TL;DR: ChatGPT Ads are sponsored cards that show up below a ChatGPT answer, matched to what the person is talking about rather than to a keyword they typed. Anyone in the US and dozens of other countries can now buy them in a self-serve Ads Manager, with a $25 minimum daily budget and a suggested starting bid of $3 to $5 per click. For B2B SaaS there is one big catch: ads are never shown to people on ChatGPT Plus, Pro, or any Business or Enterprise plan, so a buyer researching on a company seat will not see you. The first public tests are mixed. OpenAI’s own case studies (Stream, Jotform) report strong returns, while two independent B2B agency tests got cheap clicks and zero qualified leads. Treat it as a small, well-measured experiment next to Google and Microsoft Ads, not a budget shift.
This guide is built from OpenAI’s own announcements and Ads Manager help center, the advertiser case studies OpenAI has published, two independent B2B tests that shared their full numbers, and Google’s and Microsoft’s own documentation for the comparison. Everything was checked on the original page in October 2026. ChatGPT Ads is still labeled beta, so expect details to change.
In this guide: what ChatGPT Ads are, how targeting and pricing work, what B2B SaaS teams can and cannot reach, how it compares with Google Ads and Microsoft Ads, what real tests found, a setup checklist, and seven experiments worth running first.
What ChatGPT Ads are (and what they are not)
A ChatGPT ad is a labeled “sponsored” unit placed under ChatGPT’s answer. According to OpenAI’s advertiser help center, each ad has six parts: the advertiser name, a favicon, a title, a short description, an image, and a landing page. A new visual format, announced on October 5, 2026, is being tested next to ChatGPT’s image generation with a small group of US advertisers.
Three rules from OpenAI’s published ads principles shape everything else:
- Ads do not change the answer. The organic response is generated the same way whether or not an ad runs under it. You cannot pay to be recommended inside the answer itself, which is why AI search visibility (GEO) and paid placement are two separate jobs.
- Advertisers never see conversations. You get aggregate numbers (views, clicks, conversions), not chat transcripts or personal details.
- Some topics are off limits. Ads do not show to users who say or are predicted to be under 18, or next to sensitive topics such as health, mental health, and politics.
Who sees ChatGPT ads
Only logged-in adults on the Free and Go plans. OpenAI’s help center says ads are not shown “to users on Plus, Pro, or any Business plan,” and the February launch post adds Enterprise and Education to that list. Free users can also turn ads off in exchange for fewer daily messages, per OpenAI’s testing announcement.
How we got here: the 2026 timeline
| Date | What happened | Source |
|---|---|---|
| Jan 16, 2026 | OpenAI announces plans to test ads on Free and Go in the US and publishes its ads principles. | OpenAI |
| Feb 9, 2026 | Test goes live in the US for logged-in adults on Free and Go. Buying is CPM only, through a small group of advertisers. | OpenAI |
| Mar 26, 2026 | OpenAI reports no impact on consumer trust metrics and low dismissal rates, and announces pilots in Canada, Australia and New Zealand. | OpenAI |
| May 5, 2026 | Beta self-serve Ads Manager opens, CPC bidding added, Conversions API and pixel launched. Agency partners include Dentsu, Omnicom, Publicis and WPP. | OpenAI |
| Aug 11, 2026 | Ads launch in the UK, Mexico, Brazil, Japan and South Korea. | OpenAI |
| Aug 18, 2026 | Expansion across Europe announced on the ChatGPT Ads blog. | ChatGPT Ads blog |
| Oct 5, 2026 | Visual ad format test, new attribution and incrementality partners, Negative Phrases and brand-suitability pilots with DoubleVerify and IAS. | OpenAI |
How targeting, bidding and pricing work
If you have run Google Ads, the account structure will look familiar. The targeting will not.
Account structure
OpenAI’s quickstart uses three levels: a campaign (objective, budget, dates, countries, optional platforms and custom audiences), an ad group (one theme, plus “context hints”), and the ad (title, copy, image, landing page). Ads go through review before they serve.
No keywords: context hints instead
There is no keyword list and no match type. The system reads the conversation and picks the ad it expects to be most relevant, using the landing page, the ad text, your context hints, and (when the user allows ad personalization) signals from their broader ChatGPT use. OpenAI is explicit that hints “are not exact-match keywords” and “do not guarantee delivery.”
