Skip to content

How ChatGPT ads work, and what they cost

Alex Moura Co-Founder and CEO of Digital Star Marketing. Fractional CMO for insurtech and MGA teams, with hands-on experience in paid media and HubSpot attribution. Published · Updated
A woman in her thirties on a sofa looks down at her phone, comparing renters insurance, a cardboard moving box beside her, soft afternoon light. Two marketers at a sunlit desk review printed ad mockups and a pricing sheet with a pen, laptop closed beside them.

A ChatGPT ad is a single card that shows up under ChatGPT’s answer. Adults on the Free and Go plans see it, and it’s matched to what they’re talking about, not to a keyword you bought. You need at least $25 a day. Start with Intelligent bidding, which adjusts bids within your budget. Insurance is US only, and OpenAI approves each insurer one at a time.

ChatGPT ads changed fast in 2026. At launch they were sales-managed only, with big minimum commitments and no conversion bidding. By summer, all of that had changed.

Three defaults catch new insurance accounts: a $100 daily budget filled in for you, objectives that ignore people who saw the ad, did not click, and quoted later, and a setting that lets AI rewrite copy your compliance officer signed.

Who sees ChatGPT ads, and where?

Adults 18 and over on the Free and Go plans see ads, signed in or not.

OpenAI’s ads team confirmed both in its October webinar. Plus, Pro, Business, Enterprise and Edu users see nothing. OpenAI says ads run in 89 countries. For insurance, the policy limits you to the US.

Signed-out users get ads suitable for every audience. Mature categories only reach signed-in adults. Insurance is fine either way.

The ad is one card under the answer, usually introduced by a line ChatGPT writes: most often “This might also interest you,” sometimes a sentence about the conversation, sometimes nothing. You don’t control it. The card holds an image panel (or 4 to 6 images in a multi-image ad), the advertiser’s name, a grey Ad label, a headline and a description, with no web address. On desktop the card also ends with a Learn more link. OpenAI’s help center says one ad unit per response, and that unit may carry more than one image. OpenAI is also testing placements with more than one advertiser, according to Vital Design. OpenAI says the final winning bid sets what you pay. That applies whichever bidding option you use.

Ads never show in conversations OpenAI classes as sensitive, including personal and mental health, politics, privacy and child safety.

The six parts of a ChatGPT ad card: lead line, image panel, advertiser name with the grey Ad label, headline, description and the menu
Where the ad sits and what it holds. The label is always shown.

How does matching work when there are no keywords?

You write context hints: short descriptions of the offer, who it helps, and the situation they are in. OpenAI’s support team told Symphonic Digital that hints “guide matching but aren’t exact-match keywords and don’t guarantee delivery.”

ChatGPT then decides whether your ad fits. OpenAI’s help center lists the signals: the context and intent of the current conversation, your landing page, your title and copy, your context hints and targeting, and, when the person has ad personalization on, signals from their wider ChatGPT use, including past chats, memory and how they treated earlier ads. An auction decides which eligible ad shows. When OpenAI sets your bid, it weighs how likely someone is to click or convert against what that opportunity costs.

The ad can change as the conversation develops. In one live chat the card changed five times in six answers, so a follow-up question is another chance to show up.

The match can also be loose. Brookings found a DoorDash ad on a pad Thai recipe, and ads on “give me tips for work-life balance.” In the insurance sample, 328 of 1,279 ad-triggering prompts used no coverage word such as “insurance,” “policy” or “quote.” About 100 of those named a carrier directly; many of the rest had nothing to do with insurance: Liberty Mutual’s “Switch and Save” on a phone-deals prompt, GEICO on a question about Valvoline oil, and Lemonade’s pet ads on dog-vet questions.

Some loose matches are wasted. The useful ones reach people at a moment that creates a need, such as a move or a new car, before they ask about insurance.

What does it cost?

The minimum is $25 a day. The budget field pre-fills $100, so change it.

