instead of the system's actual conventions.
A design system documented only for humans leaves the AI guessing, and the guesses look generic. Astryx removes that gap by design.
For a growth team building landing pages or internal tools with AI assistance, that means fewer rounds of "make it look less like AI." It means more time on the parts of the page that are actually differentiated.
When did you last audit what AI thinks of your brand?
Don't rely on an unreviewed algorithm to manage your brand’s reputation. Our scorecard pulls back the curtain on how AI platforms frame your brand sentiment and which sources they use to vet your authority.
Inside your custom report:
Visibility score and share of voice: The percentage of AI prompts your brand appears in, benchmarked against your top competitors across all major models.
Sentiment and positioning: Whether AI describes your brand favorably, neutrally, or negatively, and whether you're showing up as the primary recommendation or an afterthought.
Optimization roadmap: The specific content structure and authority signals you need to improve your score.