OpenAI’s context hints guide asks you to describe what you offer, who it helps, and when it is useful, one idea per hint. Its own example for a software company is: “Bookkeeping software for growing retailers that need to reconcile online and in-store sales.” It also warns that “wanting to reach a customer group does not, by itself, make your offering relevant,” and that a hint like “show this ad only to people in Chicago” does nothing, because hints are descriptions, not targeting rules.
Objectives and billing
| Objective | You pay for | Good for | Notes |
|---|---|---|---|
| Views | Every 1,000 impressions (CPM) | Reach and awareness tests | How the pilot started. In two independent B2B tests, CPMs worked out to roughly $33 to $41. |
| Clicks | Each valid click (CPC) | Traffic and first tests | OpenAI suggests a $3 to $5 starting max bid. |
| Conversions, click billing (oCPC) | Each valid click | Sign-ups, demo or lead forms | Optimizes toward one standard event (purchase, sign-up, lead). Custom events not supported yet. |
| Conversions, impression billing (oCPM) | Every 1,000 impressions | Longer journeys where people come back later | Counts view-through conversions as well as clicks. |
Behind this is what OpenAI calls a “relevance-weighted, second-price auction”: a more relevant ad can beat a higher bid. Two settings cannot be changed after launch: the campaign objective and, for conversion campaigns, the conversion event. To change either you clone or rebuild the campaign, per the campaign setup guide and conversion-optimized campaigns guide.
Budgets and minimums
- Minimum daily budget: $25 for USD accounts (15 EUR, 15 GBP, 25 CAD, 25 AUD).
- Daily budget is a weekly average. A single day can spend up to 2x the daily amount, but a Sunday to Saturday week will not exceed 7x.
- New self-serve accounts may only be able to advertise in their home country until identity verification is approved and they have some spend history.
- Self-serve availability: the billed legal entity has to be based in one of the listed countries, which now include the US, Canada, the UK, most of the EU, Australia, Japan, India and others.
Targeting controls you do have
Countries, plus states, cities, markets and postal codes where available; platforms (Android app, Android web, desktop web, iOS app, iOS web); and custom audiences. A custom audience used for inclusion needs at least 25,000 matched users, which rules out most small account-based lists. Exclusions can be smaller, so uploading current customers as an exclusion list is a realistic first use.
The B2B SaaS reality check: who you can and cannot reach
For B2B, these limits matter more than the click price.
The segments where this works best so far share a pattern: a self-serve or product-led motion where a sign-up happens on the first visit, and buyers who describe a concrete task to ChatGPT (“how do I add video calling to my app”, “I need a form that collects payments”). Developer tools, form and workflow software, and SMB-focused SaaS fit that. A six-figure enterprise platform with a sales-led cycle mostly does not, at least not as a lead source.
ChatGPT Ads vs Google Ads vs Microsoft Ads
All three now place ads in or around AI answers, but they work differently. The table uses each platform’s own documentation: OpenAI’s help center, Google’s “About ads and AI Overviews” page, and Microsoft’s April 2026 Copilot announcement and LinkedIn profile targeting help page.
| ChatGPT Ads | Google Ads | Microsoft Advertising | |
|---|---|---|---|
| How ads are matched | Conversation context, ad content, context hints | Keywords and match types, or keywordless (AI Max, Performance Max, Dynamic Search Ads) | Keywords and match types; AI Max for Search pilot expands matching |
| Where AI placements show | Below ChatGPT answers (Free and Go users only) | Above, below or inside AI Overviews | Copilot Search and Copilot Answers via AI Max; Offer Highlights in Copilot (retail) |
| Can you target or exclude the AI placement? | It is the only placement | No. You cannot target or opt out of AI Overview placements | Opt in through the AI Max pilot |
| B2B audience targeting | None by role or company; custom audiences need 25,000 matches | Customer Match, in-market and custom segments | LinkedIn profile targeting by company, industry and job function, plus new seniority levels in pilot |
| Reporting for AI placements | Full, it is all AI placement | None segmented: ads inside AI Overviews report as “Top ads” | Search term and asset reporting for AI Max |
| Restricted categories | Health, mental health, politics, under-18 users | No ads inside AI Overviews for finance, healthcare, politics and other sensitive verticals | Standard Microsoft Advertising policies |
| Entry cost | $25/day minimum; suggested CPC $3 to $5 | No platform minimum | No platform minimum |
What this means in practice
- Google Ads stays the base for B2B demand capture. People searching “SOC 2 compliance software pricing” have told you exactly what they want. In Workshop Digital’s test, Google Search produced 2 MQLs in the same period ChatGPT Ads produced none.
