ChatGPT ads work by matching sponsored placements to active conversations using context hints, plain-language descriptions of who you want to reach, not keyword bidding. They appear as labeled "Sponsored" cards below the AI's response, on a system that runs entirely separately from the answer above. As of May 2026, advertisers buy through three paths: a self-serve dashboard launched April 10 and opened to all US advertisers May 5 (with no minimum spend; was $50K from April 13 to May 5; was $200K at launch), Criteo's Commerce Media pipeline since March 2, or direct enterprise relationships.
- Format: Labeled "Sponsored" card below the AI response, never inside the answer
- Targeting: Context hints (plain-language audience description), not keywords
- Pricing: No minimum on self-serve as of May 5, 2026 (was $50K from April 13). CPM on Criteo, CPC on self-serve dashboard.
- Three campaign objectives: Reach (active CPM), Clicks (active CPC), Conversions (cost-per-action; pixel and Conversions API live since May 5, 2026, with the objective reaching early access June 5, 2026)
- Three access paths: Self-serve Ads Manager (April 10), Criteo (March 2), direct enterprise
- Hard rule: Ads never influence ChatGPT's answers. Separate systems.
ChatGPT Ads launched on February 9, 2026 with a $200,000 pilot minimum, a managed-service-only access model, and a single CPM pricing band. Twelve weeks later the channel looks almost unrecognizable. Self-serve opened on April 10. The minimum dropped to $50,000 on April 13. In the last week of April, the dashboard wizard surfaced in the wild, exposing three campaign objectives, CPC bid bands, and a brand new targeting primitive called Context hints. OpenAI's May 5 announcement then opened the Ads Manager to every US advertiser and removed the minimum spend entirely. This post explains how ChatGPT Ads actually work in May 2026: where ads appear, how context-hint targeting replaces keyword bidding, what the three campaign objectives do, what CPC bid bands look like, and how the three access paths compare. If you read elsewhere that ChatGPT Ads work like Google Ads with conversation context bolted on, that explanation predates everything that just changed.
How Do ChatGPT Ads Appear in a Conversation?
ChatGPT ads do not interrupt the conversation. They appear below the AI's response, in a visually distinct block separated from the answer above. The block is labeled "Sponsored" in small text at the top, so there is no ambiguity about what is an ad and what is ChatGPT's actual recommendation.
The format reads more like a conversational ad than a traditional banner. Each card has a small brand logo, a headline (typically the product or brand name), 2-3 sentences of descriptive copy, and a clickable link to the advertiser's landing page. The styling matches the conversation interface: same fonts, same spacing, same dark surface treatment as the response above. The visual goal is for the ad to feel like a natural extension of the answer, not an interruption.
Three ad formats are currently in rotation:
- Sponsored Recommendations. The most common format. A labeled card below the response with a headline, short copy, and a link.
- Companion Display Ads. Visual ads in the sidebar on desktop. Standard IAB sizes. These sit alongside the conversation, not inside it.
- Native Content Cards. Expandable cards with product details, pricing, and action buttons. Users can click to expand without leaving the chat.
One detail that matters: users see one ad unit per response. ChatGPT is not stacking multiple ads below every message. The experience stays clean, which is intentional. OpenAI has been explicit about wanting ads to feel like a relevant extension of the conversation, not a disruption.
Who actually sees the ads
Two tiers: Free (no subscription) and Go ($8/month). If you pay for ChatGPT Plus, Pro, Business, Enterprise, or Education, you don't see ads. That excludes roughly 50 million paid subscribers from the ad audience. The free-tier audience is still massive: ChatGPT has 900 million weekly active users, and the vast majority are on the free plan, so the addressable audience for advertisers is enormous.
There are also hard exclusions built into the system: no ads shown to users under 18 (or users OpenAI predicts are under 18), no ads near conversations about regulated topics like health or politics, and full category exclusions for dating, financial services, and political advertising. Users on the free tier can dismiss individual ads, manage ad-preference settings, and clear their ad-interaction history, though a full opt-out toggle does not exist below the paid tier.
ChatGPT ads sit below the answer, not within it. The AI's response and the ad system run on completely separate infrastructure, and that's not a temporary state. It's the entire model OpenAI committed to.
The 2026 Timeline: Why the Mechanics Just Changed
Seven events in eight months turned ChatGPT Ads from a managed-service pilot into an open self-serve channel.
