A ChatGPT sponsored result sits below the answer. It never appears inside it.
That boundary is the most useful thing to understand about advertising on ChatGPT, because it decides what money can and cannot buy. It buys a card underneath the response. It doesn’t buy a sentence within it.
Here’s what the unit is, who sees it, how the match gets made, and what it costs to enter.
The Card Sits Below the Answer
The ad is a separate card placed beneath the conversational answer. It’s visually distinct from the response and labelled “Sponsored”.
OpenAI’s stated position is that advertising doesn’t influence the answers ChatGPT gives. The response gets generated on its own, and the card is matched to the conversation around it. That’s why the two sit in separate blocks rather than blending together.
If you’ve bought search ads before, the model that transfers is the shaded ad block above organic results. What doesn’t transfer is the idea that spending more moves you up inside the main content. There’s no inside to move into.
Why the separation is structural, not cosmetic
It holds because the answer and the card are produced by different systems, rather than because a designer drew a border between them. Nothing in the auction reaches back into the text generation.
For a buyer, that means the worst case isn’t a rigged answer. It’s a card sitting next to an answer that recommends somebody else.
What’s Inside the Card
A sponsored result is a small fixed set of fields rather than a freeform creative slot.
- Advertiser: your brand name and favicon, shown as the card’s source.
- Headline: guidance lands near 16 characters, since display truncates.
- Description: guidance lands near 32 characters, for the same reason.
- Image: one visual, with logos discouraged as the main subject.
- Destination: the landing page a click resolves to.
The fields themselves accept more than that, but anything long enough to be cut off is wasted. Sixteen characters is about two words, which is tighter than most paid search formats and catches advertisers out first.
The card is small on purpose
The format doesn’t leave room to make a case. It leaves room to name a product and a reason. Everything else has to happen on the landing page.
Treat the card as a doorway rather than a pitch. Advertisers who compress a value proposition into 16 characters tend to produce something that reads as neither a product name nor a benefit.
The image carries more weight than the copy
With so little text surviving display, the visual does a large share of the communicating. OpenAI discourages using a logo as the primary subject, which rules out the safest default. Show the product or the outcome instead.
Who Sees a Sponsored Result, and Who Never Does
Most coverage buries this, and it matters more than the pricing does.
Ads reach people on the Free and Go tiers. People on Plus, Pro and Business plans don’t see them at all. Neither do users under 18.
Read that against your own buyer list. Say you sell to senior people at companies that pay for software. A good share of the people you most want to reach sit on paid ChatGPT plans, where no sponsored card will ever load.
That doesn’t make the channel worthless. It does mean it reaches the top of your funnel more reliably than the bottom. Size your expectations accordingly rather than assuming parity with paid search.
The exclusion runs the wrong way for B2B
In most ad channels, the people who pay to remove ads are a minority you can afford to lose. Here the paying tiers skew professional, so the exclusion strips out the audience a B2B advertiser is chasing.
Consumer brands face the opposite picture. If your buyer is a person choosing a product for themselves, the free tier is where they are.
Availability is also limited to a short list of countries, and that list has been changing through 2026. Check the current one in Ads Manager before planning a campaign around a market, and see our dated record of every change for how it got there.
How ChatGPT Picks Which Ad to Show
There are no keywords. That’s the biggest single departure from Google Ads, and it changes how a campaign gets built. Our guide to buying and running ChatGPT ads walks through the setup end to end.
You write context hints instead. These are plain-language descriptions of the conversations and situations where your product is relevant, entered at the ad group level. You’re describing a circumstance rather than bidding on a string.
Your hints aren’t the only input. The matching also reads your landing page content, your ad title and your ad copy. A page that’s vague about what you sell will weaken delivery however carefully the hints are written.
Relevance is weighted, not only price
The winner comes out of a relevance-weighted second-price auction. The most relevant ad wins rather than the highest bid, and the winner pays based on the next bid rather than its own.
The practical consequence is worth sitting with. Outbidding a better-matched competitor isn’t a reliable strategy, and improving the match usually costs less than raising the bid.
Hints guide delivery, they don’t guarantee it
Context hints steer matching without promising your card appears in any particular kind of conversation. Plan for a range of placements rather than a target list.
This is the part that frustrates buyers coming from search, where a match type gives you a fence. Here you get influence over delivery, not control of it.
