ChatGPT Ads: What Paid Buyers Need to Know Before the First Campaign

TL;DR​

• ChatGPT Ads are sponsored chat_card units below ChatGPT answers, labeled "Sponsored" and visually separate from the model's reply. They reach Free and Go users in supported regions; Plus, Pro, Business, Enterprise, and Education do not see ads (OpenAI Help Center). OpenAI expanded self-serve beyond the US test to the UK, Mexico, Brazil, Japan, and South Korea on August 11, 2026 (Testing ads in ChatGPT).
• Early public data (third-party, US-heavy, July 2026): SE Ranking sampled 50,006 commercial prompts and saw ads on ~26%; their own spend test reported ~97k impressions, ~1,263 clicks, and ~1.3% CTR with weak downstream sign-ups. Sensor Tower estimates put active advertisers at ~300 in April → 820+ in July. Treat as directional — not industry benchmarks.
• Targeting is conversational, not keyword-based: you write context hints on ad groups describing the chats where your offer fits. There is no keyword volume or pre-flight CPC forecast; you plan with budgets and post-delivery insights instead.
• Account setup runs through Ads Manager Beta — business onboarding, Persona verification, billing, manual review, then Settings → API Keys for MCP. Campaigns won't deliver until account info (name + logo) and review finish (Account Setup help).
• The OpenAI Ads MCP server on MCPBundles wraps the API for Claude, Cursor, and other agents: account overview, context-hint and copy audits, paused campaigns, geo lookup, creative upload, conversion setup, and delivery reads from chat.

You're evaluating a channel that barely existed in paid media a year ago. ChatGPT Ads started as a US test in early 2026 and is still rolling out market by market.

Northbridge Legal runs Google Search for intake keywords and wants a second line on ChatGPT without another console tab for every standup. Their marketer's first question isn't "where's the button?" — it's whether yesterday's test bought anything worth keeping, and whether the account is still stuck in review.

Worth knowing what you're buying before you touch budget. Also worth knowing what you're not buying (keyword volume, chat transcripts, pre-flight CPC quotes).

What the user actually sees​

When an ad serves, the user gets a tinted card under the assistant's answer, clearly marked Sponsored. It isn't woven into the response text. OpenAI says ad selection runs on separate systems from the chat model, so the sponsored unit doesn't change the organic answer (Testing ads in ChatGPT).

The self-serve creative is the chat_card: title, short body, square image, favicon, destination URL. I've seen teams write 45-character titles that still truncate on mobile — put the promise in the first line.

Who sees ads today: logged-in adult users on ChatGPT Free and Go. Plus, Pro, Business, Enterprise, and Education stay ad-free. Rollout is phased; check Ads in ChatGPT for current regions.

I've also watched someone paste "United Kingdom" into geo targeting and stare at a 400. Location ids from search, not country names — you'll hit that wall fast if you skip the lookup step.

What early public data suggests​

There's no Keyword Planner for ChatGPT yet. What exists instead is a handful of July 2026 studies and OpenAI's own aggregate reporting inside Ads Manager.

SE Ranking's prompt study (US, 50,006 commercial prompts, collected July 23, 2026) found ads on 25.94% of queries — roughly in line with Google's AI Mode ad rate in the same methodology. Most placements were a single sponsored card under the answer, not a multi-ad block. They also flagged that ~14% of ad impressions looked off-topic relative to the prompt, and that ~96% of advertisers weren't cited in the organic answer above the card — paid placement and AI visibility are separate problems.

SE Ranking ran their own campaigns in parallel: 97,000+ impressions, 1,263 clicks, 1.30% average CTR, and very few sign-ups afterward. Their read for B2B SaaS: traffic is cheap-ish, but heavy AI users often sit on ad-free Plus/Pro plans, so the ad-eligible pool may not match your ICP. E-commerce with product feeds and conversion volume might look different.

Business Insider on Sensor Tower data (July 2026, mobile US) reported ChatGPT users on Free/Go seeing roughly twice as many ads per hour as in April, and estimated advertisers grew from ~300 to 820+ over the same window — with mainstream retail brands joining the mix.

Other vendors publish higher ad-appearance rates on shopping-style prompts (Otterly.ai reported ~76% of shopping answers carrying an ad in a US July snapshot). Different question sets, different numbers — useful for trend direction, not for quoting as "the" ChatGPT ad rate.

Inside your own account, OpenAI gives the seven metrics you'd expect after delivery: impressions, clicks, spend, CTR, CPC, CPM, and conversions when measurement is wired. Agency blogs cite anecdotal $20–80 CPM and $1–5 CPC bands; I haven't seen OpenAI publish official benchmarks. Plan from daily minimums and your first week of rows, not a third-party quote.

Getting into Ads Manager​

You can't API your way around account creation. OpenAI ties each Ads API key to one ad account, and that account has to exist in Ads Manager first.

Northbridge's marketer started at ads.openai.com with a business email — not the personal Gmail she uses for ChatGPT Plus on the side. The flow, spelled out in OpenAI's Account Setup guide, looks like this:

Sign up and business details. Sign in with an OpenAI account (create one inline if needed). Enter business name, site, industry, and the country + currency for the ad account. OpenAI treats those as sticky — pick the market you're actually going to spend in.