Spend can reach twice your daily budget on a given day and seven times it across seven days. A $25 budget can bill $50 on a Tuesday. Tell finance before the first invoice.

When you switch to a target, base it on what you’ve actually been paying per click or per started quote, then compare that with what the result is worth to your business. A lower target won’t automatically make quotes cheaper. It can mean your ad shows less, so watch quote volume as well as cost. The old $3 to $5 starting bid was for Max CPC. Use your own results to choose a target.

WhoWhatCost per click
Workshop Digital, B2B, $2,319 over three campaignsIndependent test, numbers published$3.14 ($3.44 on its financial-services campaign)
Symphonic Digital, B2B, about $500Independent testAbout $9. A $3 bid got zero impressions. It took a $12 max bid to spend.
PPC One, 15 campaigns it compiledAgency roundupMedian $4.41, range $0.76 to $14.80
Digiday, buyers nine weeks after launchTrade reportingCPM $25 to $45, down from about $60 at launch

Billing is postpaid. Your card is charged when unpaid spend reaches a threshold that starts low and rises, or at month end. The country, currency and time zone you pick at setup cannot be changed.

Which objective should an insurer pick?

There are three objectives: Views, Clicks and Conversions. Views pays per thousand impressions. Clicks pays per click. Conversions has two billing options. Billed on clicks, it learns from people who clicked and then converted. Billed on impressions, it also learns from people who saw the ad, did not click, and converted within a day. You cannot switch a campaign’s objective in place. To change it, build a new campaign.

Pick Conversions. If Ads Manager offers it billed on impressions, take that. If your account only offers Conversions billed on clicks, that works too.

ObjectiveYou pay forIt learns fromFor an insurance test
ViewsImpressions (CPM)ViewsNo. You cannot bid on a quote with it.
ClicksClicks (CPC)ClicksOnly until your conversion events are live. Then build a new Conversions campaign. You cannot switch the objective in place.
Conversions, billed on clicksClicksConversions after a clickWorks. Misses the person who saw the ad and acted without clicking.
Conversions, billed on impressionsImpressionsConversions after a click, plus eligible view-through conversions within a dayYes, where your account offers it. OpenAI says it also learns from people who saw the ad and acted within a day without clicking.

Why the last row matters: OpenAI says 52.7% of eligible one-day view-through conversions happen within an hour of the impression. Many people act soon after seeing the ad, without clicking, and billed on clicks the campaign does not learn from them. The figure describes timing. It does not show that the ad caused extra conversions, or that impression billing will perform better for you.

For your first test, start with Intelligent bidding, formerly Maximize Results (Autobid). OpenAI adjusts your bids based on the result you want and the budget you set. Use the test to learn what a started quote actually costs before choosing a target.

From November 9, 2026, if you’re creating a conversion campaign and want to set a cost target, you’ll choose Target CPA instead of Max Bid. New click campaigns will offer Target CPC.

A $50 Target CPA tells OpenAI to aim for an average of $50 per started quote. Some quotes may cost more, others less, and the average can still end up above or below $50. Judge the campaign by what you actually paid for the quotes you got. Target CPC works the same way for clicks. Choosing conversions as your goal doesn’t mean you only pay when someone converts. Depending on your billing setting, you still pay for clicks or impressions.

If you’re already using Max Bid or Max CPC, those settings cap your bids. They don’t set an average cost per result. You can start switching those campaigns to Target CPA or Target CPC on November 9. The old options retire on January 1, 2027, so plan the switch before then and follow the steps Ads Manager shows for your campaign.

What targeting exists?