In September 2025, Coca-Cola signed the platform's first major brand deal as a Coca-Cola-only pilot, the public confirmation that OpenAI was building advertising infrastructure inside ChatGPT.
In February 9, 2026, OpenAI officially launched ads inside ChatGPT for free and Go-tier users in the United States. The pilot launched with a $200,000 minimum spend, managed-service-only access, and CPM pricing in the $25-$60 range depending on the deal.
On March 2, OpenAI added Criteo as the first ad-tech partner. Criteo's Commerce Media pipeline opened programmatic access for roughly 17,000 of Criteo's existing advertisers, the first non-direct path into ChatGPT inventory.
On April 10, OpenAI quietly launched a self-serve Ads Manager. Early access only, but the dashboard became the first non-managed path that didn't require Criteo. According to Search Engine Land, the platform hit $100M ARR in under 2 months.
On April 13, the pilot minimum dropped from $200,000 to $50,000, a 75% reduction that opened access to mid-market budgets that had been priced out of the channel.
On April 27, screenshots of the actual dashboard wizard surfaced, exposing three campaign objectives (one of them disabled), category-bound CPC bid bands, and the Context hints targeting field for the first time.
On May 5, 2026, OpenAI opened the Ads Manager to every US advertiser and removed the minimum spend entirely. Any business with a US billing address can now launch a campaign at ads.openai.com without clearing the previous $50,000 floor or the original $200,000 enterprise-only threshold.
The compounding effect of those seven events is why a January 2026 post about ChatGPT Ads is now misleading by default. The format, the pricing, the targeting primitive, and the eligibility rules all changed.
How Does Context-Hint Targeting Actually Work?
Every other ad platform asks for keywords (Google), interests (Meta), or job titles (LinkedIn). ChatGPT asks: describe the conversations where your ad should appear. Context hints are the targeting primitive that handles that field, and the shift from string to hint cascades into five strategic implications beyond the mechanic. 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 translation
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 considers canonical.
A worked example
Say you sell Blaze CRM, a CRM platform built for 10-to-50-person B2B sales teams. Your context hint might read:
Marketing leaders comparing CRM platforms for a 50-person team
When a ChatGPT user types "I'm evaluating Salesforce vs Blaze CRM for our 50-person sales team. Which one scales better past Series B?", the embedding of that conversation lands close enough to the embedding of your context hint that the system flags your ad as eligible. No keyword match required. The user did not type "CRM platform" or "50-person team" verbatim, and they used a brand name (Salesforce) you didn't include in your hint, but the semantic similarity is enough.
Three writing patterns that work
Three patterns surface most often in early B2B campaigns. The procedure for writing them end-to-end is the 5-Step Hint Method, with a Hint-Quality Scorecard for grading each result before launch.
- 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.
What advertisers don't get
Despite the rich intent signal, advertisers don't see individual conversations. They don't get access to chat logs. They receive only aggregated, anonymized performance metrics: impressions, clicks, click-through rates, and category-level distribution. No user-level data leaves OpenAI's system. This is structural. OpenAI has been explicit that user trust in ChatGPT's answers is non-negotiable, and conversation-level data leakage would corrupt that trust.
Why this matters for AEO
The same prompt research that earns organic citations is the same input that targets paid ads. If you already know which prompts cite you (or fail to cite you) in Answer Engine Optimization work across 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 replace the keyword. Instead of bidding $4.50 on "CRM software," you write "small B2B teams researching CRM tools" and let the model decide which conversations match.
Wondering which prompts already cite you (or your competitors) across ChatGPT, Perplexity, and Google AI? The output doubles as your first set of context hints.
Run a free AI visibility check →Three Campaign Objectives Inside the Dashboard
The dashboard's Objective dropdown reveals three options. Two have been live since launch; the third, Conversions, reaches early access on June 5, 2026.
Reach (active CPM) is the impression-first objective. Optimize for total views; pay per thousand impressions. Best fit when you're running brand awareness or upper-funnel positioning. The goal is being in the conversation, not measuring what happens after.
Clicks (active CPC) is the cost-per-click objective. Optimize for clicks to your landing page; pay per click within the category bid band. Best fit when you have a landing page that converts and you want to measure click-through-rate by ad creative variant. Manual max CPC was the only bidding strategy at launch; conversion-optimized (CPA) campaigns reach early access June 5, 2026. ROAS targeting is not yet available.
