Research
/
2026
Meta Ads: AI-First Advertiser Study
By the time advertisers reach Meta, they’ve already asked AI for help with their ad campaigns. I studied what kind of ads Meta should show them in that moment.
Project details
Company
Meta
Role
Research lead, end to end
Methods
Unmoderated think-aloud, survey, and participatory design sketches (n=18)
Year
2026
Overview
The Problem: Small business advertisers now figure out their ad problems with AI before they ever reach Meta, but Meta’s ads still treat them like beginners.
The Method: I ran a 3-step unmoderated study with 18 advertisers. Each one asked AI about a real ad problem, talked through the answer on a screen recording, then sketched the Meta ad they wanted to see next.
The Core Insight: AI gave people the right answer but couldn’t show them how to act on it inside Meta. The moment right after an AI session is a high-intent window Meta is missing.
The Impact: My 5-state framework became a creative briefing lens for Meta’s small business ad creative team and was shared with 3 leadership teams.
What did I notice?
Something changed, fast.
For years, when I asked small business advertisers where they go for help, the answer was Google, the Help Center, or Facebook groups. In about 18 months, that changed to one answer across the board: AI.
Why does it matter?
Meta shows up like it’s the first stop. It’s not.
By the time a business owner sees a Meta ad, they’ve usually already diagnosed their problem with AI. 14 of 19 were troubleshooting or improving their ads, not getting started. They don’t need an introduction. They need help doing the fix inside Meta.
How did I study it?
A 3-step unmoderated study with 18 advertisers.
I wanted people’s reactions in the moment, not what they remembered later. So each advertiser:
Asked their own AI tool about a real problem with their ads
Talked through the answer out loud while recording their screen
Sketched the Meta ad they’d want to see right after
What did a session look like?
The advice was right. Doing it was the problem.
One agency owner had ads that got clicks but no sales. ChatGPT correctly told him to check his conversion numbers in Ads Manager. He had no idea where to find them. “It told me what to do but not how to actually do it.”
AI also kept pointing people to Meta features they’d never heard of. A service business owner found out about Engaged Shoppers after months of running ads. An online store learned about Meta’s free Ad Library of competitors’ ads. “Why wasn’t I told about this?
What did I find?
Five headspaces after AI, and each needs a different ad.
I sorted all 18 sessions by how people felt right after using AI. Someone ready to go needs one click into Ads Manager. Someone let down needs to see that Meta’s own AI can help.
What did I recommend?
Segment by what they just learned, not who they are.
Every AI session is a signal. When someone asks ChatGPT why their ads aren’t converting, they’ve just told you their exact problem and what they need next. That’s more specific than any demographic.
Test problem-first ads against “Get started” ads, built around the real problems from this study.
Prototype an AI-to-Meta bridge: one page that maps problems people diagnose in AI to the exact place they get fixed in Meta.
Add a post-AI line to every creative brief: not just who the advertiser is, but where they just came from.
Where does it fall short?
I was upfront about its limits.
The sample leans engaged. People who sign up for studies are already curious, so the “AI gave me nothing” group is likely underrepresented.
18 people show patterns, not percentages. Sizing the 5 states needs a survey.
US-only, mostly solo operators. Bigger spenders, agencies, and other markets need their own studies.
What came of it?
Outcome
5-state framework adopted as a creative briefing lens by Meta’s small business ad creative team
Shared with 3 leadership teams: Creative X, Business AI, and Small Business Growth Marketing
Reframed how the org talks about advertisers who come to Meta after using AI
Next step: a survey to size the 5 states, test ads built from them, and explore whether Meta can detect post-AI intent in its own products
Next project
