The $500 ChatGPT Spent Before Lunch: A Small Business Tests Paid AI Visibility

The first sign of trouble arrived as a text from the bank.

Five separate charges. All partial amounts. All from OpenAI. Nobody at Out of the Box Advisors had expected that.

The firm runs a small business coaching practice. They have worked with owners since 2012. They also test marketing tools with their own money before advising clients. So when OpenAI opened a beta self-serve Ads Manager in May 2026, they bought in. The logic was sound. Customers now ask ChatGPT the questions they used to type into Google. BrightLocal's 2026 Local Consumer Review Survey, based on 1,002 US adults, found 45% of consumers used an AI tool for a local business recommendation in the past year. That number was 6% the year before.

So the team set a $500 budget and launched a real campaign for their own firm. What happened next is a field report every small business owner should read. They published the whole thing in August 2026, including corrections to their own first conclusions.

> Does paying for visibility in ChatGPT work for a small business right now? Sometimes, but you may not be able to prove it. Out of the Box Advisors spent $500 on ChatGPT ads and watched the budget burn in a little over an hour. OpenAI reported 140 clicks. The firm's own analytics could verify only about 13% of them. A second campaign of $175, paced across a week, delivered 68% of its paid clicks. The difference was not the ad. It was tracking and pacing. Clicks from inside a chat app often lose their referrer label. They land in your analytics as Direct traffic, unnamed and easy to miss. So the honest answer is this. Paid visibility inside ChatGPT is real, early, and messy. Treat it as an experiment with a budget you can afford to lose. Verify every click in your own tools before you scale anything.

Why did the clicks I paid ChatGPT for never show up in my analytics?

Most often, the clicks did arrive, but their label fell off on the way. Chat apps strip referrer data, so paid visits land as Direct traffic in Google Analytics. Tracking tags can fail too. The fix is simple. Isolate the campaign date and watch Direct traffic for new visitors.

This is the strangest part of the whole story. The first campaign looked like a total ghost. OpenAI's dashboard reported 140 clicks. Wix and Google Analytics 4 showed almost nothing. The campaign data took about a day to appear at all. Meanwhile the $500 was already gone.

Then the firm did something rare. They went back and checked their work. With the campaign date isolated, their Direct traffic roughly tripled that day. Every one of those visitors was new. So the clicks had landed. They were just invisible, filed under a label that says nothing.

Their tagged links were supposed to fix that. The UTM tags did not survive the trip. This matters far beyond ads. If a customer taps a link inside ChatGPT, Perplexity, or Gemini, the referrer often dies inside the app. Your analytics quietly call it Direct. You conclude AI sends you nothing. You may be wrong.

The billing added its own confusion. Five partial charges hit the card instead of one clean transaction. The firm checked OpenAI's help docs, which note that multiple charges can be pre-authorizations. The later reporting showed the charges were real campaign spend.

How do I know if ChatGPT or Perplexity actually sends me customers?

Check two places. First, filter your analytics for referrers like chatgpt.com and perplexity.ai. Second, watch Direct traffic on days you run AI campaigns. A spike of new visitors without referrers is often your AI traffic. Tag every link, then click it yourself and confirm the tag survives.

This habit matters on Google too, not just in chat apps. Pew Research Center analyzed real browsing data from March 2025 and found users clicked a traditional result 8% of the time when an AI summary appeared, versus 15% without one. Traffic is shrinking and getting harder to label at the same time. Measurement is no longer optional.

The second campaign proved the pattern. Same landing page, $175 budget, working tags this time. Sixty-eight percent of paid clicks arrived on the site. Measured by cost per real visitor, the second run was roughly six times more efficient than the first. The only visible difference was pacing. The first campaign burned its budget in about an hour. The second spread it across a week. OpenAI's reporting never explained why. The firm's own numbers did.

One more detail is worth stealing. Their site gets about 1,000 visitors in a typical month. Days after publishing their report, they put a modest boost budget behind it on an ordinary platform. That traffic showed up the same day, three to four times the normal baseline. Same website, same analytics. The instruments work. They caught one channel instantly and missed the other entirely.

What should I check before I pay for visibility on ChatGPT?

Three things. Confirm your tracking catches traffic, pace the budget over days, and spend only what you can afford to lose. The platform itself can look great on paper while your own numbers tell a different story.

Credit where it is due. After the first test, OpenAI shipped conversion tracking, location targeting, and daily budget controls. Three of the five gaps the firm listed got fixed within months. That is fast, and it is a real sign the channel may mature into something worthwhile.

But maturity is not here yet. The single most important feature of any ad platform is trust in the numbers. When the platform says 140 clicks and your own tools can account for fewer than 20, you are buying a story, not traffic. A $3.50 cost per click sounds wonderful. The firm paid it. A cheap click that never verifiably arrives is not cheap. It is just expensive in a quieter way.

There is also a free option worth more than any ad right now. ChatGPT, Perplexity, and Google's AI Overviews all prefer to name businesses they can verify from public sources. Plain facts on your website, consistent listings, specific reviews. That work compounds. Ad spend stops the day you stop paying.

The coaching firm said it best in their own verdict. Boring channels that pay the mortgage beat exciting ones that eat it. New platforms deserve small, careful tests. Not faith.

I watch how businesses get named across ChatGPT, Perplexity, Gemini, and Google's AI Overviews, and I use Ozvor to check my own visibility in minutes. If you want that same clear picture of yours, start at ozvor.com.