Flight log / Transmission 001

ChatGPT ads: the complete guide for advertisers (2026)

A new challenger just entered the auction. It doesn't take keywords, it reads the whole conversation, and it went from zero to a reported $1B annualized run rate in under 200 days.

Search ads start with a query. ChatGPT ads start with a conversation. That one difference changes how you target, what you write and what a click is worth.

This is the field guide we wanted when the auction opened: what's confirmed, what's still noise, and how to run a first test without lighting budget on fire. We refresh it monthly because the platform moves faster than most quarterly plans.

ChatGPT ads at a glance

What are ChatGPT ads?

ChatGPT ads are paid placements that appear underneath a ChatGPT response when the conversation is relevant to what an advertiser sells. Each ad carries the advertiser name, a logo, a short headline, a description, an image and a landing page link. OpenAI keeps the ad visually separate from the answer and labels it as sponsored. The answer itself is not for sale.

The closest mental model is a search ad that triggers off a whole conversation instead of a two-word query. Someone who spends three messages working out which CRM fits a 20-person sales team has told the system more about their intent than any keyword ever could. That's the opportunity. It's also why the old playbook only half works here.

Who sees ChatGPT ads, and who doesn't?

Ads show to logged-in adult users on the Free plan and the low-cost Go plan in markets where ads are live. OpenAI states it does not show ads to users on Plus, Pro or any Business plan, or to anyone who tells OpenAI, or whom OpenAI predicts, is under 18. Ads are also kept away from sensitive and brand-unsafe conversations; early coverage of the test cited health, mental health and politics as excluded topics.

For B2B, that means your reachable audience skews toward free users and earlier-stage researchers. Plan the offer accordingly: a useful guide or benchmark will usually outperform a "book a demo" ask on this surface.

How does ChatGPT decide which ad to show?

OpenAI says ads are selected primarily on relevance to the context and intent of the conversation. Advertisers influence that with three inputs:

1. Context hints at the ad group level. Plain-language phrases describing the conversations, topics or needs where your product fits.
2. Your ad copy, meaning the title and description.
3. Your landing page content, which the system also reads.

Eligible ads then compete in a relevance-weighted, second-price auction. Relevance can beat a bigger bid, the same principle Quality Score applies in Google Ads. A sharp ad and a matching page are worth real money here too.

The part that makes search marketers twitch: OpenAI is explicit that context hints are not exact-match keywords or targeting rules and do not guarantee delivery for particular conversations.

How to write context hints that match

OpenAI's own example contrasts "running shoes" with "cushioned everyday running shoes for beginners training for their first 5K." Specific beats short, every time. Describe who the buyer is, what they're trying to do and the moment they're in.

Build one ad group per distinct need, not per product. OpenAI's guidance says to split ad groups when they need different messaging or landing pages. Same discipline that keeps search accounts healthy: tight theme, matching ad, matching page.

How much do ChatGPT ads cost?

Still no rate card. But OpenAI's help center now prints the floor and the starting line, which beats guessing from trade press:

• Minimum daily budget: $25 for USD accounts. It's an average over 7 days, so a single day can spend up to 2x and the week caps at 7x. You can set a campaign-total budget instead.
• Starting bid: OpenAI recommends a $3 to $5 max CPC for Clicks campaigns.
• What you're billed for: impressions on Views, valid clicks on Clicks. Conversions campaigns optimize toward a conversion but bill per click or per 1,000 impressions, your pick.

The price history, with dates attached:

• February 2026 pilot: about $60 CPM, with early commitments reported in the six figures.
• April 2026: reported minimums dropped as self-serve opened, and CPMs were reported as low as $25.

Bidding changed in August 2026. New ad groups now default to Maximize results, which sets bids automatically to get the most clicks or conversions out of your budget. OpenAI is upfront that it doesn't guarantee any CPA, CPC or ROAS target. If you want a hard ceiling, switch the ad group to Manual: Max bid. By the end of August, OpenAI said CPC and outcome-optimized bidding made up the majority of campaigns.

