AI Ads Are Getting Smarter. Your Measurement Isn't Keeping Up.
OpenAI Adds Custom Audiences to ChatGPT Ads
First-party targeting just landed in the last place advertisers expected it.
What's the News:
OpenAI has reportedly introduced Custom Audiences for its ad platform, allowing advertisers to upload email or phone-based audience lists for inclusion/exclusion targeting and bid adjustments. The article notes lists can include up to 5MM people, with a minimum matched audience size of 25K.
Why It Matters:
This signals that OpenAI advertising is moving beyond purely contextual targeting and toward more familiar performance media infrastructure. For marketers, that means ChatGPT ads may become more viable for direct response, retargeting, suppression, lifecycle marketing, and audience-based bidding — not just broad prompt-level relevance. This is a meaningful step toward making OpenAI a more practical performance channel, but it also reinforces why marketers need to approach emerging AI ad platforms with both curiosity and control.
Marketers Should:
Test ChatGPT as a high-intent, conversational placement — not a copy-paste of your Meta audience strategy
Set guardrails early: audience quality, exclusions, brand safety, and measurement before you scale spend
Use first-party data with intent, not just because the option now exists
Measure against incrementality and real business outcomes, not against “it’s new and everyone’s testing it”
GA4's New AI Assistant Channel Isn't Telling You the Whole Story
Attribution has been broken in media for <strike>years</strike> (ever). This is one of the first meaningful experiences SEO has had with Attribution’s shortcomings.
What's the News:
In May, GA4 rolled out an “AI Assistant” channel that automatically tags sessions when it recognizes a referrer from a tool like ChatGPT or Gemini. On paper, it sounds like the fix marketers have been waiting for. In practice, only an estimated 60–80% of true AI-originated visits carry a clean referrer header at all — the rest still land in Direct or get scattered across Referral and Organic Search. (TechWyse)
Why It Matters:
Search for “ChatGPT” as a source in your own GA4 and you’ll likely find it split across three different channel groupings: direct, referral and now this AI Assistant. It’s proof the real impact of AI search is bigger than what your session data shows. And the deeper problem isn’t referrer data. It’s that AI search and Google search work in parallel: a user researches on ChatGPT, gets a brand recommendation, then searches that brand by name on Google and converts there. Organic search gets the credit. AI search — the channel that actually drove the decision — gets nothing.
Marketers Should:
Stop treating last-click attribution as the source of truth. It wasn’t accurate before this update, and it isn’t now
Build a measurement framework that accounts for AI’s influence upstream of the click, not just sessions tagged “AI Assistant”
Track visibility and down-funnel engagement, not just session volume, to size AI’s real impact
Treat this GA4 update as a starting point for the conversation with leadership, not the final answer
New Research: IP-Based CTV Targeting Misses 3 Out of 4 Times
The "precise, deterministic" targeting you're paying for is mostly a guess.
What's the News:
New research from identity vendor Adstra and InterMedia Advertising found that IP-based CTV targeting is accurate just 23% of the time, meaning three out of every four ads aimed at a specific household miss. Device-level CTV IDs performed far better, holding a consistent link to the same person 71% of the time, a 24-point advantage over IP mapping. (Adweek)
Why It Matters:
Advertisers have been treating IP-targeting as deterministic and precise, but new research shows that’s far from the truth. In reality, it’s only accurate 23% of the time. IP-targeting is much more probabilistic than deterministic. While this is most pronounced in CTV, many advertising channels and measurement models rely on the same IP-based foundation
This is another data point that shows believing in precise targeting and measurement is increasingly a delusion.
Marketers Should:
Invest in modern measurement tools like MMM and incrementality testing to measure marketing channel effectiveness rather than increasingly flawed platform data
Scrutinize the targeting signals you’re sending to ad platforms. As AI takes on a larger role in optimization, the targeting, creative and measurement signals you set it are critical for it to optimize towards the right thing
Install a bot detection software that monitors the quality of traffic coming in through digital channels to spot your biggest hot spots
Google Images Just Became a Discovery Engine
The plain search box is gone. A personalized, scrolling feed just took its place.
What's the News:
Google Images turned 25 this month and marked the occasion by retiring its plain search box for a personalized, real-time scrolling gallery with saveable collections. Rollout starts on desktop in the US (English) over the coming weeks, and it’s signed-in only. (Search Engine Land)
Why It Matters:
Google Images is no longer just a lookup tool. It’s becoming a discovery surface. Ranking for a query isn’t enough when Google is actively recommending images based on a signed-in user’s inferred interests and saved collections. For ecommerce, travel, fashion, and home brands, that’s a new acquisition channel worth tracking on its own, separate from standard organic search.
Marketers Should:
Clean up image metadata and structured data now. Personalized feeds reward well-tagged visual content, and an afterthought product photo won’t surface in a feed actively being pushed to users
Segment Google Images referral traffic in GA4 today, before behavior shifts, so you have a clean baseline to measure against
Remember this is signed-in and personalized: the same product can surface differently to different users based on interest signals, not keyword match — plan your visual content strategy accordingly
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