ChatGPT ads place a sponsored unit below an answer when the conversation looks relevant to your offer. For a SaaS founder the question is not whether the channel is new, but whether your buyer can see it and whether a small test can tell you anything. This guide checks the reach limits OpenAI documents, explains how ads are matched without keywords, sets a budget ceiling from OpenAI's pacing rules, and compares the test with a first Google search campaign. It is useful without any Soloop account.
What ChatGPT ads are in October 2026
OpenAI started testing ads in the United States on February 9, 2026, and now offers a self-serve Ads Manager in a beta. The platform is still changing: OpenAI says delivery, inventory, ad formats and buying options will evolve, and that it has no performance benchmarks yet across advertisers, industries or campaign types. Treat any cost-per-click figure you read elsewhere as someone else's experience, not a forecast.
Ads may appear for users on the Free and Go plans. Plus, Pro, Business, Enterprise and Edu accounts do not have ads. Accounts identified as under 18 do not see ads, and Temporary Chats show none.
OpenAI Help: Ads in ChatGPTBelow the end of a response, labeled as sponsored and visually separated from the answer. Ads are not eligible near sensitive or regulated topics such as personal health, mental health or politics.
OpenAI Help: Ads in ChatGPTThe system considers the current conversation, the ad's landing page, title and copy, the advertiser's context hints and targeting, and, when personalization is on, other signals from the user's ChatGPT use. Ads do not change ChatGPT's answers.
OpenAI Help: Ads in ChatGPTAggregated reporting such as views and clicks. Advertisers never receive chats, memories, names, emails or precise location.
OpenAI Help: Ads in ChatGPTChatGPT Ads is in beta after a pilot from February through April. Ads Manager Beta supports campaign creation, billing, reporting, conversion measurement and CPC bidding. There are no cross-advertiser performance benchmarks yet.
OpenAI Help: ChatGPT Ads FAQDecide whether your buyer can see the ad
The reach limit matters more for SaaS than for most advertisers. If your buyer uses ChatGPT through a company Business or Enterprise workspace, or pays for Plus, that buyer does not see ads at all. A tool sold to a founder on the free plan is in reach; a tool sold to an enterprise IT team that works inside ChatGPT Enterprise may not be.
Answer these questions before you open Ads Manager. A "no" is not a failure; it tells you which channel to test first.
- Plan: do the people you sell to use ChatGPT on the Free or Go plan for the problem your product solves? If you do not know, ask five of them which plan they use.
- Conversation: would they describe the problem to ChatGPT in their own words, such as "how do I schedule three crews across forty weekly jobs"? Ads follow conversations, so a problem people never ask about gives the system nothing to match.
- Country: is the legal entity that will pay for the ads based in a country where self-serve Ads Manager is available? New self-serve accounts may at first advertise only in their home country.
- Category: is your product outside the excluded and restricted categories in OpenAI's ad policies?
- Landing page: does one page explain the offer to this buyer, and does your site allow OpenAI's crawlers OAI-AdsBot and OAI-SearchBot? OpenAI requires landing pages not to block them.
- Counted action: can you count one useful action from a visit, such as a trial signup or a first completed task?
- Affordable loss: can you lose the whole test budget, at least the minimum daily budget for the full test, without touching money the product needs?
Write context hints instead of keywords
Search ads start from words a person types. ChatGPT ads start from what a conversation is about. OpenAI calls the advertiser's input context hints: short descriptions of what you offer, who it helps and when it is useful. They inform relevance but are not exact-match keywords or targeting rules, and they do not guarantee delivery in any particular conversation.
Write each hint as a situation, not a category. "Scheduling software" says little. "Crew scheduling for small landscaping companies running weekly lawn routes" names the buyer, the job and the moment. Use separate ad groups when two needs require different messages or landing pages.
OpenAI's ad guidance asks for several distinct title and copy variations per offer, each with a different angle, and for copy that explains practical value rather than a slogan. Link each ad to the most relevant page, not the homepage, and add UTM parameters to the destination URL so your own analytics can separate this traffic.
A campaign sets the objective, budget, dates, countries and optional platforms and audiences. Ad groups organize ads around one product, service or need and hold context hints. Ads hold the title, copy, image and landing page.
OpenAI Help: create campaignsClear, specific phrases that describe what you offer, who it helps or when it may be useful. They are not exact-match keywords or targeting rules.
