---
title: "ChatGPT Ads (OpenAI Ads) in 2026: How They Work, What They Cost & How to Run Them"
description: "How OpenAI Ads work, real CPM/CPC/CTR benchmarks from live campaigns, how context hints differ from keywords, B2B best practices, and how to run OpenAI Ads alongside Google, Meta, and LinkedIn in Synter."
date: "2026-06-15"
draft: false
category: "guides"
tags:
  - "Guide"
  - "OpenAI Ads"
  - "ChatGPT Ads"
  - "AI Agents"
---

## What ChatGPT Ads Are and How Advertisers Buy Them

A ChatGPT ad is a single sponsored card that appears below the
assistant&apos;s response. It sits visually separated from the
organic answer and carries a "sponsored" label. The card
resembles a product or service listing, showing an item name, a
price, and the sponsored tag. It can feature one or more items from
a single advertiser. OpenAI keeps ads away from sensitive topics
like health, mental health, and politics, and it suppresses them in
Temporary Chats, logged-out sessions, and the ChatGPT Atlas browser.

Only logged-in adults on the Free and Go tiers see ads. Plus, Pro,
Business, Enterprise, and Education users stay ad-free. Your reach
skews toward people who haven&apos;t paid to remove the experience.
OpenAI began

testing ads in the U.S.

on February 9, 2026, expanded to Canada, Australia, and New Zealand
in March. It then announced rollout to the UK, Mexico, Brazil,
Japan, and South Korea in May.

The pilot ran on an impression-based CPM model with a flat rate near
$60 and minimum commitments around $200,000. When multiple
advertisers compete for the same conversation, OpenAI runs a
relevance-weighted second-price auction and shows the ad most
relevant to the chat first. You pay just above the next-highest
effective bid, and relevance acts as a multiplier rather than a
tiebreaker. A tightly written ad on a $3 bid can outrank a generic
ad on a $7 bid.

CPC bidding

launched May 5, 2026

alongside self-serve access at ads.openai.com, with no minimum
spend, a Conversions API, and pixel-based measurement. The minimum
dropped in stages from $200,000 a month down to zero.

Named pilot brands include Target, Ford, Adobe, and Williams-Sonoma,
with agency partners

Dentsu, Omnicom, WPP, and Publicis

building integration paths. The platform crossed $100 million in
annualized revenue within six weeks and onboarded more than 600
advertisers. OpenAI is targeting $2.5 billion in ad revenue for
2026, so the spend is real and the buyer pool is growing fast.

## How to Set Up ChatGPT Ads: Step-by-Step

Start at

ads.openai.com

, which opened to all US businesses on May 5, 2026 with no minimum
spend. The advertiser registers the account in their own name.
Agencies get invited in after the account exists, which trips up
agencies used to standing up accounts on a client&apos;s behalf.

The ownership rule deserves a hard stop before you touch anything.
If you are an agency, do not create the account. Have the client
register it first, then add your team as a user. Account ownership
cannot be transferred retroactively. A misstep here means tearing
down and rebuilding from scratch.


Build the account structure

The hierarchy follows the pattern every paid search manager already
knows. Campaigns sit at the top, ad groups underneath, and
individual ads inside each group. Context hints attach at the ad
group level, so plan your groups around distinct buyer situations
rather than products. One group per landing page works cleanly
because the system reads landing page content as part of its
matching.


Pick a bidding objective

OpenAI gives you two objectives. Reach optimizes for impressions and
bills on CPM, which suits awareness work where you want the
sponsored card seen by people in the right conversation. Clicks
optimizes for traffic and bills on CPC, which fits pipeline and
signup goals. The CPC objective launched May 5, 2026 alongside
self-serve access. Do not run both objectives in the same ad group.


Wire up measurement

Install the conversion pixel on your site and connect the
Conversions API for server-side event tracking. The pixel captures
post-click events like signups and purchases. CAPI backs it up when
browser tracking degrades. OpenAI describes platform-reported
conversions as a leading indicator only, so treat your CRM as the
truth layer.