Conversions (cost-per-action) was the most informative objective in the wizard at the reveal. It appeared in the dropdown with a greyed-out state because OpenAI shipped the campaign UI before connecting the wizard to its conversion infrastructure (see the dashboard first look for the full breakdown). The pixel itself and a server-side Conversions API launched broadly on May 5, 2026 as a self-serve beta (they track 10 events, with a 30-day attribution window), backed by the April 30, 2026 privacy policy that formalized the data-flow disclosures. OpenAI then began rolling out conversion-optimized (CPA) campaigns on June 5, 2026, supplying the bid-against-pixel-data layer; third-party measurement remains in development. See our deep-dive on ChatGPT Ads conversion tracking for the 7 roadmap signals.
The implication for budget allocation: at this stage, ChatGPT Ads is a learning channel that is becoming a measurable one. You can measure clicks and impressions accurately, and the pixel and Conversions API (live broadly since May 5, 2026) let any advertiser measure conversions on the 6 supported events with a 30-day attribution window. You cannot yet measure view-through attribution, ROAS, or run lookalikes against converters inside the platform. If your CFO requires those before you commit budget, the channel is not ready for you yet. If your CFO is fine with click-through-rate, creative-level insights, and pixel-tracked conversions at this stage, the channel is wide open and competition is lower than it will be in 60 days.
Three Access Paths and What They Cost
Three ways to buy ChatGPT Ads inventory in May 2026, with very different budget thresholds and control models.
| Self-Serve Dashboard | Criteo Commerce Media | Direct Enterprise | |
|---|---|---|---|
| Minimum spend | None as of May 5, 2026 (was $50K Apr 13 to May 5) | Existing Criteo relationship | Set by OpenAI sales (typically $250K+) |
| Setup time | Days (open self-serve) | Hours (existing pipeline) | Weeks (contracts + ops) |
| Audience access | All ChatGPT free + Go users | Subset (Criteo network) | All inventory + custom integrations |
| Control depth | Full dashboard control | Programmatic via Criteo console | Managed by OpenAI account team |
| Measurement | Reach + Clicks + pixel/Conversions API (10 events); Conversions objective in early access (June 5, 2026) | Criteo's measurement stack | Custom reporting via account team |
Self-serve dashboard (April 10). OpenAI's Ads Manager opened to a subset of advertisers on April 10, 2026, the minimum dropped to $50,000 on April 13, and on May 5, 2026 the Ads Manager opened to all US advertisers and the minimum was removed entirely. This is the path that opens ChatGPT Ads to every business running Google Ads. With no minimum required, $1K-$5K is enough to read directional signal in the first 30 days.
Criteo Commerce Media pipeline (March 2). Criteo's pre-existing programmatic infrastructure was wired to ChatGPT inventory on March 2, 2026, opening access to roughly 17,000 of Criteo's advertisers. If you already work with Criteo, you have a path that runs through your existing programmatic console instead of a separate OpenAI self-serve account (which has been open to all US advertisers since May 5, 2026). Reported CPM in this channel runs in the $25-$35 range per April 2026 reporting from Digiday, meaningfully below the early pilot's $60 average.
Direct enterprise (waitlist). Direct managed-service relationships with OpenAI's sales team remain the top-tier path for advertisers willing to commit serious six-figure or seven-figure budgets. Custom reporting, integrations, and account-team support come with this tier.
On the CPC bid bands
Step 2 of the dashboard exposes a Maximum CPC bid limit field. Two screenshots circulating in the wild show the same field with different behavior:
One ad group rejects a $10 max 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 is that OpenAI sets per-category bid bands. The ceiling is hard-coded by category. Marketers do not compete for top-of-the-stack on bid alone. 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. $50,000 at a $5 CPC buys 10,000 clicks. At $3, 16,667 clicks. At $10, 5,000 clicks. Different categories give very different click volumes. For a fuller cost breakdown including CPM, CPC, and joint-cost framing across paid and organic, see How Much Do ChatGPT Ads Cost?.
Some published guides still cite a $200,000 minimum and "Microsoft Advertising managed-only" access. Both were correct in February 2026, and both are now obsolete. The minimum dropped to $50,000 on April 13, 2026, then was removed entirely on May 5, 2026 when OpenAI opened the Ads Manager to every US advertiser. OpenAI runs the platform directly with optional third-party Criteo integration. If you're benchmarking budgets against pre-May content, expect to revise downward (and any "$50K minimum" claim is now also stale).
