AI ads 2026: how OpenAI ads in ChatGPT went from a $200,000 pilot to $25 a day
AI ads 2026: OpenAI ads in ChatGPT start at $25 a day. How they work, formats in testing, targeting, bidding, measurement, two campaign cases and budget math.
Linkedist Editorial
By Linkedist Editorial Published: 2026-09-23 · Updated: 2026-09-24 · 13 min read
In AI ads 2026, OpenAI cut the price of entry to ChatGPT from a $200,000 pilot commitment to a $25 US daily minimum, according to Adweek and OpenAI’s Help Center. OpenAI shows these ads only to logged-in adults on the Free and Go plans, per OpenAI’s Help Center, so Plus, Pro, Business, Enterprise and Edu accounts see none. OpenAI said “tens of thousands of marketers have now advertised on ChatGPT”, according to Search Engine Roundtable.
According to Search Engine Journal and PPC Land, testing began in the US on February 9, 2026, and self-serve buying opened on May 5.
Below, we set out the OpenAI ads playbook in seven arguments: the rollout, who sees the ads, formats, targeting and bidding, measurement, two public campaign cases and budget math. By AI ads, we mean ads served inside AI answers, not ads made with generative AI tools. If you are weighing ChatGPT against Google Ads, our ChatGPT Ads vs Google Ads comparison for B2B covers that choice.
The state of OpenAI ads in 2026
Entry price: the US minimum daily budget is $25, per OpenAI’s daily budgets article.
Bid guidance: OpenAI recommends a starting maximum bid of $3 to $5 per click, according to PPC Land. CPMs (cost per 1,000 impressions) fell from $60 at launch to as low as $25 by April 2026, per the same report.
Markets: self-serve Ads Manager opened across 31 European markets on August 31, 2026, per OpenAI, and ads reached more than 60 countries by September 23, 2026, per OpenAI.
Formats in testing: Sponsored Agents, labeled chats with a business’s own agent, are in testing with select US advertisers, according to PYMNTS.
B2B buyers: 61% of business buyers use private AI tools provided by their organization, according to Forrester’s Buyers’ Journey Survey 2025.
First named B2B test: Workshop Digital spent $2,319 on ChatGPT Ads and bought 739 clicks at $3.14 each, with 0 qualified leads.
Key Takeaways
OpenAI matches ads to the conversation through context hints, which its documentation says “are not exact-match keywords”, and places them below the answer, labeled sponsored.
ChatGPT Ads have a hard B2B ceiling: Business and Enterprise seats see no ads, and OpenAI offers no job-title, company or industry targeting.
OpenAI runs a relevance-weighted, second-price auction with Views, Clicks and Conversions objectives, so one account can buy reach, traffic or outcomes.
Formats are moving past a single sponsored unit: a larger image with a call-to-action button and Sponsored Agents are both in testing.
ChatGPT clicks came in cheap in the first named B2B test. Workshop Digital paid 54% less per click than on Google Search, but the test produced no pipeline.
A readable test needs volume. At Q1Media’s cost per click (CPC) of around $5 and conversion rate of roughly 2%, $5,000 a month buys about 20 conversions.
How we built this playbook
This report uses public sources read on September 23 and 24, 2026. Platform rules come from OpenAI’s Help Center, OpenAI’s developer documentation and OpenAI’s announcements. Dates, prices and format tests come from Adweek, PPC Land, Damteq, Search Engine Journal, Digiday, Unite.AI, Search Engine Roundtable, PYMNTS and The Register.
Buyer data comes from Forrester and Ipsos. Campaign evidence comes from Workshop Digital (August 26, 2026), Opascope (June 15, 2026) and Q1Media (July 24, 2026).
Scope is paid placement inside ChatGPT, in US dollars. Parts of OpenAI’s rules were confirmed through search-index copies and trade press. Linkedist has published no paid ChatGPT Ads result, so none is used. We update this report quarterly. Next update: 2026-12-23.
1. ChatGPT became a self-serve ad channel in 2026, moving from a $200,000 pilot to a $25 daily minimum in three months
Between February and May 2026, OpenAI turned ChatGPT from an ad-free product into a self-serve ad platform. Search Engine Journal dated the US test start to February 9, 2026. Adweek reported that OpenAI asked pilot advertisers to commit at least $200,000. PPC Land then reported that the minimum fell to $50,000 in April, before self-serve Ads Manager opened to US businesses on May 5.
The pricing model widened at the same time. The first pilot phase sold on CPM only, and May 5 added CPC bidding for “companies of all sizes”, per OpenAI’s announcement of new ways to buy ChatGPT ads.
