I spent $1K on ChatGPT ads — here’s what actually happened

ChatGPT ads are no longer a closed beta for big spenders. Anyone can now sign up, get approved in about a week or two, and start running campaigns. But is it actually worth putting real budget into this new ad platform yet?

After spending around $1,200 on ChatGPT ads for a Google Ads agency offer, here’s a breakdown of how the system works, what the results looked like, and who should treat this as an experiment versus a real acquisition channel.

What ChatGPT ads look like right now

The current ChatGPT ads manager is extremely basic. It’s clearly still in beta, and that shows up in the limited controls, minimal reporting, and lack of optimization levers.

When a user on a free ChatGPT account types a query, they get the usual AI-generated response. Your ad appears below that response, typically at the very bottom of the page. It’s a simple unit: a short title, a brief description, a link, and an optional image. There are no fancy formats, no carousels, and no obvious placement controls yet.

If you want a deeper walkthrough of the early interface and setup flow, you can also check out our first look in this ChatGPT ads manager beta guide.

Campaign setup: objectives, budget, and targeting

Creating a campaign is straightforward, but also very limited compared to platforms like Google Ads or Meta.

Available objectives

Right now, you can choose between:

• Clicks – the only practical objective if you want traffic today.
• Reach – focused on impressions rather than traffic.
• Conversions – visible in the interface but marked as “coming soon” as a bidding option.

You can set up conversion tracking (more on that below), but you can’t yet optimize bidding directly for conversions.

Location targeting

Geographic targeting is also very narrow. You can currently choose from just four countries:

• United States
• Canada
• Australia
• New Zealand

There’s no city, state, or regional targeting yet—just country-level selection.

Budget options

You get two budget types:

• Campaign budget – a total amount for the whole campaign.
• Daily budget – a set amount to spend per day.

It’s easy to accidentally choose the wrong one, and there’s no smooth way to flip an existing campaign between budget types. If you misconfigure it, you’ll likely end up recreating the campaign from scratch.

Bidding and costs: the $3 floor and real CPCs

Bidding is done at the ad group level using a maximum CPC (cost per click). This is where things start to feel a bit odd.

When you enter a bid, ChatGPT gives a delivery indicator:

• At $3.00 per click: “strong delivery”
• At $2.99 per click: “may not deliver”

This strongly suggests there’s a hard floor of around $3 per click on the network. Below that, your ads may barely show, if at all.

In this test, the initial bids were set between $3–$5, but the campaign barely spent. To force the system to actually start delivering, the max CPC was increased to $25. That doesn’t mean you pay $25 per click; it just signals you’re willing to pay up to that amount so the system will start serving impressions.

Once the campaign began spending, the actual average CPC settled around $13 for competitive B2B terms (Google Ads agency-related conversations). That’s still high, but far below the $100–$200+ CPC you might see on Google Ads for the same kind of intent.

How targeting works: “context hints” instead of keywords

ChatGPT ads don’t use traditional keywords. Instead, you give the system “context hints” that describe the types of conversations where your ad should appear.

These hints might include:

• The audience (e.g., business owners, marketing leaders).
• The intent (e.g., looking to hire a Google Ads agency).
• The topics or problems being discussed.

Think of it as writing a short brief for the algorithm: “Show my ad when users are talking about X, trying to solve Y, or asking Z.”

Because this is new territory, there isn’t a proven best practice yet. In this test, the hints were generated with the help of an AI assistant (Claude) to make them more detailed and specific. Even so, there’s no visibility into which queries or themes actually triggered the ads, so it’s hard to know how well your hints are working.

Ad creative and tracking setup

Ad creation is simple and text-focused:

• Title: up to 50 characters.
• Description: up to 100 characters.
• Final URL: your landing page (you’ll want to add UTM parameters).
• Image: optional, for a small visual element.

Because reporting inside ChatGPT is so limited, UTM tracking is essential. Add parameters like utm_source=chatgpt and utm_medium=cpc so you can see performance in Google Analytics or another analytics tool.

There are still some bugs. In this test, adding UTMs initially caused issues with the ad, requiring a fresh ad to be created. Once that was resolved, tracking started working again, but it’s another sign that the platform is still rough around the edges.