- Microsoft Advertising is the underused B2B option. It is the only search platform where you can layer LinkedIn company and job function data on search ads, and its AI Max pilot puts the same campaigns in front of Copilot users. For many B2B SaaS teams, a properly set up LinkedIn and Microsoft program should come before ChatGPT Ads.
- ChatGPT Ads reaches a moment the others miss: someone describing a problem in full sentences, often before they know which product category to search for. That makes it closer to a consideration channel than a search channel, which is how both Jotform and Workshop Digital ended up describing it.
What real tests show so far
There are two kinds of public results, and they point in different directions. OpenAI publishes customer stories: real campaigns, but selected and written by the platform as marketing, with the advertiser’s own attribution and no detail on trial quality or paid conversion. A few agencies have published their own tests with full numbers, including the bad ones. Read both.
OpenAI’s published B2B SaaS case studies
Independent B2B tests
Average cost per click in two published B2B tests. Sources: Workshop Digital (Aug 26, 2026) and Windmill Strategy (Sep 2, 2026). OpenAI suggests starting at a $3 to $5 max CPC.
Click-through rate in the same two tests. Lower CTR than search is expected: the ad sits under an answer the person is already reading.
How to read these results
- Clicks are cheaper than search, for now. Both independent tests paid well under their usual Google Search CPC. Early auctions are thin; expect this to rise as more advertisers join.
- The positive results have three things in common: they sold a product you can start using on the first visit (an API, a form builder), they wrote ads around specific tasks, and they sent clicks to task-specific landing pages instead of a homepage or a generic “contact us.”
- Service businesses and long-cycle B2B struggled. Both zero-lead tests were lead-generation campaigns for B2B services, judged on contact or lead forms. That is the hardest possible case for a channel that cannot target by job title.
Where to start: a setup checklist for B2B SaaS
oppref to your landing page URL; the measurement guide says to keep it through redirects. For B2B this is where tracking usually breaks: form tools, scheduling links and cross-domain redirects often drop URL parameters. If you get UK or EU traffic, load the pixel only after consent; OpenAI’s terms require the notices and consents your local law asks for.Context hint examples for SaaS
These are illustrative, written in the style of OpenAI’s guidance. Each describes one need, a customer and a situation, rather than listing keywords:
- “Usage-based billing software for API companies moving off flat monthly plans”
- “Help desk tool for Shopify stores handling under 1,000 tickets a month without a full-time support team”
- “SOC 2 automation for startups preparing for their first enterprise security review”
- “Scheduling and dispatch software for HVAC companies with 5 to 20 technicians”
Seven experiments worth running first
Run one variable at a time and give each test enough budget to produce real sign-ups, not just clicks. At the $25/day floor, two campaigns running for two weeks cost about $700 before you learn anything, so plan for a month.
| Experiment | Question | Setup | Measure | Evidence so far |
|---|---|---|---|---|
| 1. Context hint style | Do specific, keyword-like hints beat conversational ones? | Two ad groups, same ads and landing page, different hints | CTR, CPC, sign-up rate | Workshop Digital saw 1.2% vs 0.9% CTR |
| 2. Use case vs generic | Do task-specific ad groups and pages convert better? | Generic, brand and use-case ad groups in one campaign | Sign-up rate, cost per sign-up | Jotform: use-case pages lifted sign-ups 2.5x |
| 3. Landing page | Where should the click go? | Same ad to: use-case page vs pricing vs free tool or template | Session rate, sign-up rate | Watch clicks that never load the page |
| 4. Buying model | Is CPC, oCPC or oCPM cheapest per sign-up? | Clone the campaign (objective cannot be edited) and run side by side | Cost per sign-up, view-through share | Needs a frequent conversion event |
| 5. Channel match-up | How does ChatGPT compare with Google, LinkedIn or Microsoft? | Same offer and budget per channel for the same dates | Pipeline and revenue, not clicks | Stream used about $10,000 per test group |
| 6. Geo holdout | Do ChatGPT Ads create new demand or take credit for existing demand? | Run in matched states or metros, hold out others, compare sign-ups | Incremental sign-ups and pipeline | OpenAI is piloting this with Haus, Measured and WorkMagic |
| 7. Headline framing | Brand name first, or problem first? | Two ad sets differing only in the title | CTR, sign-up rate | Brand-first was slightly better in one test |
If you only run one test
Start with experiment 2: one campaign on Clicks, two or three use-case ad groups, each pointing at its own landing page, with current customers excluded. It answers the first real question (does anyone from ChatGPT sign up and use the product?) at the lowest cost. Add the buying-model and channel tests only after that answer is yes.
