Every few years a new ad platform opens and the same two camps form overnight. One camp declares it the future and pushes clients in before the tracking even works. The other camp dismisses it until the case studies pile up, then arrives late and pays mature-platform prices.

ChatGPT Ads opened to advertisers in January 2026, and both camps are already loud. We’ve been building on the platform since the early days, with real budgets and real tracking, so this is my attempt at the third position: an honest fit check. Who this channel fits right now, who should wait, and how to think about it without kidding yourself in either direction.

What You’re Actually Buying Today

Strip the narrative away and here’s the product: Reach campaigns, billed on CPM, serving a chat card format natively inside ChatGPT conversations. That’s it. No click bidding, no conversion bidding. Clicks and Conversion objectives are on the roadmap, but today you are buying impressions in front of people who are actively researching inside an AI assistant.

The platform’s own reporting is impression-only, so everything you learn beyond the impression comes from your own tracking stack: pixel, UTMs, GA4. (I’ve written up the full measurement setup in ChatGPT Ads Measurement: Pixel, GTM, and Separating Paid From Organic, and the build itself in the practitioner’s walkthrough.)

So the fit question is really: is a native impression inside a research conversation worth CPM money to your business, at a stage where you can’t bid toward outcomes yet?

For some businesses, genuinely yes. For others, genuinely not yet.

Who It Fits Right Now

Considered-purchase businesses whose buyers research conversationally. If your customer’s journey starts with questions (“what should I look for in”, “compare X and Y”, “is it worth paying more for”), that research is increasingly happening inside ChatGPT instead of across ten Google searches. Travel, higher-ticket e-commerce, B2B software, professional services. Being present inside that research conversation is the entire value proposition of this channel, and these are the categories where it’s strongest.

Brands with real awareness budgets and patience. This is a top-of-funnel buy today. If your media plan already includes CPM channels (YouTube, Display, CTV) and you evaluate them on reach and downstream lift rather than last-click ROAS, ChatGPT Ads slots naturally into that row of the plan. You already have the mental model.

Teams that want the early-mover learning curve. Inventory is less contested, the format is new enough that decent creative stands out, and the platform knowledge you build now compounds. When conversion objectives arrive, the advertisers with months of pixel data and creative learnings will start from ahead. That knowledge advantage is a legitimate reason to pilot, as long as you’re honest that it’s a learning investment.

Businesses with their measurement house in order. Because the platform tells you almost nothing, your own stack has to. If your GA4 is clean, your GTM is well-run, and you have the discipline to UTM-tag everything, you can actually read this channel. That’s the entry requirement.

Who Should Wait

Anyone who needs conversions this quarter to justify the spend. If the money for this test would otherwise fund proven demand capture, and your business needs that demand, keep funding the proven thing. A channel that can’t bid toward conversions yet should never displace one that reliably delivers them.

Pure impulse and emergency categories. If your customer goes from need to purchase in minutes (food delivery, emergency home repair, low-cost impulse buys), a research-conversation placement is solving a problem your buyer doesn’t have. Search and local capture that intent far better.

Businesses that can’t deploy the tracking. No GTM access, contested GA4, no one to own UTM discipline: without those, you’ll spend the budget and learn nothing. Fix the measurement foundation first. It’ll pay off on every channel, not just this one.

Tiny budgets that can’t sustain a real test. If the pilot budget is so small that it needs to “work” within three or four weeks, the test is designed to fail. Wait until you can fund patience.

The Early-Data Trap: Engagement vs Conversions

Here’s the pattern we’ve seen firsthand, and it’s the single most important thing to internalize before you launch.

Early on, the engagement numbers look genuinely good. Sessions from tagged ChatGPT traffic behave like real, interested visitors. They stick around, they view products, they act like people mid-research. Meanwhile, the conversion column sits at or near zero for weeks.

Both signals are real. The engagement tells you the platform is putting you in front of the right people. The empty conversion column mostly tells you your buying cycle is longer than your campaign is old. Someone researching a considered purchase in ChatGPT this month may not buy for two or three months, and when they do, last-click attribution will hand the credit to the branded search or direct visit they came back on.

That second part matters: check first-touch attribution in GA4, not just last-click. First-touch views regularly surface assisted conversions from ChatGPT traffic that last-click hides completely. If you judge this channel purely on last-click conversions at week four, you will kill tests that were quietly working. This is where human judgment has to stay in the loop: the dashboards will not make this call correctly on their own, and no automated rule should be allowed to make it either.

The honest framing: engagement data is your leading indicator, conversion data is your lagging one, and the gap between them is your category’s buying cycle. Plan the test around that gap instead of pretending it doesn’t exist.

What a Sensible Pilot Looks Like

If the fit check says go, here’s the shape we recommend:

A quarter, minimum. Commit to 90 days before you judge. Shorter than a buying cycle isn’t a test, it’s a coin flip.
A budget you won’t flinch on. Sized so nobody panics in week three. Modest and steady beats large and nervous.
Tracking before spend. Pixel deployed through GTM, events mapped, every URL UTM-tagged. If this isn’t done, the launch date moves, not the tracking standard.
Simple structure. One Reach campaign, two or three ad groups, a small set of genuinely different chat cards. You’re learning, not scaling.
Pre-agreed success criteria. Decide up front what good looks like at day 90: engagement quality of tagged sessions vs your other paid channels, first-touch assisted conversions, cost per engaged session. Write it down before launch so nobody moves the goalposts in either direction.
A senior human reviewing weekly. New platform, thin native reporting, no conversion bidding. This is exactly the environment where judgment earns its keep and autopilot burns money.

Who We Tell Not to Run It

I’ll end with the list we actually use when someone asks, because honesty here saves everyone a bad quarter. We do not recommend ChatGPT Ads right now for:

Businesses where this budget would come out of working demand capture. Fund the thing that feeds you first.
Urgent-need local services. Your buyer isn’t researching in ChatGPT while their basement floods.
Anyone who needs last-click proof within 30 days. The platform cannot give it to you yet, and we won’t pretend otherwise.
Teams without measurement basics in place. Untracked spend on an impression-only platform is donation, not marketing.
Anyone buying it because it’s new. Novelty is not a strategy. Fit is.

If you land on the “fits now” side of this list and want the pilot run by someone who has already built on the platform, deployed the tracking, and lived through the early-data trap with real budgets, that’s exactly what our ChatGPT Ads management engagement covers. And if you land on the “wait” side, waiting is a fine answer. The platform will still be there when your fit improves, and we’ll say that to your face too.

Should You Advertise on ChatGPT Yet? An Honest Fit Check