Conversion tracking: easy to misconfigure

You can set up conversion tracking under Tools → Conversions. The platform gives you a script to install on your website and a specific event snippet for your conversion (for example, a form submission or registration completed).

The documentation, however, is confusing. It suggests installing the script across your site “preferably near the top” of the head, but doesn’t clearly explain that the event snippet for a specific conversion should only fire on the conversion page (like a thank-you page).

If you paste the full snippet on every page, you’ll end up firing conversions on all visits, completely corrupting your data. In this test, a tracking specialist had to step in to implement it correctly. Expect to do a bit of extra research or involve a developer if you’re not comfortable with tracking scripts.

Real performance from $1,200 in spend

Here are the actual numbers from the initial test:

• Spend: about $1,200
• Clicks: 92
• Average CPC: around $13
• Average CPM: around $120
• Click-through rate: roughly 1% across campaigns
• Conversions: 0 (no form fills from this traffic so far)

That 1% CTR appears to be fairly typical for the current setup. The system seems to settle around that level, at least with the limited controls and placements available today.

One big challenge: the internal reporting doesn’t match analytics data perfectly. Because UTM tracking wasn’t in place from day one, some clicks likely ended up in “direct” traffic or other buckets in GA4. Even with UTMs, you still don’t get granular insights into what’s working.

Reporting and optimization: almost no visibility

The “Insights” tab in the ads manager is extremely barebones. You can see:

• Impressions
• Clicks
• Average CPC by day
• Average CPM by day

That’s it. No breakdowns by query, conversation theme, device, or audience segment. No demographic data. No placement reporting. No search term reports. And no clear way to optimize beyond adjusting bids, budgets, and your context hints.

This makes it very difficult to improve performance over time. You’re essentially trusting the black box and hoping the algorithm is putting your ad in front of the right people.

The hidden catch: ads only show to free ChatGPT users

One of the most important details: ChatGPT ads only appear for free-tier users. Paid subscribers don’t see ads.

That has big implications for B2B advertisers. Many business owners, marketers, and decision-makers are likely paying for ChatGPT Plus or a higher tier. If your ideal customer is a heavy AI user, there’s a good chance they’re on a paid plan—and will never see your ads.

For B2B offers like a Google Ads agency service, that raises serious questions about audience quality. You may be paying premium CPCs to reach a pool of users that skews more casual, cost-sensitive, or less business-focused.

For B2C or e-commerce brands, the story might be different. Free users could still be a valuable audience, especially for consumer products. But there’s no data yet to clearly prove which verticals benefit most.

Is ChatGPT ads worth testing right now?

Based on this early $1K+ test, ChatGPT ads feel very much like an experimental channel, not a mature performance platform.

Who might want to test it now:

• Brands with a dedicated “test” budget and low expectations.
• Marketers who want to be early adopters and learn the system before competitors.
• Teams who are okay with paying for data and experimentation, not guaranteed ROI.

Who should probably wait:

• Businesses that need predictable, trackable results.
• B2B advertisers whose ideal customers are likely on paid ChatGPT plans.
• Anyone who can still profitably scale on Google, Meta, or other proven channels.

The platform will almost certainly evolve. Expect better reporting, more objectives (true conversion bidding), and richer optimization tools over time. We’ve seen this pattern before with other AI-driven ad products, like when Google launched Performance Max and later added more transparency and controls. If you’re curious how AI is reshaping other tools, you might also find our hands-on tests in this Excel AI tools breakdown useful for comparison.

Final thoughts

Right now, ChatGPT ads are simple, expensive on a per-click basis, and very opaque. You get minimal insight into what’s actually driving impressions and clicks, and you’re limited to free-tier users only.

If you have budget you’re comfortable “burning” in exchange for early learnings, it can be a worthwhile experiment. But if you’re choosing between scaling proven channels and funding a black-box beta, most advertisers will be better off putting that money into Google, Meta, or other established platforms—for now.

As OpenAI pushes further into ads, expect rapid changes. The real question isn’t whether ChatGPT ads will improve—they will—but how quickly they’ll become transparent and performant enough to earn a serious share of your marketing budget.