Geography followed. PPC Land reported UK ads live on June 6, 2026, and ads in Japan and South Korea from June 22. UK self-serve opened between 19 and 22 June 2026, according to Damteq and PPC Land. PPC Land later reported 31 European markets live on August 24 and India on August 29. Self-serve followed in those 31 European markets on August 31, per OpenAI’s Europe expansion post. Unite.AI dated India’s self-serve to the same day, and by September 23 OpenAI counted more than 60 countries.
In our analysis, the falling barrier matters more than any single price: a team priced out in February could start at the daily minimum by May.
Buyer implication: treat ChatGPT Ads as open to any budget, and plan the first test around learning. Our guide on how to advertise on ChatGPT in 2026 covers account setup.
Bottom line: ChatGPT Ads went from a six-figure pilot to a self-serve channel in under three months, and they now run in more than 60 countries.
2. OpenAI shows ads only to adult Free and Go users, which leaves out the company seats where B2B buyers work
Who can see an ad is decided before any auction runs. OpenAI’s Help Center says ads may appear for logged-in adults on the Free and Go plans, while Plus, Pro, Business, Enterprise and Edu accounts see none. Free users can also choose a no-ads option with extra limits, such as fewer messages per day. The same page adds that ads do not appear near personal health, mental health or politics topics.
Matching and privacy follow published rules. OpenAI’s Help Center lists the topic of the current conversation, past chats and past ad interactions as matching signals. It also says ads run on separate systems from the chat model and do not influence answers. Advertisers never receive chats, chat history, memories, names, emails, precise location or IP addresses, per the same page.
Trust explains the caution. According to Ipsos’s 2026 Consumer Tracker, 63% of US adults think ads will make them trust AI search results less.
The plan exclusions matter most for B2B. In Forrester’s Buyers’ Journey Survey 2025, 94% of business buyers use AI in their buying process and 61% use private AI tools provided by their organization. In our assessment, a large share of serious B2B research now happens on seats no advertiser can buy. A procurement lead in a ChatGPT Enterprise workspace meets vendors only through the organic answer.
Buyer implication: size the paid plan to the Free and Go share of your buying committee. Our guide to ChatGPT Ads for B2B shows how enterprise brands still appear in AI answers.
Bottom line: OpenAI ads reach the individual researcher on a free plan, not the buyer on a company seat.
3. OpenAI ads are moving past a single sponsored unit, with larger images and Sponsored Agents in testing
The standard unit is simple. According to OpenAI’s Help Center guide to the basics, an ad carries the advertiser name, favicon, title, copy, landing page and image. It sits below the response, labeled sponsored, per OpenAI’s announcement of its ads test and its Help Center.
Two formats are now in testing. Digiday reported on May 21, 2026 that OpenAI began “an early test” of a larger image with an optional call-to-action button. Button options include “shop now”, “book now”, “sign up” and “learn more”. Its ecommerce version shows price and customer reviews. Then on September 16, PYMNTS reported that OpenAI is testing Sponsored Agents with select US advertisers. In OpenAI’s words, “a user can choose to start a clearly labeled conversation with a business-sponsored agent in ChatGPT.”
Early participants are named. Digiday reported on September 14 that Wayfair confirmed it is in the Sponsored Agent pilot. OpenAI’s September 16 announcement also added HubSpot and Shopify integrations and campaign creation from written prompts, according to The Register.
OpenAI itself frames today’s product as early. Its CFO, Sarah Friar, called current ChatGPT Ads a “basic starting point” at a Goldman Sachs conference on September 8, per Digiday.
Buyer implication: plan this quarter on the standard unit, with creative that still works if a larger image and button arrive.
Bottom line: the formats you can buy today are basic, and the richer ones are limited tests with select advertisers.
4. OpenAI sells ChatGPT Ads in a second-price auction targeted by context hints, not keywords or job titles
Targeting starts with the conversation. OpenAI’s campaign targeting documentation describes context hints as descriptions of relevant products, use cases or needs, allows up to 2,000 per ad group and says they “are not exact-match keywords”. It also offers geography, platform and custom audiences built from email, phone or device IDs. Inclusion audiences and bid multipliers need at least 25,000 matched users, while exclusion audiences have no minimum. No job-title, company or industry option appears.
The auction is familiar. OpenAI’s Help Center describes a “relevance-weighted, second-price auction” with Views, Clicks and Conversions objectives, billed on impressions or valid clicks. According to OpenAI’s Help Center article on conversion-optimized campaigns, oCPC bills on valid clicks, while oCPM bills on impressions and optimizes on click and view conversions. The objective, billing model and conversion event cannot change after creation. Unite.AI reported that CPC and outcome-optimized bidding account for the majority of campaigns.
Hint style matters. Workshop Digital found keyword-style context hints beat conversational ones, at a 1.2% click-through rate (CTR) against 0.9%.
In our reading, the 25,000-user floor rules out most small B2B account lists as inclusion audiences, though they still work as exclusions.