Tag every click with a consistent UTM scheme. Use
utm_source=chatgpt, utm_medium=cpc or cpm, utm_campaign for the
campaign name, and utm_content for the ad variant. These parameters
persist through the click and give you attribution your analytics
can read.


What you cannot connect yet

The Ads Manager runs as a standalone tool today. Integrations with
Search Ads 360, Skai, and major MMM vendors are not available. Adobe
and Criteo offer integration paths, and CRM connectors for
Salesforce and HubSpot are slated for later in 2026. Until those
ship, you manage OpenAI Ads outside your usual stack or pull the
data through a tool that already supports it.

## What ChatGPT Ads Cost: Real Benchmarks

Synter&apos;s own live OpenAI Ads campaigns give the clearest
picture of what these ads actually cost, because most published
numbers are platform estimates rather than observed spend. Synter
runs five separate geo campaigns from a single account (US, UK,
Canada, Australia, New Zealand), each on impression-based CPM
bidding with conversion events layered on top. The numbers below
come from real per-day delivery windows in early-to-mid 2026.


CPM, CPC, and CTR in Synter&apos;s campaigns

CPM in Synter&apos;s campaigns sits in a $30 to $45 band across the
four largest markets. The US peak day delivered 34,347 impressions
and 116 clicks for $1,410, an implied CPM near $41 and a CPC of
$12.16. The UK ran cheaper. One strong UK day produced 20,277
impressions and 99 clicks for $666, a CPC of $6.73 and a CPM around
$33.

CPC lands between $5 and $13 once a campaign has real volume. Canada
ran $6.54 to $11.37 on its better days. Australia ran $5.34 to
$13.65. CTR across all markets falls between 0.3% and 0.75%, with
roughly 0.5% as the typical figure. That CTR sits below paid search
because the placement is contextual and in-conversation, closer to a
high-intent native band than a typed-query result.

Thin ad groups distort the math. A low-volume Canada day spent $86
for 3 clicks against 2,802 impressions, a $28.62 CPC that means
nothing except that the ad group lacked delivery. New Zealand,
running a $1,000 budget across three ad groups, showed the same
problem. Consolidate ad groups early so each one accumulates enough
impressions to find a stable price.


How Synter&apos;s data compares to published benchmarks

The secondary numbers track Synter&apos;s observed ranges.
OpenAI&apos;s pilot used a flat $60 CPM behind a $200,000 monthly
minimum. The self-serve launch opened a $25 to $60 CPM range that
varies by category (

flyweel.co

). On CPC, OpenAI recommends a $3 to $5 starting bid, and bids under
$3 routinely fail to clear delivery. B2B campaigns run hotter, with
some agencies reporting $8 to $15 CPCs because the buyer pool is
narrower (

justglobal.com

).

The barrier to entry collapsed fast. The pilot minimum was $200,000
per month, dropped to $50,000 in April 2026, then fell to zero when
self-serve opened on May 5, 2026 (

flyweel.co

). A meaningful diagnostic test now runs $5,000 to $15,000 over two
to four weeks, enough volume to see whether your hints and landing
pages clear the auction.

Read these figures as directional, not as platform averages. They
come from one advertiser, Synter, in the paid-media and B2B SaaS
category, and CPMs swing day to day as early fill stays volatile.
Your CPCs will shift with vertical, geo, and creative. Budget in CPM
terms, expect mid-single-digit to low-double-digit CPCs at volume,
and watch starved ad groups inflate the number.

## How Context Hints Work (and How They Differ from Keywords)

Context hints are natural-language phrases you write at the ad group
level to describe the situations your buyer is in, not the products
you sell. OpenAI&apos;s ad system treats them as semantic guidance
to the auction, never as bid terms. The platform is explicit.
"Context hints are not keywords. They do not guarantee
placement in specific conversations and do not use exact-match
logic." (

justglobal.com

)

The auction reads four inputs at once. Your context hints, your
landing page content and URL, your headline and description copy,
and your relevance-weighted second-price bid all feed the same
decision. Get any one of them right and the system matches your ad
to the right conversation.

