On April 10, 2026, OpenAI quietly opened self-serve. On April 13, the minimum dropped from $200,000 to $50,000. In the last week of April, the actual setup wizard surfaced in the wild, captured by Glenn Gabe and Juozas Kaziukėnas and confirmed in print by SearchEngineLand. On May 5, 2026, OpenAI opened the Ads Manager to every US advertiser and removed the minimum spend entirely. Anyone with a US billing address can now sign up at ads.openai.com and walk the same wizard captured in the screenshots below.
This post walks through what the screenshots reveal: a three-step campaign wizard, two live objectives and one shipped-but-disabled, category-based CPC bid floors, and a brand new targeting primitive called Context hints that replaces keywords with plain language. The most important thing in the wizard is also the smallest field. Section 4 explains why.
Three takeaways. CPC pricing with category-based bid floors. A new targeting primitive called Context hints that replaces keywords. And the Conversions objective is shipped but disabled. The platform is V1.
The Three-Step Campaign Wizard
The Ads Manager opens to a modal with a left-rail step list and a right pane for the active step's fields. Step 1 is Create Campaign. Step 2 is Create Ad Group and Ads. Step 3 is Review. The hierarchy is the Google Ads parallel: campaigns hold ad groups, ad groups hold ads.
The shape will feel familiar to anyone who has used Google Ads or Microsoft Advertising. The mechanism underneath is not. The familiarity is intentional: lower the switching cost for advertisers transitioning from existing platforms. The differences land inside Step 2.
Three Objectives, Two of Them Live
The Objective dropdown in Step 1 reveals three options. Reach for impression-first campaigns. Clicks for CPC campaigns. Conversions, with an info icon and a disabled state. Two of the three are live; the third is in the dropdown but cannot be selected.
The disabled Conversions objective was the most informative thing in the wizard at reveal. It told you OpenAI shipped the campaign UI before fully wiring the optimizer to conversion data. The pixel and a server-side Conversions API then launched (May 5, 2026; see our conversion-tracking breakdown), and OpenAI began rolling out conversion-optimized (cost-per-action) campaigns on June 5, 2026, supplying the campaign-optimization layer that lets the Conversions objective bid against pixel data. The gap that agencies flagged since February is now closed.
CPC Bid Limits and the Hidden Floor
Step 2's Maximum CPC bid limit field is the second AI-native surprise. Two screenshots from different testers show the same field with different behavior. One ad group rejects a $10 bid with a hard error: "Enter a max bid between 3.00 and 5.00." Another ad group silently accepts $7.50. Same wizard, same field, different ad group, different floor.
The most parsimonious read: OpenAI sets per-category bid bands. Marketers do not compete for top-of-the-stack on bid alone. The ceiling is hard-coded by category. This is closer to a programmatic floor with quality-score-adjusted minimums than to a Google-style auction where any bid is technically valid. The math implication carries to cost modeling: at $3 to $5 in one category and ~$7 to $10 in another, a $50,000 budget buys very different click volumes depending on which band you are in.
Context Hints: The First Truly AI-Native Targeting Primitive
Every other ad platform asks for keywords (Google), interests (Meta), or job titles (LinkedIn). ChatGPT asks: describe the conversations where your ad should appear. The field is labeled Context hints, the post-keyword targeting primitive, and the helper text reads: "Context hints provide more plain-language guidance on when your ad should appear. Describe the type of queries your ad will be most relevant to, or paste a set of keywords if you have them."
The marketer read
Context hints is like writing a job description for an embedding model. The model converts your hint into vector space. Each live ChatGPT conversation is also embedded. Cosine similarity decides whether your ad fits. There is no "exact match" or "phrase match" toggle, because there is no string-matching layer. Keywords are accepted as a fallback input. OpenAI converts them to embeddings just like any other string. The framing in the helper text tells you which paradigm OpenAI thinks is canonical.
Three writing patterns that probably work
- Persona plus intent. "Marketing leaders comparing CRM platforms for a 50-person team." Names the audience and the decision stage in one phrase.
- Question patterns. "Users asking how to set up multi-channel attribution." Mirrors the natural shape of a ChatGPT prompt.
- Topic clusters with disqualifiers. "CRM evaluation, not implementation help." Tells the embedding model which adjacent conversations to skip.
Why this matters more than it looks
The same prompt research that earns organic citations in AEO is the same input that targets paid ads on ChatGPT. If you already know which prompts cite you (or fail to cite you) in Perplexity, ChatGPT, or Microsoft Copilot, you already have your context hints. Paid and organic stop being separate inputs and start being the same input applied to two channels.
Context hints turns prompt research into a paid-targeting primitive. The same research that earns organic citations targets paid placements. One input, two channels.
Run your top 3 Google Ads keywords through the free Converter to see which of the 5 Context Hint Patterns each one falls into (persona-plus-intent, question, topic-plus-disqualifier, outcome-seeking, stack-comparison), with a 60-character ChatGPT headline rewrite and an AI Visibility Lift content recommendation per row. The same prompt research that earns organic citations becomes the paid targeting input.