  • Location. Country, and in the US, state, DMA (a TV market) and ZIP code. Exclusions exist. For an insurer this is the setting that does the compliance work. Include only the states you are licensed in, and do not rely on exclusions alone. Location is approximate, so your quote flow must confirm the applicant’s state before it shows a rate.
  • Platform. iOS app, Android app, iOS web, Android web, desktop web. Leave all five on for a test.
  • Custom audiences. Your own list of emails or phone numbers, hashed. Inclusion needs 25,000 matched people, and OpenAI recommends 100,000. Exclusion has no minimum, so with compliance sign-off, and a privacy notice that covers sharing with an ad platform, a hashed policyholder list can keep your acquisition ad away from existing customers from day one. Not available in the EEA.
  • Context hints. The main lever. Free text per ad group. Chapter 4 covers them.
  • Negative Phrases. Announced October 5 for qualifying advertisers with brand-safety needs. It is not a performance tool, and a new account should not plan around it.

There is no interest targeting and no behavior targeting of the Meta kind, and you never see the conversation. As Search Engine Journal puts it: “Advertisers don’t receive users’ conversations or chat histories.” You see impressions, clicks and conversions by ad, ad group, device and country, and nothing more.

What rules apply to insurance?

OpenAI’s ad policy puts financial services, insurance included, in the US-only, approved-advertiser category. Proof of licensure may be required. Health insurance is restricted to approved US advertisers. Credit repair, debt settlement, debt assistance and alternative investments like bullion are prohibited. OpenAI publishes no verification or review times. Insurance can get extra verification and manual review, so allow a week or more.

Compliance should check three things. First, creative enhancements. As of Friday, October 30, 2026, OpenAI turns creative enhancements on for new and existing campaigns unless you have opted out. They can rewrite and translate your text, resize your image and change your landing page address. Enhanced ads “can run automatically without requiring you to review or approve each version.” Turn them off for the account in Ads Manager, under Manage settings. Then check each campaign, because a campaign’s own setting overrides the account default. Opt-outs you set before October 30 carry over.

Second, location. The Brookings audit found that accounts in lower-income ZIP codes were significantly more likely to receive ads, with no evidence that race played a part. That is the platform’s delivery pattern, not yours. But if you hand-pick ZIP codes, you own the choice. Target by licensed state and let the rest be.

Third, your own record. State regulators can ask what you advertised, where and when. Before each campaign goes live, save a dated PDF of the ad card as it renders, the context hints, the state targeting and the landing page, and keep it with your compliance file. If creative enhancements are ever on, you cannot know every version that ran, which is one more reason to keep them off.

FAQs

Can I see which prompts triggered my ad?

No. You get performance by ad, ad group, device and country. The libraries in Chapter 3 are the only public window into prompts, and they are built with researcher-written questions.

Is there a carousel or video format?

There is a multi-image ad. Since October 30, 2026 it is in open beta for all advertisers: choose Multiple images under Ad format in Ads Manager and upload 4 to 6 square images. Product-feed carousels are for retailers with a catalog. OpenAI reviews the text, the images and the landing page together.

Can a person reply to my ad?

Yes. OpenAI’s help center says a user can message an advertiser through an ad, and the advertiser sees only those messages. The user can also hide the ad, report it, read “About this ad,” or ask ChatGPT about it. Write the card as something you would be happy to answer questions about.

Can my agency run this from its own account?

No. Use one ad account per advertiser, in the insurer’s own legal name, because business verification and licensure checks run against it. The insurer creates it and invites the agency as Admin, Member or Viewer.

Want it run for you? Talk to an insurance marketer who runs ChatGPT ads.

Alex Moura
Written by

Alex Moura

Alex Moura is Co-Founder and CEO of Digital Star Marketing and a fractional CMO for insurtech and MGA teams, with hands-on experience in paid media and HubSpot attribution.

Make your ChatGPT ads count quotes, not clicks.

Tell us what lines you write and where you are licensed. Alex Moura will check your tracking, your quote form and your state plan, then map the first test with you.

Bobby Touran

★★★★★

“Digital Star is hands-on, creative and thoughtful at every step!”

Bobby Touran, CEO and Founder at Rainbow