That structure changes how you write the inputs. A Google keyword
bids on a typed query and offers exact, phrase, and broad match
types. A ChatGPT context hint reads the full conversation a person
is working through and matches probabilistically with no match types
at all.


You also get no audience segments and no user-level data of any
kind, only aggregated performance metrics (

justglobal.com

). The conversation is the targeting signal.


Write hints as problems, not products

The format trips up most advertisers on the first try. "Project
management software" fails because it names a category the user
would never type into a chat. "Manager trying to reduce missed
deadlines and improve team visibility" works because it
describes the moment a buyer hits the problem your product solves.

Write from the user&apos;s perspective in plain conversational
language, never marketing language. Make each hint specific enough
to exclude irrelevant matches without narrowing reach to nothing. A
practical trick. Ask ChatGPT how it describes your product category,
then mirror that phrasing in your hints.


Treat the landing page as a targeting input

The selection system reads your landing page URL and content as part
of the match. A generic homepage that covers ten use cases sends a
muddy signal and weakens your auction eligibility for any single
conversation.

Build one dedicated page per ad group, each one thematically tied to
its hints. If your ad group targets sales forecasting conversations,
point it at a page about forecasting, not your product overview. The
closer the page reads to the conversation, the more often the
auction picks your ad.

| Dimension | Google Ads keywords | ChatGPT context hints |
| --- | --- | --- |
| Match logic | Exact, phrase, broad | Semantic, no match types |
| Targeting unit | Keyword strings | Conversational phrases |
| Guarantee | Bid on specific queries | No placement guarantee |
| Signal source | Typed search query | Full chat context |
| User data | Query plus demographic and behavioral signals | Conversation context only, no demographics or cookies |

## ChatGPT Ads vs. Google Ads: Targeting Compared

Google reads the query a user types into a search box. ChatGPT reads
the full conversation a user is working through with the assistant.
That single difference cascades through every other part of how the
two platforms target, control, and measure ads.

Google gives you exact, phrase, and broad match types, negative
keywords, and a documented auction with Quality Score. You bid on
specific strings and you control which strings you exclude. ChatGPT
offers none of that. Context hints act as semantic guidance, not bid
terms, and OpenAI does not publish its auction model or expose a
keyword-style negative list (

justglobal.com

).

User data is where the two platforms diverge most. Google layers
demographic and behavioral signals on top of the query. ChatGPT
hands advertisers conversation context only, with no demographics,
no browsing history, and no cookie data. Advertisers receive
aggregated views and clicks and nothing else (

openai.com

).


You should run both, at different points in the funnel. Google
captures the user who already knows what to search for and types it.
ChatGPT captures the user still framing the problem, working through
"how do I fix this" before they ever phrase it as a query.
Pair them and you reach the same buyer at two distinct moments, one
where intent is explicit and one where it is still forming.

| Dimension | Google Ads | ChatGPT Ads |
| --- | --- | --- |
| Match logic | Exact, phrase, broad match | Semantic, probabilistic only |
| Signal source | Typed search query | Live conversation context |
| User data | Query plus demographic and behavioral signals | Conversation context only |
| Negative controls | Negative keyword lists | None exposed |
| Auction transparency | Documented, Quality Score | Relevance-weighted, model undisclosed |
| Measurement maturity | Deep, conversion-rich | Early, aggregated metrics |

## B2B Best Practices for OpenAI Ads

Build your campaigns around where the buyer is in their decision
rather than around your product lineup. A buyer asking "how do
I reduce missed deadlines" sits at a different distance from
purchase than one searching "vendor pricing." Map each
stage to a matching intent signal, CTA, and bidding goal, and keep
those goals separate.


Never put awareness and conversion objectives in the same ad group (

Just Global

). The two attract different conversations and force the auction to
optimize against itself.