Identity verification through Persona. OpenAI routes you through Persona for ID and business checks. Have registration docs and a director ID handy; most teams I've talked to clear this in one sitting, then wait on OpenAI's side.

Account info before anything serves. After access is granted, Settings → Account info needs an advertiser name and logo exactly as they'll render in the chat_card favicon slot. Ads won't deliver with blank brand fields — easy to miss if you're rushing to campaign build.

Billing. Add a payment profile before campaigns can spend. Daily minimums apply at the campaign level ($25 USD / £15 GBP in the markets OpenAI documents today).

Manual review queue. Submitted accounts land in review. Days, not minutes. Northbridge got API access while campaigns were still in_review — useful for building paused structure, useless for learning from live delivery.

Country availability moves; OpenAI keeps the live list in Ads Manager Availability. One account owner per business, then invite teammates from Settings when you're ready.

Connect MCP after the account exists​

Once Ads Manager shows your account and review status, grab an API key: Settings → API Keys → Create. Copy it when shown — same discipline as any ads platform secret.

On MCPBundles, open the OpenAI Ads MCP page, paste the key, run the connection test, save. One key, one ad account; there's no account picker on the wire.

Northbridge's first chat prompt after connect: read my ad account — currency, timezone, review status, and whether anything is blocking delivery. That maps to the account read tool and catches the "still in review" state before she builds campaigns.

Context hints replace keywords​

Google Ads matches queries. Someone typed "best crm for agencies" and you bid on that intent.

ChatGPT Ads match conversation context. On each ad group you write context hints — for an accounting product targeting small service firms, try accounting software for service businesses under 50 employees.

Advertisers get aggregated performance: impressions, clicks, spend, CTR, CPC, CPM, and conversion counts when measurement is wired. They never get chat text or user email. Reporting stays thinner than Google Search because of that privacy boundary.

At auction, OpenAI weighs your hints, ad copy, and landing-page relevance to the live conversation (unless the user opted out of ad personalization). There's no Keyword Planner or search volume data; you set CTR and CPC targets after the first week of impressions.

Second-price auction, relevance-weighted. A tight hint on a $3 bid can beat a vague card on $7.

Search volume and CPC bands before spend? Keep Google Ads MCP in the same thread for keyword research — complementary channel, not a substitute for ChatGPT placement.

Campaign structure and geo​

Ads Manager and the API mirror Google Ads hierarchy: account, campaign, ad group, ad.

Campaigns carry objective, budget, schedule, and geo targeting. Ad groups carry theme, context hints, and bid config. Ads carry the chat_card creative.

In the US you can also target states, DMAs, and ZIP codes where supported (Create Campaigns for ChatGPT). Daily minimums vary by market — $25 USD in the United States, £15 GBP in the United Kingdom — with other floors listed in OpenAI's campaign help article. Budgets and bids are stored in micros (one million micros equals one currency unit).

New creatives sit in review until OpenAI approves them. I keep new campaigns paused until review passes; one accidental active toggle on a half-built structure can burn $25 in minutes.

CPM, CPC, and oCPC​

You'll use three objectives for almost all ChatGPT spend (Conversion-Optimized Campaigns).

CPM (impressions) when you want presence in relevant conversations. CPC (clicks) when the card should drive landing-page visits. oCPC (conversions, open beta as of August 2026) still bills per valid click but steers delivery toward a signup or checkout pixel you wired first.

oCPC isn't pay-per-conversion on the invoice — billing stays click-based while the auction favors clicks more likely to convert.

Measurement when outcomes matter​

Delivery reporting shows impressions, clicks, spend, CTR, CPC, and CPM. Conversion reporting needs a conversion pixel, a conversion event setting, and optionally a Conversions API key for server-side events.

Those are separate reads in the API — daily efficiency vs attributed conversion counts — and I treat them as two questions in standup, not one blended dashboard.

A safe first build from chat​

Northbridge's marketer usually starts with account review status, then UK geo ids, then a paused campaign with a £15 daily cap and context hints for B2B SaaS buyers comparing project tools. She uploads the square image first — chat_card units need a file id from upload before the ad exists — and leaves everything paused until review clears.

When search volume still matters, she runs keyword ideas on Google in the same thread before she unpauses anything on ChatGPT. Same assistant, two channels.

The MCP surface also includes one-call triage her standup actually uses: account overview (full tree plus serving flags), context-hint audit, ad copy audit, and a performance audit that flags dead or weak ads over a date window — so she's not chaining six list calls every Monday.

Wire it through the OpenAI Ads MCP when you'd rather do that ladder from Claude or Cursor than Ads Manager.

Official docs worth bookmarking​

Key docs as the product stabilizes: Ads in ChatGPT, Testing ads in ChatGPT, Ads Manager Beta overview, Account Setup, Create Campaigns for ChatGPT, and the Advertiser API overview — that's the surface the MCP tools call.