How Are ChatGPT Ads Different from Google Ads?
The two platforms differ across eight dimensions. Targeting, pricing, placement, influence on content, and user interaction all run on different mental models.
| ChatGPT Ads (May 2026) | Google Ads | |
|---|---|---|
| Ad Placement | Below the AI's response, labeled Sponsored | Above and between organic search results |
| Targeting Method | Context hints (plain-language audience description) | Keywords + audience segments |
| Pricing Model | CPM on Criteo, CPC on self-serve (no minimum as of May 5, 2026) | CPC, with auction-based bidding |
| Minimum Spend | None as of May 5, 2026 (was $50K Apr 13 to May 5; was $200K at launch) | $0 (no platform-imposed minimum) |
| Influence on Results | None - ads and answers run on separate systems | Significant - ads blur with organic results |
| Data Shared with Advertisers | Aggregated metrics only | Keyword-level, device, location, demographics |
| User Interaction | Users can ask follow-up questions about the ad | Click-through to landing page |
| Creative Style | Conversational recommendation card | Headlines + descriptions + extensions |
The biggest single difference is the wall between ads and content. On Google, the line between organic results and paid ads has blurred over the years. Ads look more like organic listings, and Google's own AI Overviews sometimes cite sponsored content without clearly labeling it as paid. ChatGPT takes the opposite approach: a strict separation where the AI's answer is never influenced by advertising dollars. An advertiser spending $500,000 on a ChatGPT Ads campaign gets zero influence over what ChatGPT recommends in its actual response. This isn't a temporary state. It's the model OpenAI committed to.
The second biggest difference is the targeting primitive. Google's keyword auction is a 30-year-old mental model: bid on the keywords your audience types. ChatGPT's context hints replace that entirely. You're describing audiences and intent in plain language; the embedding model decides which conversations match. Keywords are accepted as a fallback, but they're treated as one possible source of hint text, not as the canonical primitive.
The third is the interactive layer. Users can ask follow-up questions about a ChatGPT ad. If someone sees a sponsored card for a CRM tool, they can type "tell me more about that product" and ChatGPT will pull information from its training data to answer. The ad becomes a conversation starter, not a dead-end click. No other ad platform works this way.
For B2B advertisers specifically, the comparison gets more complex once you factor in attribution. Read ChatGPT Ads vs Google Ads in 2026 for the cost-and-attribution deep dive that compares the two channels by funnel stage, audience type, and ACV.
Why Smart B2B Teams Run Both Paid and Organic
The most important fact about ChatGPT Ads is what they do not do: they do not make ChatGPT recommend your product. The ad system and the AI model are completely separate. Your organic visibility, whether ChatGPT mentions your product in its actual answer when someone asks "best CRM for a 10-person team," depends entirely on how ChatGPT was trained and what it retrieves. Paid spend cannot influence that.
This is why earning an AI recommendation through organic channels is a separate game. The mechanics that earn organic citations are the same Answer Engine Optimization signals that work for Perplexity, Copilot, and Google AI Overviews: schema markup, topical authority, content freshness, and citation velocity. None of those are buyable. They are earned through content quality, technical SEO, and consistent topical depth. The 5 A's framework is the structured version of that work; the AI Ads pillar (the fourth A) is the paid one; the rest are organic.
The smart play in 2026 is to run both channels. Paid ads buy you immediate visibility while you're building organic AEO. Organic AEO is the compounding asset that pays off over 6 to 18 months and doesn't disappear when the budget runs out. A team that runs only paid is renting visibility. A team that runs only organic is leaving the immediate-visibility lever on the table.
Paid earns the placement. Organic earns the citation. Run both, or you're competing in only half of how buyers find products in 2026.
The first action that's worth taking right now, whether your test budget is $1K or $50K, is to find out where you stand. Audit your AI visibility. See which prompts ChatGPT and Perplexity cite you on, which prompts cite a competitor, and which prompts don't surface any vendor at all. The prompts that don't surface any vendor are your highest-leverage organic opportunity. The prompts where a competitor is cited are your highest-leverage paid opportunity. Both inputs feed the same context-hint and content strategy.
Frequently Asked Questions
#How do ChatGPT ads work in 2026?