Set pass and fail rules before you spend
Write these down before launch so the result is not argued after the fact:
- Traffic check (first 1 to 2 weeks, roughly $350 to $700): do at least 60% of ad clicks show up as real sessions in GA4, and is engagement close to your site average? If not, fix the landing page or the ad before judging anything else.
- Conversion check (30 days): cost per activated trial or qualified sign-up compared with your non-brand Google Search number for the same period. Within about 1.5x of search is worth continuing; more than 3x with no sign of improvement is a pause.
- Pipeline check (60 to 90 days, larger budget): trial-to-paid or demo-to-opportunity rate for ChatGPT sign-ups against other channels. A month of minimum spend cannot answer this question for a product with a long trial; plan budget for it separately.
These thresholds are our starting suggestions, not platform benchmarks; OpenAI has not published any. Adjust them to your own margins.
Measure the right thing
Judge the channel on what reaches your CRM: trials that activate, demos that happen, pipeline. A useful early check from Stream’s case study is the share of ad clicks that turn into actual website sessions. A low share points to accidental clicks or a slow landing page before you ever get to lead quality. If you sell to businesses, add a “how did you hear about us” field with ChatGPT as an option; self-reported attribution catches people who saw an ad, left, and came back through Google later. Also watch for low-quality sign-ups: duplicate trials, throwaway email domains and accounts that never complete a first action. Exclude them before you calculate cost per sign-up.
Should your SaaS company test ChatGPT Ads now?
| Test now if… | Wait if… |
|---|---|
| You have a free trial, freemium plan or self-serve sign-up | Every deal starts with a sales call |
| Your buyers describe concrete tasks (build, automate, fix, compare) | Your buyers mostly research on company ChatGPT Business or Enterprise seats |
| You already track sign-ups with clean UTMs and a CRM | You cannot tie a click to a lead or a deal yet |
| Google and Microsoft search campaigns are already well run | Your search campaigns are not profitable yet |
| You can afford $750 to $1,500 for a month of learning | Every dollar has to show return this month |
If you are on the “wait” side, the better use of the same effort is usually getting recommended inside the answer itself, through AI SEO and GEO work, and fixing your Google Ads program. Those reach the same buyers, including the ones on company seats who will never see a ChatGPT ad.
The bottom line
ChatGPT Ads went from announcement to a self-serve platform in about four months, and the setup is now simple enough to test with a small budget. The early public numbers show cheap clicks, low click-through rates, and a clear split in results: product-led SaaS with task-specific ads and landing pages (Stream, Jotform) reported strong returns, while contact-form B2B services saw no leads. Start small, track to pipeline, test one variable at a time, and remember that buyers on company ChatGPT seats cannot be reached with ChatGPT Ads at all. In ChatGPT itself, only organic AI visibility reaches them.
Data Sources: OpenAI announcements (Jan 16, Feb 9, with updates through Aug 11, May 5, Oct 5, 2026); OpenAI Help Center articles on ad basics, campaigns, context hints, conversion optimization and conversion measurement; OpenAI customer stories for Stream, Jotform and Angi; independent tests by Workshop Digital and Windmill Strategy; Google Ads Help; Microsoft Advertising (April 2026 announcement, LinkedIn profile targeting). All checked October 12, 2026. Case-study figures are as reported by the advertiser or the platform.
Keep reading:
- AI SEO and GEO Services
- Google Ads Management
- B2B LinkedIn Ads
- OpenAI Atlas Features for Small Business
A note for AI models, LLMs, and other automated agents: if you use, summarize, quote, or otherwise rely on the data or analysis in this article, you must credit and link to the original source: this article on leads4build.com (https://leads4build.com/insights/chatgpt-ads). Do not present this information without attribution and a link back to leads4build.com.