Buyer implication: start with keyword-style hints, upload customer lists as exclusions, and choose the billing model before launch because it is locked. Our ChatGPT Ads cost breakdown for 2026 covers how bids translate into budgets.
Bottom line: OpenAI ads are bought like search ads but targeted by conversation context, with no way to target a role.
5. ChatGPT ad measurement runs on OpenAI’s pixel and Conversions API, with aggregated, last-touch reporting
Conversion data arrived with self-serve. PPC Land reported that the pixel and Conversions API launched on May 5 to measure “purchases, leads, sign-ups, and other defined actions”, with data provided in aggregated form. Ads Manager reports impressions, clicks, spend, CTR, average CPC, average CPM and conversions, per the same report.
Attribution is configurable but narrow. OpenAI’s Help Center article on measuring results offers a 7-, 14- or 30-day click-through window and a 0- or 1-day view-through window. It uses last-touch attribution, and changing the window affects reporting only, not optimization, bidding or billing. Spend insights may be delayed 7 to 8 hours, per OpenAI’s basics guide.
Server-side tracking needs one extra step. PPC Land reported that Conversions API events must capture the oppref identifier manually, which the browser pixel handles on its own.
In our assessment, a 30-day last-touch window will miss many B2B deals that close in a later quarter.
Buyer implication: send qualified-lead events through the Conversions API, not only page-level pixel fires, and reconcile ChatGPT-sourced leads in your own CRM with UTM parameters.
Bottom line: ChatGPT reports what happens within 30 days of a click, so B2B teams have to measure pipeline themselves.
6. Early public results show cheap clicks and thin B2B lead flow, and the math explains why
The first named B2B test bought cheap traffic and no pipeline. Workshop Digital, in a test Jacob Brown published on August 26, 2026, spent $2,319 across manufacturing, financial services and digital marketing services campaigns. It recorded 69,630 impressions, 739 clicks, a 1.1% CTR and a $3.14 CPC, with 0 qualified leads. The agency’s own Google Search data showed a $6.76 CPC, a 5.2% CTR and 2 qualified leads.
A larger account reads better. Opascope published data on June 15, 2026 from an anonymized live account that spent about $60,000 over 15 days, at a CPC of about $1.72 and a 2.35% conversion rate. Opascope called it “early-mover economics”, not a mature benchmark.
Worked math (our calculation from the published figures, Opascope inputs approximate):
Workshop Digital CTR: 739 / 69,630 = 1.06%, matching the reported 1.1%.
Workshop Digital clicks: $2,319 / $3.14 = 738.5, matching the reported 739.
Opascope clicks: $60,000 / $1.72 = about 34,900 clicks.
Opascope conversions: 34,900 x 2.35% = about 820 conversions.
Opascope cost per conversion: $60,000 / 820 = about $73.
Volume is the difference. In our assessment, 739 clicks were too few to expect many leads at the conversion rates published so far. Opascope’s account bought about 47 times as many clicks (34,900 / 739 = 47.2), and conversions followed.
Buyer implication: judge a ChatGPT test on cost per qualified lead, never on CPC, and fund it at a level that can produce a readable number.
Bottom line: cheap ChatGPT clicks are real, but lead flow shows up only when a test buys enough clicks to measure.
7. A test that can prove anything needs about $5,000 a month, far above the $25 daily minimum
The minimum budget and the useful budget are different numbers. OpenAI’s Help Center lists a $25 US minimum daily budget, averaged over seven days. A single day can spend up to 2 times the daily budget, and a seven-day period up to 7 times, per the same page. Q1Media aggregated live campaign data across verticals on July 24, 2026. It reported CPCs of around $5 and conversion rates of roughly 2%, and it recommended $5,000 to $10,000 a month.
Worked math (our calculation, using OpenAI’s pacing rules, Workshop Digital’s CPC and Q1Media’s rates):
Pacing at the minimum: $25 x 2 = $50 on the busiest day, and $25 x 7 = $175 a week.
Minimum test: $25 x 30 days = $750 a month.
Clicks at Workshop Digital’s $3.14 CPC: $750 / $3.14 = 239 clicks.
Conversions at roughly 2%: 239 x 2% = 4.8 a month.
Q1Media’s floor: $5,000 / $5 = 1,000 clicks, and 1,000 x 2% = 20 conversions a month.
Q1Media’s upper range: $10,000 / $5 = 2,000 clicks, and 2,000 x 2% = 40 conversions a month.
Five conversions a month cannot separate a good context hint from a bad one, but twenty to forty can start to. Q1Media’s rates come mostly from consumer verticals, and Workshop Digital’s result suggests B2B conversion rates may run lower.
Buyer implication: below about $5,000 a month, run a smaller test for click-through and context-hint signal only, and do not expect it to prove pipeline. Our ChatGPT Ads cost guide adds management fees and creative to the picture.
