3 Months of Running ChatGPT Ads: What We Know Now
A first look at ChatGPT Ads - targeting, reporting, where it fits next to LinkedIn and Google, and whether it's worth testing yet. Based on live campaigns we've been running on the platform since launch.
Who spoke
It's still early days. ChatGPT Ads launched CPM-only, automated bidding only went default six weeks ago, and there's no prompt-level data or third-party verification yet - cost and quality benchmarks just aren't reliable enough to call yet.
Targeting runs on conversation context, not job titles, keywords, or retargeting. It matches what someone's actually asking - comparing vendors, evaluating attribution tools.
It's a third, different layer - not a replacement: LinkedIn builds recognition before someone has a problem, Google catches people who already know what they want, and ChatGPT catches them mid-comparison.
What the session covered
Your buyers are asking ChatGPT which vendor to shortlist. Now there's a small, labeled ad sitting right below the answer they get.
At VertoTalks #25, Emil Zagorov walked through three months of running campaigns on the platform - what we've learned, and what we'd tell a B2B marketing leader before they spend a dollar.
We covered four things in detail:
- Targeting - runs on conversation context, not job titles or keywords; it matches the shape of what someone's actually asking, like comparing vendors or evaluating attribution tools because leadership is questioning ROI.
- Reach - ads currently only show to Free and Go tier users. Plus, Pro, Business, Enterprise, and Education see none, and there's no reliable read yet on how much of the B2B buyer base that excludes.
- Reporting - seven numbers: impressions, clicks, spend, CTR, avg CPC, avg CPM, conversions. No demographics, reach, frequency, prompt-level data, placement breakdown, or competitor visibility.
- Fit - additive, not a replacement: LinkedIn builds recognition before someone has a problem, Google captures people who already know what they want, ChatGPT catches them while they're actively comparing.
But three months in, there isn't enough reliable signal to responsibly put hard cost and quality numbers next to our LinkedIn and Google benchmarks: the platform was built CPM-only with performance tools bolted on later, automated bidding only went default six weeks ago, and the numbers that do exist are self-reported with no third-party verification yet.
Our recommendation: test it. Choose one top ICP and your biggest pain point, skip the massive budget, run it for six to eight weeks, and checkpoint weekly instead of waiting for a final number. Count pipeline influence, not just direct conversions - and build tracking and CRM connection before spending a dollar.
If you're weighing whether this belongs in your channel mix at all, our ChatGPT Ads service page covers how we approach it for clients.
Who spoke
Your host
Any questions about this session or future VertoTalks? Reach out to Nikol directly.
Get notified about upcoming VertoTalks sessions - subscribe to our calendar and never miss a session.
Subscribe to calendar