Write creative that states the problem

Lead headlines with the problem or the outcome, not your brand name.
"Transform Your Business" tells a reader nothing, so cut
it. Write the description as one sentence of value plus a clear
action, and drop superlatives like "#1 platform" or
"industry-leading." OpenAI&apos;s ad system reads your
copy as a matching signal, so vague language costs you both
relevance and clicks.

Use product UI screenshots, dashboards, and workflow visuals instead
of stock lifestyle photography. A picture of your actual results
matches a buyer working through a real problem. A photo of someone
smiling at a laptop matches nothing.


Know which categories can run

SaaS and cloud software, marketing and revenue technology, HR tech,
analytics, security, and business productivity tools all qualify.
Financial services, legal services, healthcare, and data brokerage
stay restricted. Confirm your category before you spend setup time,
because OpenAI also blocks sensitive conversation topics like
health, mental health, and politics from carrying ads at all.


Close the measurement gap yourself

Click-based attribution loses data in low-cookie environments, so
add a "How did you hear about us?" question to your
post-conversion flow. That single field recovers attribution that
UTM tracking misses on AI traffic. CRM connections to Salesforce,
HubSpot, and Microsoft Dynamics are planned for later in 2026 for
closed-loop reporting, so until then, track lead quality downstream
rather than counting form fills.

Run a meaningful test now while the auction stays thin. Just Global
flags the early-mover advantage directly. Competition will rise as
more advertisers register, which means today&apos;s CPCs reflect a
market that hasn&apos;t filled up yet. With

29% of B2B buyers

already using AI tools as their primary discovery channel, the
audience is real even though the auction is young.

| Stage | Intent signal | CTA | Bidding |
| --- | --- | --- | --- |
| Awareness | How do I... / What is the best way to... | Guide, checklist, educational resource | Lower bids, optimize for reach |
| Consideration | X vs Y / How does the category work | Comparison page, free trial, demo | Competitive, optimize for clicks |
| Decision | Vendor pricing / Vendor reviews | Direct offer, consultation | Aggressive, optimize for conversion |

## Early Results: Is Anyone Actually Running ChatGPT Ads?

Yes, and the spend is real. Synter&apos;s June 11, 2026
account report
documented 129,663 impressions and $4,428 in spend across five
markets from a single live advertiser account. Those numbers come
from working campaigns with measured CPMs and CTRs, not a sandbox
demo.

OpenAI&apos;s own pilot data points in the same direction. The
company reports that ads have not measurably dented consumer trust
in ChatGPT, that users dismiss ads at low rates, and that relevance
keeps improving as the model learns which placements fit a
conversation. Fewer than 20% of eligible users see an ad on any
given day, which keeps the experience light enough that most people
don&apos;t notice the ads.

The platform-level signals are louder. OpenAI Ads crossed

$100 million in annualized revenue within six weeks

of opening self-serve access, with more than 600 advertisers
onboard. OpenAI projects $2.5 billion in ad revenue by the end of
2026 and $100 billion by 2030. Demand-side behavior backs the bet,
with 29% of B2B buyers now using AI tools as a primary discovery
channel.


What the data still doesn&apos;t show

Be skeptical of anyone quoting ChatGPT Ads ROAS right now. No named
advertiser has published verified return-on-ad-spend figures. The
public sources carry no vertical-specific CPC breakdowns beyond
Synter&apos;s own first-party numbers. The flat-rate pilot pricing
of

$60 CPM

tells you what early test budgets looked like, not what a SaaS or
retail campaign actually returned.

Synter&apos;s benchmarks fill part of that gap, with one important
caveat. The CTR range of roughly 0.3% to 0.75% and CPCs of $5 to $13
at volume reflect one advertiser in the paid-media and B2B SaaS
category. Your CPMs and clearing prices will move with vertical,
geo, and creative, and early fill stays volatile day to day.

The honest read is that ChatGPT Ads has moved past curiosity into a
market with real money and named partners like Target, Ford, and
Adobe. The performance literature is thin because the platform is
young, so treat first-party data as your best guide and run your own
diagnostic before you trust any published average.