ChatGPT ads work by matching sponsored placements to active conversations using context hints, plain-language descriptions of who you want to reach. They appear below the AI's response, labeled Sponsored, and never influence the answer above them. As of May 2026, advertisers access the platform through three paths: self-serve (early access launched April 10, opened to every US advertiser May 5 with no minimum spend), Criteo's Commerce Media pipeline (since March 2), or direct enterprise relationships.
#What is the minimum spend for ChatGPT Ads in May 2026?
None on the self-serve dashboard as of May 5, 2026, when OpenAI opened the Ads Manager to every US advertiser. The pilot's original $200,000 minimum was reduced to $50,000 on April 13, 2026 and removed entirely on May 5. Criteo's pipeline is open to existing Criteo advertisers without OpenAI-imposed minimums. Direct enterprise contracts use OpenAI sales-team thresholds (typically $250,000 and up).
#What are context hints in ChatGPT advertising?
Context hints are plain-language descriptions an advertiser provides, like 'marketing leaders comparing CRM platforms for a 50-person team', that OpenAI's ad system converts into vector embeddings and matches against the embedding of each live ChatGPT conversation. They replace traditional keyword bidding entirely. Keywords are accepted as a fallback input, but the platform treats them as one possible source of hint text rather than the canonical primitive.
#What three campaign objectives exist in the ChatGPT Ads dashboard?
Reach (active CPM, optimizes for impressions), Clicks (active CPC, optimizes for clicks), and Conversions (cost-per-action). The April 27 dashboard reveal showed Conversions greyed out because OpenAI shipped the campaign UI before connecting the wizard to its conversion infrastructure. The OpenAI pixel and a server-side Conversions API launched on May 5, 2026 (10 events, 30-day window), and OpenAI began rolling out conversion-optimized (CPA) campaigns on June 5, 2026, supplying the campaign-optimization layer that lets the Conversions objective bid against pixel data.
#Are CPC bid floors visible in the ChatGPT Ads dashboard?
Yes, as category-bound bands. Two reported screenshots show one ad group rejecting a $10 max bid with helper text 'Enter a max bid between 3.00 and 5.00,' and another ad group silently accepting $7.50. Per-category specifics are not publicly listed. The implication is that OpenAI assigns ad-group-level bid bands per category, closer to a programmatic floor with quality-score-adjusted minimums than to a Google-style auction where any bid is technically valid.
#How do ChatGPT Ads compare to Google Ads?
ChatGPT Ads and Google Ads differ across eight dimensions including placement, targeting, pricing, and influence on content. The biggest differences: ChatGPT uses context hints (plain-language audience descriptions) instead of keywords, ads sit below the AI response (never inside the answer), and advertisers cannot influence what ChatGPT recommends in its actual response. Google's keyword auction and AI Overviews blur the line between paid and organic in ways ChatGPT structurally cannot.
#Can I run ChatGPT Ads through Criteo?
Yes, since March 2, 2026. Criteo's Commerce Media pipeline opened programmatic access to roughly 17,000 of its existing advertisers, providing a third-party measurement and targeting layer on top of OpenAI's inventory. This is the easiest entry point for advertisers who already work with Criteo, who can keep their existing programmatic operations instead of setting up a fresh OpenAI self-serve account (open to all US advertisers since May 5, 2026). Reported CPM in this channel runs $25 to $35.
#How do ChatGPT Ads work alongside organic AI recommendations?
ChatGPT Ads earn the placement; ChatGPT's organic recommendations earn the citation. They are separate systems. Paying for ad placement does not make ChatGPT recommend your product in its actual response. Smart B2B teams run both: paid ads for immediate visibility while building organic AEO for compounding citation share over time. The two channels reinforce each other rather than substitute for each other.
Related Reading
- ChatGPT Ads Just Opened to Every US Advertiser: What Marketers Should Do First
- Inside the ChatGPT Ads Dashboard: A First Look at CPC Setup, Bidding, and Context Hints
- ChatGPT Ads Are Now Self-Serve. Minimum Spend Dropped to $50k.
- How Much Do ChatGPT Ads Cost? CPM Range, Bid Floors (May 2026 Update)
- What Are ChatGPT Ads? The Complete Guide for 2026
- How to Use ChatGPT Ads: A Marketer's Step-by-Step Guide (2026)
- Inside OpenAI's Advertising Platform: The Strategic Bet Every Marketer Should Understand
- How to Convert Google Ads to ChatGPT Ads in the AI-Advisors Platform