OpenAI Help: create ad groupsUse several distinct, benefit-focused variations; link to the most relevant page; UTM tags are allowed; landing pages must not block OAI-AdsBot or OAI-SearchBot.
OpenAI Help: create adsSet the ceiling from OpenAI's budget rules
Decide the total you are willing to lose first. Then choose the budget type that enforces it. A campaign-total budget caps spend over the campaign's full duration, and OpenAI calls it the stricter control. A daily budget is an average over a Sunday-to-Saturday week: one day can spend up to twice the amount, and one week up to seven times it.
The minimum daily budget is 25 US dollars for accounts billed in dollars (15 euros, 2,500 yen; other currencies are listed in OpenAI's table). At the minimum, a daily budget can spend $50 on one day and $175 in one week. A daily budget does not end a test; the end date or a campaign-total budget does.
Three details catch first-time advertisers. On your first campaign, the start and end dates may only be editable after creation, so do not publish just to reach those settings. Ads may keep serving for up to 24 hours after you pause, and that spend is billable. Spend reporting may lag 7 to 8 hours, so a zero total does not mean nothing is accruing.
- Write the ceiling: the total amount you can lose.
- Choose the test length and divide the ceiling by the number of days. If the result is below the minimum daily budget in your currency, shorten the test or raise the ceiling.
- Use a campaign-total budget equal to the ceiling, and set the end date before the campaign goes live.
- Choose the Clicks objective for a first test, so you pay per click and can read traffic directly. Conversion-optimized campaigns have their own prerequisites.
- Name the one action you will count and confirm that a test visit is recorded.
- Write the stop rules, with dates, before launch.
Views is billed per 1,000 impressions (CPM), Clicks per valid click (CPC). Conversions optimizes toward a selected event and bills per click (oCPC) or per 1,000 impressions (oCPM), not per conversion.
OpenAI Help: billing and paymentA daily budget may spend up to twice the amount on one day and no more than seven times it in a Sunday-to-Saturday week. A campaign-total budget limits spend over the full campaign. The minimum daily budget is USD 25, EUR 15, GBP 15 or JPY 2,500, among other currencies.
OpenAI Help: create campaignsSelf-serve accounts are billed after delivery: the saved card is charged when unpaid spend reaches an assigned payment threshold, and any balance at month end. The threshold is not a spending cap. Ads can serve for up to 24 hours after a pause, and that cost is billable.
OpenAI Help: billing and paymentWorked example: a fictional first test
Everything in this example is fictional, including the product, the wording, the budget and the dates. The figures are planning assumptions. They are not benchmarks, customer results or output from a Soloop task.
CrewSlate is an invented scheduling tool for landscaping companies with two to ten crews. Eight owners interviewed by the founder said they use the free ChatGPT plan, and three of them described asking it how to plan weekly routes. The founder can count a trial signup and, separately, the first week a new account schedules all its crews in CrewSlate.
| Dimension | Brief | Reason, or what would change it |
|---|---|---|
| Buyer and conversation | Owners of small landscaping companies asking ChatGPT how to plan crews and weekly routes. | Interview notes show the plan they use and how they phrase the question. If most owners pay for Plus, they cannot see the ad and the test stops here. |
| Context hints | "Crew scheduling for landscaping companies with two to ten crews"; "planning weekly lawn-care routes across crews"; "replacing a whiteboard or spreadsheet crew schedule". | Each hint names the buyer and the moment. Hints are not keywords, so the founder reads which conversations the ads reached, not which words matched. |
| Ads | Three variations: one on saving the Monday planning hour, one on moving jobs when it rains, one on replacing the whiteboard. | Each variation tries a different angle, as OpenAI's guidance suggests. All three link to the same landscaping page with a UTM tag. |
| Landing page | One page for landscaping owners, with the trial as the only next step. Robots rules allow OAI-AdsBot and OAI-SearchBot. | The homepage also speaks to cleaning and pest-control companies, so a click there could not be interpreted. |
| Objective and budget | Clicks (CPC). Campaign-total budget $300 from October 19 to October 30, 2026: 12 days at an average of $25, the US minimum daily budget. | $300 / 12 = $25. A daily budget of $25 could instead spend $50 on one day and $175 in one week, so the total cap is the clearer ceiling. |
| Counted actions | Trial signup, and separately the first week with all crews scheduled. | A signup shows interest. A full week of schedules shows use. Payment is a later, separate fact. |
| Mid-test check | October 24, at least eight hours after the last change: compare spend with the ceiling, read impressions and clicks by ad, and check that signups are recorded. | Spend reporting can lag 7 to 8 hours. If impressions are near zero, the hints or the page may not match real conversations. |
| Stop rules | Stop at $300 or on October 30, whichever comes first. Pause early if the page breaks or the signup event stops recording. | Pausing can still allow up to 24 hours of billable delivery, so pause before the money runs out, not after. |
| Decision after | At least one account that schedules a full week: plan a second test with new hints. Clicks but no signups: fix the page first. Few impressions: test Google search instead. | Twelve days at this size cannot establish a cost per customer. It shows whether the path from conversation to product works at all. |
ChatGPT ads or a first Google search test?