## Managing OpenAI Ads Alongside Google, Meta, and LinkedIn

OpenAI Ads Manager runs in its own window, with no connection to the
rest of your stack. Search Ads 360, Skai, and the major mix-modeling
vendors have not shipped named OpenAI Ads integrations. You pull
impressions and spend from one tab, then paste them into a
spreadsheet next to your Google and Meta numbers to see the full
picture. That breaks the moment you run five geo-split campaigns at
once.

Synter closes that gap
with a named function, pull_openai_ads_performance, that reads
impressions, clicks, spend, CTR, and CPC directly from your OpenAI
Ads account. The same agents that read the data also build the
campaigns. Synter structured its own live account as five country
campaigns (US, UK, CA, AU, NZ) with topic-stacked ad groups beneath
each, and reports every metric in the same view as Google, Meta,
LinkedIn, Reddit, TikTok, X, Amazon, and Spotify.

The track record sits behind it. Synter connects to 43 canonical ad platforms
through direct APIs, has managed more than $26 million in spend
across 600+ advertisers, and documents CPA improvements above 70% (
Synter blog). OpenAI
Ads is one platform inside that set rather than a bolted-on export.
You watch CPC drift on a thin New Zealand ad group and a heavy US
campaign on the same dashboard, then act without switching tools.


Agents that execute, not just advise

Most cross-channel tools stop at recommendations and wait for a
human to click approve. Synter calls its model Autonomous AI
Execution. The agents make the changes inside the ad account
directly, consolidating a low-volume ad group that spiked to a $28
CPC or shifting budget toward the UK campaign that returned a $5.41
CPC. Approval-first legacy tools hand you a list of suggestions and
leave the execution to you. Synter does the work in the account.

The same agents run from the command line. Synter&apos;s MCP server
connects to Claude Desktop, so you can issue plain-language
instructions across X Ads, Spotify Ads, Amazon Ads, and the rest.
Ask it to compare CPM across your five OpenAI Ads markets, or to
lift the daily cap on the campaign with the best CTR, and it
executes against the live account. Running OpenAI Ads next to every
other platform stops being a reporting chore and becomes one
conversation.



Sign up for Synter

## Frequently Asked Questions

### How much do ChatGPT ads cost?

CPM bids run $25 to $60 per thousand impressions depending on category, and CPC bids start at $3 to $5 with B2B clicks reaching $8 to $13 or higher. Synter's live OpenAI Ads campaigns have observed CPCs spiking to $20 to $30 on thin ad groups. As of May 2026, OpenAI dropped the minimum spend to $0, so you can test with a small budget.

### Who sees ads on ChatGPT?

Ads appear only to logged-in adult users on the Free and Go tiers. Plus, Pro, Business, Enterprise, and Education users see no ads at all. Synter accounts for this structural limit when planning reach, because your audience excludes anyone on a paid ChatGPT subscription.

### Can any business run ChatGPT ads now?

Yes. The ChatGPT Ads Manager opened to all US businesses on May 5, 2026 at ads.openai.com with self-serve access. Synter connects directly to that account so you can launch and report without learning a separate dashboard.

### How does ChatGPT ad targeting work?

Targeting runs on conversational context rather than typed keywords. You write context hints at the ad group level that describe the situation a user is working through, and the auction matches your ad to relevant live conversations. Synter manages these hints across geo-split campaigns and topic-stacked ad groups.

### Are ChatGPT ads worth it for B2B?

For B2B, the early-mover advantage outweighs the higher click costs of $8 to $15. A 2025 survey found 29% of B2B buyers now use AI tools as their primary discovery channel. Synter helps you reach those buyers before the auction gets crowded.

### Can I manage OpenAI Ads in the same tool as Google and Meta?

Yes. Synter pulls OpenAI Ads performance through a direct API connection and reports it alongside Google, Meta, and LinkedIn. You run every platform from one interface instead of stitching siloed dashboards together.