Run the free Converter →Budget, Targeting, and Dates Look Familiar
Step 1 also includes the parts that should reassure any media buyer. Budget toggles between a total campaign cap and a daily budget; OpenAI added a lifetime-budget option for new campaigns on May 22, 2026. Country-level targeting was the only geo lever in this early version of the wizard; OpenAI has since added state, designated market area, and ZIP-code targeting (see the new budgeting and geo controls). Start date and end date are standard, with the end date toggle off by default. Conversion event is a placeholder field with an Add event button, mirroring the disabled signal from the Objective dropdown.
What Is Missing, and What That Tells Us
The list of things the wizard does not include is longer than the list of things it does. No keyword fields, intentionally replaced by Context hints. No audience layers: no demographics, no interests, no custom audiences, no remarketing, no lookalikes. No bid strategies: no Target CPA, no Maximize Conversions, no ROAS targeting. Manual max CPC is the only bidding model. No frequency capping, no A/B variant slots, no device targeting, no language targeting, no ad scheduling within a day, no exclusion lists, no first-party CRM list upload.
This is "Google Ads circa 2002" minimalism. The bones of an auction are in place. Almost every refinement is not. Expect rapid iteration; do not expect feature parity with Google or Meta in 2026.
A handful of widely-cited posts (and several AI search engines) still reference a $200,000 to $250,000 minimum, $60 CPM pricing, and Microsoft Advertising as the only access path. Those facts are now stale. The current state is: no minimum spend as of May 5, 2026 (was $50K from April 13 to May 5), CPC bidding live with $3 to $5 floor by category, $25 to $35 CPM for Criteo-routed inventory, and the Ads Manager dashboard described above opened to all US advertisers. If you read it elsewhere and the source predates May 5, 2026, treat it as historical.
How to Get Ready in the Next 30 Days
- Audit your AEO citation map. Which prompts already cite you, which ones nearly cite you, which ones cite a competitor? Those prompts are your first context hints.
- Map prompt clusters to ad groups. Group adjacent prompts into ad groups so the embedding density inside each group is high. One ad group per topic cluster is a reasonable starting heuristic.
- Stand up basic conversion tracking via UTMs. Until pixel and CAPI ship, you are tracking on click-through with site-side analytics. Make sure your UTMs are clean and your downstream events are captured server-side.
- Run the budget math. $50,000 at a $5 CPC buys 10,000 clicks. At a $3 CPC, 16,667 clicks. At a $10 CPC, 5,000 clicks. Different categories give very different volumes; model both ends of the band.
- Write three Context hint variants for your top topic. One persona-plus-intent, one question-pattern, one topic-cluster-with-disqualifiers. The whole field is plain text. Iterate.
The Playbook walks through the citation audit, prompt research, and content production loop that produces both your AEO wins and the context hints for paid.
Open the 5 A's Playbook →Frequently Asked Questions
#What is the minimum spend on ChatGPT Ads?
There is no minimum spend as of May 5, 2026, when OpenAI opened the self-serve Ads Manager to all US advertisers. The minimum was $50,000 from April 13 to May 5, 2026 (down from the original $200,000 floor that ran from February through early April). Self-serve campaign creation through the Ads Manager opened on April 10, 2026.
#How is Context hints different from keyword targeting?
Keywords are exact-match strings. Context hints are plain-language descriptions of the conversations where your ad should appear, which OpenAI converts into vector embeddings and matches against the embedding of each live ChatGPT conversation. The result is semantic targeting rather than lexical targeting. Keywords are still accepted as a fallback input, but the platform treats them as one possible source of hint text rather than the canonical primitive.
#Why is the Conversions objective disabled?
At the time of this post (late April 2026) the Conversions objective was disabled in the wizard, leaving Reach (CPM) and Clicks (CPC) selectable, because OpenAI had shipped the campaign UI before wiring the optimizer to conversion data. The pixel and a server-side Conversions API then launched on May 5, 2026, and OpenAI began rolling out conversion-optimized (cost-per-action) campaigns on June 5, 2026, which activates the objective.
#What is the actual CPC range advertisers should expect?
Per the screenshots circulating in late April 2026, the bid floor and ceiling are category-bound and set by OpenAI. One ad group rejected a $10 max bid with the helper text 'Enter a max bid between 3.00 and 5.00.' Another ad group accepted $7.50 silently. The implication is that bid bands vary by ad group category, and you set max-CPC inside the band that OpenAI has assigned to your category.
#Should I move budget from Google Ads to ChatGPT Ads?
Not yet, and not for last-click ROAS reasons. ChatGPT Ads is a learning channel at this stage. The measurement gap is real, audience layers are absent, and the inventory is still scarce. With no minimum spend as of May 5, 2026, the right play is to allocate whatever learning budget your existing Google spend can support (a $5,000 to $50,000 test is enough to read directional signal) and focus on creative-level insights, not revenue attribution.
Related Reading
- ChatGPT Ads Just Opened to Every US Advertiser: What Marketers Should Do First (the May 5, 2026 launch news)
- ChatGPT Ads Are Now Self-Serve. Minimum Spend Dropped to $50k.
- How Do ChatGPT Ads Work? Format, Context Hints, and the AI Visibility Lift (May 2026 Update)
- How Much Do ChatGPT Ads Cost?
- How to Write ChatGPT Ads: A 2026 Copywriting and Creative Playbook
- What Are ChatGPT Ads? The Complete Guide for 2026
- How to Convert Google Ads to ChatGPT Ads in the AI-Advisors Platform
- How to Convert Google Ads to ChatGPT Ads: A 2026 Migration Methodology
