For about three years the answer to "can I advertise inside ChatGPT" was no. That changed in 2026, when OpenAI opened a self-serve Ads Manager and Canada came in as a supported market. A local business can now buy placements in the product its customers are already asking for recommendations.
Which raises the question everybody asks second: is it worth anything yet?
The thing itself
A ChatGPT ad is a sponsored placement shown underneath an answer, matched to what the conversation is about. It's labelled as an ad. It doesn't appear inside the assistant's reply, and the assistant doesn't say your name because you paid it.
Three details do most of the work in deciding whether this is for you:
| How it works | |
|---|---|
| Who sees it | Free-tier users. Paying subscribers don't see ads at all |
| Where it sits | Beneath the answer, labelled, visually separate from the reply |
| What it matches on | The context of the conversation, not a keyword somebody typed |
That first row is the one people miss, and it cuts both ways. You're excluded from the segment most willing to pay for software, which for a B2B SaaS company is a real problem. For a roofer, a dentist or a law firm, the free tier is closer to "most people", and the exclusion barely registers.
Targeting without a keyword
Search advertising has a shape everyone knows. Somebody types a string, you've bid on that string, an auction runs, your ad appears. Every part of that assumes a keyword exists.
In a chat there isn't one. Somebody describes a situation over three messages, changes their mind halfway, and never types anything you could have bid on. So the matching happens against the context: what this person appears to be trying to do.
Practically, that changes what you write in the brief. You aren't listing the phrases you want to buy. You're describing the moments where your business is the right answer, plus geography and budget. "Someone comparing quotes for a bathroom renovation in Mississauga and worried about being overcharged" is a usable brief. "bathroom renovation mississauga" is not, because nobody typed it.
What nobody can tell you yet
Here's where most articles on this would hand you a benchmark. We're not going to, because there isn't an honest one.
The platform is months old. Formats are changing, the controls are changing, and there's no stable public body of cost data across verticals the way there is for Google Ads after fifteen years. Any cost per click or cost per lead you read for this channel right now is either one advertiser's account or a guess with a chart around it.
What that means in practice is not "avoid it". It's that the test has to be built so the answer is cheap either way: a small capped budget, one clearly defined offer, a landing page that matches the promise, and a way to tell a ChatGPT-sourced enquiry from every other kind. If you can't do that last part, don't start, because you'll spend the money and end the quarter with an opinion instead of a number. Our AI SEO and analysis service exists partly for that reason, and the attribution problem is worth reading about first.
Should you run one
Three honest cases.
Yes, test it now if your customers are consumers, your offer is simple enough to explain in a card, and your Google cost per lead has climbed far enough that a new channel is worth a month of budget. Being early on an ad surface your competitors haven't opened an account on is a genuine, if temporary, advantage.
Not yet if you sell to businesses whose staff are on paid ChatGPT accounts, or if your attribution is loose enough that you couldn't prove the result afterwards. You'd be buying a plausible story.
Do the other half first in almost every case. Ads buy attention for as long as you pay. The work that gets you named in the answer keeps working after you stop, and it pays off across Claude, Perplexity and Copilot too, not only the platform you bought from. If you only have budget for one, that's the one with the longer tail. It's also why we run both halves as one programme rather than selling the ads on their own.
The part that isn't optional either way
Whichever you choose, the ad sends people to a page, and the page has to make the same promise the ad did. That's an old rule and it doesn't change here. What does change is who else is reading it: the same page is what an assistant parses when deciding whether to recommend you unpaid, so a landing page that converts well and says nothing checkable is a wasted asset on one of the two surfaces.
If you're already spending on Google and Meta, this is a line item, not a strategy. Treat it as a test with a budget cap and a date to review it, and it costs about what finding out should cost.
The short version: the placement is real, it's buyable in Canada today, and it's early enough that nobody can quote you a reliable benchmark. The recommendation sitting above it is the more valuable of the two, and paying for the ad does not buy it.
Frequently asked
Can you advertise on ChatGPT?
Yes. In 2026 OpenAI opened a self-serve Ads Manager, and Canada is a supported market, so a local business can buy placements. A ChatGPT ad is a labelled sponsored placement shown beneath an answer, matched to what the conversation is about. It doesn't appear inside the assistant's reply, and the assistant doesn't name you because you paid. Paying subscribers don't see ads at all, so the audience is free-tier users.
Who sees ads in ChatGPT?
Free-tier users only. Paying subscribers don't see ads, which cuts both ways. For a B2B software company it excludes the segment most willing to pay for tools. For a roofer, a dentist or a law firm, the free tier is closer to "most people" and the exclusion barely registers. Whether that audience fits your customers is the first question to settle before budgeting anything.
How does targeting work in ChatGPT ads without keywords?
Matching happens against the context of the conversation, meaning what the person appears to be trying to do, rather than a string they typed. So the brief describes the moments where your business is the right answer, plus geography and budget. "Someone comparing quotes for a bathroom renovation in Mississauga and worried about being overcharged" is usable; "bathroom renovation mississauga" is not, because nobody typed it. The closest Google Ads analogy is a well-targeted contextual placement rather than Search.
How much do ChatGPT ads cost?
Nobody can give you an honest benchmark yet. The platform is months old, formats and controls are still changing, and there's no stable public body of cost data across verticals the way there is for Google Ads. Any cost per click or cost per lead you read right now is one advertiser's account or a guess with a chart around it. Build any test so the answer is cheap either way: a capped budget, one clear offer, and tracking that isolates ChatGPT enquiries.
Should my business test ChatGPT ads?
Test now if your customers are consumers, your offer fits in a card, and your Google cost per lead has climbed enough that a month of budget on a new channel is worthwhile. Hold off if your buyers use paid ChatGPT accounts or your attribution couldn't prove the result afterwards. In almost every case, do the unpaid work first: getting named in the answer keeps working after you stop paying and carries over to Claude, Perplexity and Copilot.
Filed under