For a first test, the Clicks objective is the safer call: you pay for traffic you can measure instead of impressions on a surface whose click behavior you haven't learned yet. Our judgment: start with a manual max bid inside OpenAI's $3 to $5 range so a learning week can't surprise you, then try Maximize results once you know what a click is worth.

How do you set up a ChatGPT ads campaign?

Self-serve campaigns live in Ads Manager at ads.openai.com. Start the account early: billing has to be in place before anything delivers, and some new accounts can only advertise in their home country until identity verification and a spend threshold clear. The step-by-step setup guide walks every screen; here's the short list.

1. Create the account and add billing. Pick country, currency and time zone carefully (they can't be changed later), then add a card, billing address, business name and invoice email.
2. Confirm your logo. It shows next to your brand name on every ad.
3. Set up measurement first. Install the pixel or connect the Conversions API, and add UTM parameters to every landing page URL. No telemetry, no test.
4. Create a campaign and pick the objective: Views, Clicks or Conversions. Choose carefully. Changing it later means a new campaign. Set a daily budget of at least $25 (USD) or a campaign-total budget.
5. Build ad groups around single needs and write three to five specific context hints for each.
6. Write several ads per ad group. OpenAI's advice is volume with variety: multiple title and description variations per offering, each with a different angle. It doesn't publish character limits, and third-party reports disagree, so trust the counter in Ads Manager. Don't make your logo the main visual; it already sits next to your name.
7. Choose the bid strategy. Maximize results is preselected for eligible new ad groups. Switch to Manual: Max bid if you need a ceiling.
8. Point each ad group at the page that answers that need. More on that below.

ChatGPT ads vs. Google Search ads

Round one isn't a knockout. Google still owns the "I know what I want" query. ChatGPT is winning the "help me figure out what I need" moment that used to sprawl across ten browser tabs. Treat it as an addition to search, not a replacement. The full head-to-head is in ChatGPT ads vs. Google Search ads.

What happens after the click?

Here's the part most early guides skip. A ChatGPT user clicks after a back-and-forth where they've already explained their situation. They land with more context than a searcher, and with less patience for a page that makes them start over.

If someone just asked how to get a 15-person team through a first SOC 2 audit and lands on a homepage that opens with "the leading compliance platform," you've paid for the click and dropped the thread. The page should pick up where the conversation left off: same problem, same audience, same next step. That's message match, and on a conversational surface it matters more, not less.

How should you measure ChatGPT ads?

• Use the Conversions API where you can. Server-side events survive browser privacy changes better than a pixel alone.
• Tag everything with UTMs so ChatGPT traffic is a clean segment in your analytics and CRM.
• Judge on pipeline, not CTR. Click-through on a shiny new placement measures curiosity. Qualified pipeline measures fit.
• Be skeptical of early case studies. OpenAI has shared examples like one ecommerce advertiser reporting 3x return on ad spend over 28 days, and a partner reporting that over 80% of ad-driven traffic came from new customers. Those are single, self-reported data points with no sample or confidence attached.

What we don't know yet

Here's what nobody, OpenAI included, has published yet:

• B2B performance at scale. The public wins so far are ecommerce. There's no credible B2B benchmark yet.
• How hints actually get weighted. OpenAI describes the inputs, not the math. Expect to learn by testing.
• Attribution depth. Aggregated reporting makes multi-touch attribution harder than on search or social.
• Pricing stability. CPMs moved more than 50% in a quarter. That can happen again in either direction.
• How automated bidding behaves on small budgets. Maximize results chases volume. On a niche B2B ad group with thin conversion data, nobody has shown what that does to cost per qualified lead.

Should B2B advertisers test ChatGPT ads now?

Yes, if three things are true: you already run a working search program to compare against, you can track conversions server-side, and you have a few thousand dollars you can treat as tuition.

Pick two or three high-intent needs, write specific context hints, send each to a page built for that need, and run on CPC for four to six weeks. Compare cost per qualified lead against your own search baseline, not against a benchmark from someone else's account.

Still fixing search fundamentals? Fix those first. A new channel amplifies whatever is already broken.

Mission log: ChatGPT ads so far

FAQ