For most early SaaS products the two channels answer different questions. A Google search test shows whether people already type your phrase and click. A ChatGPT test shows whether people describe your problem in a conversation and act on a sponsored suggestion. If your budget allows only one test, choose the channel where you have better evidence that your buyer is present.
| Dimension | ChatGPT ads | Google search ads |
|---|---|---|
| Matching | Conversation context, your ad and page, and context hints. No keywords. | Keywords with exact, phrase or broad match, plus negative keywords. |
| Who can see it | Free and Go users aged 18 or over, outside Temporary Chats and sensitive topics. | People searching on Google where your campaign is eligible. |
| What you can read | Impressions, clicks, spend, CTR, average CPC and CPM, and conversions if measurement is set up. | The same basics, plus the search terms report: the actual searches that triggered your ads. |
| Budget rules | Minimum daily budget (USD 25); up to 2x per day and 7x per week; or a campaign-total cap. | Average daily budget; up to 2x per day and at most 30.4x per month. |
| Maturity | Beta; OpenAI publishes no performance benchmarks yet. | Mature, with planning tools such as Keyword Planner. |
| With Soloop | Not supported. Run it yourself in Ads Manager. | Soloop can prepare and launch a Search campaign after you complete setup and payment. |
Copy the first-test brief
Select and copy this brief into your own document; no account is required. Fill the plan column first, then write what would make you change it.
| Dimension | Plan for this test | Evidence or reversal condition |
|---|---|---|
| Buyer / ChatGPT plan | [Role; the plan they use] | [How you know: interviews, surveys, support messages] |
| Conversation | [The question they would ask ChatGPT, in their words] | [Where the wording came from] |
| Context hints | [Three to five hints: what, for whom, when] | [Needs that deserve a separate ad group] |
| Ads | [Two to four variations with different angles] | [Claim each ad makes; capability you inspected] |
| Landing page | [URL with UTM; the one next step; crawler access checked] | [Who else the page addresses] |
| Objective / budget / dates | [Clicks; campaign-total ceiling; start and end date] | [Average per day vs your currency's minimum] |
| Counted action | [Signup or first use; how it is recorded] | [Confirmation that a test visit was counted] |
| Stop rules / decision after | [Ceiling, end date, early-stop conditions; what you do in each outcome] | [What this test cannot tell you] |
Read the results one stage at a time
When the test ends, read each stage separately. A single cost-per-signup number from a small test hides where the path broke.
- Few impressions: few eligible conversations matched your hints and page, or your buyer is not on an ad-supported plan. Rewrite the hints once; if impressions stay low, stop.
- Impressions but few clicks: the ad does not answer what the conversation was about. Compare the angle of each variation.
- Clicks but no counted action: the page was unclear for this visitor or the next step asked too much. Check the UTM-tagged sessions in your own analytics.
- Counted actions: record each one and what the account did next. A handful of signups cannot give a reliable conversion rate or cost per customer.
Where Soloop fits
Soloop does not create or manage ChatGPT ads. If you decide to test them, use OpenAI's Ads Manager and this brief.
The Soloop manual says Soloop can prepare and launch Google Ads campaigns once you complete the campaign setup and the required payment steps; the campaign budget is separate from Soloop credits. Much of this brief carries over to that test: the buyer, the landing page, the counted action, the ceiling and the stop rules. To continue in Soloop, copy them into the conversation: My product is [URL]. Buyer [role], phrases they search [phrases], counted action [event], ceiling [amount], dates [start and end]. Prepare a Google Ads campaign for my review.
Soloop does not promise clicks, signups or revenue from any campaign. The Explore Soloop link opens the public entry, so bring the brief yourself.