What It Costs to Enter the Auction
Campaigns run on either cost per click or cost per thousand impressions, and both settle through the same auction.
OpenAI has published starting bid guidance of roughly $3 to $5 per click, and the pilot ran at $60 per thousand impressions. Treat both as guidance rather than as the going rate.
Reported real-world figures vary widely between sources. We’d rather give you the published numbers than a market rate we can’t verify. What you pay depends on relevance, competition and the topic of the conversation.
We’ve broken the pricing down in full, including the minimum-spend history and what a first test should cost, in how much ChatGPT ads cost. For a considered view of budget and account structure, our OpenAI ads agency page covers how campaigns get built and measured.
How the Auction Differs From Google Ads
Four habits carried over from paid search will cost you money here.
- Targeting: you describe situations, rather than bidding on strings.
- Fencing: there are no match types, so delivery is steered and not bounded.
- Winning: relevance is weighted, so the top bid doesn’t take the slot.
- Learning: reporting is aggregated, so there’s no query report to mine.
The common thread is that you have less direct control and more dependence on how well you’ve described your product. Search rewards operators who manage structure. This rewards operators who write clearly about what they sell and to whom.
Your landing page became a targeting input
The system reads your landing page when it decides whether your ad fits a conversation. In search, the page affects quality score and cost. Here it helps decide whether you enter the auction at all.
That’s a real change in priority. A thin page hurts delivery before it ever hurts conversion.
What You Get Back as Measurement
Reporting covers impressions, clicks, spend, click-through rate, average cost per click, average cost per thousand and conversions. There’s a conversions API and pixel-based tracking for attribution.
One limit shapes how you read all of it: data comes back aggregated, and individual conversation data isn’t shared with advertisers.
So you’ll learn that a hint delivered and converted. You won’t learn which questions produced it. If your reporting habit is mining query reports for new intent, that habit has nowhere to run here.
A Sponsored Result and an Earned Citation Are Different Units
ChatGPT can surface your brand in two structurally different ways, and they aren’t substitutes.
The sponsored result is the card below the answer. You buy it, it carries a label, and it disappears when you stop paying.
The earned citation is a reference inside the answer itself, produced because the model drew on a source that mentions you. You can’t buy it, it isn’t labelled as advertising, and it persists independently of any budget.
One sits below the boundary and one sits above it. A brand that’s absent from the answer but present in the ad slot pays to stand beside a conversation it isn’t part of.
That’s workable. It’s weaker than being cited in the response.
They fail differently, too
An ad stops the day the card stops serving. A citation decays slowly as sources age and get replaced, which gives you warning and time to act.
What This Changes If You Aren’t Buying Ads
Two things are worth acting on even with no ad budget at all.
- Competitors can appear beneath answers about you: context hints describe situations, not trademarks.
- The answer itself stays unbought: whatever earned you a citation still works.
The arrival of ads didn’t change how citations get earned. It added a second surface underneath them. The work that puts a brand inside the answer is the same work it was before any of this shipped.
Frequently Asked Questions
Do ChatGPT ads appear inside the answer
No. The sponsored card is placed below the response as a separate, labelled block. OpenAI’s stated position is that advertising doesn’t influence the answer ChatGPT generates.
Who sees sponsored results in ChatGPT
People on the Free and Go tiers. Plus, Pro and Business subscribers don’t see ads, and neither do users under 18.
Can I bid on keywords for ChatGPT ads
No. Matching uses context hints, which are plain-language descriptions of relevant conversations written at the ad group level, alongside your landing page, ad title and ad copy.
How much does a ChatGPT sponsored result cost
OpenAI has published starting guidance of about $3 to $5 per click, and the pilot ran at $60 per thousand impressions. Actual cost depends on relevance and competition, so treat published figures as a starting point.
Is a sponsored result the same as being cited by ChatGPT
No. A sponsored result is a paid card below the answer. A citation is a reference inside the answer, earned rather than bought, and it doesn’t stop when a budget does.
Can advertisers see the conversations their ads appeared in
No. Reporting is aggregated, and individual conversation data isn’t shared with advertisers.
Know Which Side of the Line You’re On
Open ChatGPT on a free account and ask the question your buyers ask. Whatever sits in the answer, you earned. Whatever sits below it, somebody bought.
Both are worth having. They aren’t the same purchase, and only one of them keeps working after the invoice stops.