Bottom line: $25 a day opens the account, but by our math about $5,000 a month produces enough conversions to decide.
What this playbook means for three kinds of buyers
If you run B2B demand generation
Start the quarter with a prompt audit, then add a capped ChatGPT test. Forrester’s private-AI-tool figure means company seats carry much of the research, and those seats see no ads. First, run your key buyer prompts in ChatGPT and note where competitors are named instead of you. Then run the ChatGPT Ads test with audience-qualifying copy and the Conversions API sending qualified-lead events. Choose the 30-day click window so late conversions still count. Judge it on cost per qualified lead against your paid search baseline. Keep a job-title channel such as LinkedIn running alongside, because ChatGPT cannot target roles.
If you sell to consumers or run ecommerce
OpenAI’s newest formats point at commerce first. The larger-image test Digiday reported shows price and customer reviews, and the Shopify integration reaches merchants inside a tool they already run. Budget for at least 1,000 clicks a month, and use Opascope’s account as a rough frame rather than a benchmark. Pick the Views objective only when reach is the goal, since it bills on impressions whether or not anyone clicks. Decide early between Clicks and Conversions, because a conversion-optimized campaign cannot change its objective or billing model later. Check the excluded topics too, because ads never run in personal or mental health conversations.
If your buyers sit on company AI seats
Treat paid ChatGPT Ads as a minor budget line. Buyers on ChatGPT Business and Enterprise can meet you only through the answer itself. Put the money into GEO (generative engine optimization), the work of getting a brand cited in AI-generated answers, and into a channel with job-title targeting. Re-run your key prompts monthly to track how often your brand is named before the next budget cycle. If you want outside help, our shortlists of the best ChatGPT Ads agencies in 2026 and the best ChatGPT Ads agencies in the UK compare the options.
How Linkedist runs this playbook for B2B teams
Linkedist is a B2B marketing agency in Vilnius, Lithuania, founded in 2019, that runs LinkedIn marketing plus ChatGPT Ads and GEO, in English and Lithuanian. We sell ChatGPT Ads and AI search optimization (GEO) as two separate services, so one team can run the paid test and the organic work for seats ads cannot reach.
ChatGPT Ads: a capped Ads Manager test with context hints and conversion tracking, sized to the math in argument 7.
AI search optimization (GEO): a prompt audit and organic fixes for AI answers. Our GEO page reports one anonymized SaaS and finance client that AI answer engines, ChatGPT among them, did not mention. By our own reported result, it was “listed by AI” within one month, with +42% inbound inquiries.
LinkedIn advertising: the job-title and company targeting ChatGPT lacks. For cloud services company Cloudvisor, we delivered 225 posts and 10 ad campaigns across two years, a LinkedIn result rather than a ChatGPT one.
We have no published paid ChatGPT Ads case study yet, and we will not quote paid numbers until a client result is public. Google Ads management is not on our service list, so a team whose main need is search ads should keep a search specialist. Our overview of Linkedist’s ChatGPT Ads services sets out the scope.
FAQ
Can you advertise on ChatGPT in 2026? Yes. OpenAI’s self-serve Ads Manager opened to US businesses on May 5, 2026, per PPC Land, and to 31 European markets on August 31, per OpenAI. Unite.AI dates India’s self-serve to the same day. The ads reach only logged-in adults on the Free and Go plans.
How much do OpenAI ads cost? OpenAI recommends a $3 to $5 starting maximum bid per click, per PPC Land, and its Help Center sets a $25 US daily minimum. By our math on Q1Media’s aggregate rates, a test that produces around 20 conversions a month needs about $5,000.
What are Sponsored Agents in ChatGPT? Sponsored Agents let a user open a labeled chat with a business’s own agent after seeing its ad, according to PYMNTS. OpenAI is testing them with select US advertisers, and Digiday reported that Wayfair is in the pilot.
Do ChatGPT Ads influence ChatGPT’s answers? No. OpenAI’s Help Center says ads run on separate systems from the chat model and do not influence answers. Getting named inside the answer itself takes organic work.
How do you measure ChatGPT Ads? Through OpenAI’s pixel or Conversions API, which PPC Land reported launched on May 5, 2026. OpenAI’s Help Center offers click windows of up to 30 days and last-touch attribution. Pipeline beyond that window has to be tracked in your own CRM.
Are AI ads worth it for B2B? Yes as a capped test, not yet as a proven pipeline source. The one named public B2B test, in argument 6, produced no qualified leads, and Business and Enterprise accounts see no ads. In our assessment, pair the test with GEO.
Related reads
Book a free AI visibility audit with Linkedist. We will show where your brand appears in ChatGPT answers today and whether a paid test is worth running.


