Source: ai-research/openai-ads-launch-2026-05-06.md — verbatim capture of ads.openai.com landing page + three OpenAI Help Center articles (Ads in ChatGPT: The Basics / Ads Manager Beta Overview / Quickstart). Help-center articles updated 8-22 hours before fetch — product is freshly live as of early May 2026. Refreshed 2026-09-29 with three podcast relays: raw/Last_Week_in_AI_256_-_Fable_5.1_Astra_Tease_Gemini_3.8_Flash.md (Andrey Kurenkov), raw/Ep._237_-_GPT-6_Astra_NYC_Bans_AI_in_Schools_Trump_Goes_All-In_on_Data_Centers.md (The Artificial Intelligence Show, Mike Kaput’s news segment) and raw/Adam_Foroughi_Applovin_CEO_-_Surviving_a_92%_Drawdown_Ads_as_ML_1.0_the_50B_Game_Ad_Market.md (All-In interview with AppLovin CEO Adam Foroughi).
Product: OpenAI Ads Manager Beta URL: https://ads.openai.com Status: Beta (per OpenAI’s own framing — “This is a beta product and will continue to evolve over time”) Available to advertisers: Self-serve via ads.openai.com signup API surface: developers.openai.com/ads (Ads API + measurement docs)
OpenAI launched a self-serve advertising platform that places paid ads below relevant ChatGPT conversations. The product targets the same user-decision-moment (that Similarweb measured 600,000 organic citations against) but as a paid surface — Free and Go tier users only, US/CA/AU/NZ only at launch (40 countries by September 2026 — see the 2026-09-29 update), with Plus/Pro/Business tiers explicitly ad-free. CPM and CPC pricing on a relevance-weighted second-price auction. The Marketing API + Ads Manager schema mirror Google Ads structure (Campaign → Ad Group → Ad), so anyone running paid search has a near-zero learning curve.
Key Takeaways
- Free and Go tiers only. ChatGPT Plus, Pro, and any Business plan are explicitly ad-free. This caps the addressable audience but eliminates the “are paying users seeing ads” concern that plagued early AdSense.
- Geos: US, Canada, Australia, New Zealand only. No EU/UK launch yet — likely tied to ad-policy review and DSA/GDPR compliance work. Superseded 2026-09-29: Free and Go users now see ads in 40 countries (Last Week in AI #256); the country list was not given. See the update below.
- Ads sit below conversations, not inside them. Per the trust principles: “Ads remain distinct from ChatGPT’s responses” — they are not blended into the answer the way sponsored snippets are in Google AI Overviews.
- Targeting mechanism is “context hints,” not keywords. Advertisers describe the conversations/topics where their product is relevant; the system does relevance matching on context hints + ad title + copy + landing page. Explicitly not exact-match keywords.
- Second-price auction, relevance-weighted. Same auction shape as Google AdWords — relevance score multiplied against bid; second-price clearing.
- Pricing benchmarks: 60 default max CPM. CPC is “Clicks” objective, CPM is “Reach” objective.
- Standard ad creative: advertiser name + favicon + title + copy + landing page + image asset. Same components as a Meta Ad or Google display ad.
- Bulk upload via CSV — campaigns can be created at scale. There is also an Ads API with API keys for programmatic operation, suggesting this is agent-friendly from launch.
- Measurement: impressions, clicks, spend, CTR, avg CPC, avg CPM, conversions. UTM parameters persist on ad clicks; conversion tracking via Ads Manager.
- Early advertisers: Best Buy, Lowe’s, VistaPrint. Best Buy quote acknowledges the shift — “consumers are increasingly turning to platforms like ChatGPT to research and make decisions” — meaning these are dollars consciously moved from other surfaces.
- Brand-safety policy: OpenAI’s stated commitment is ads only near chats that are safe and brand-appropriate. Specific policies at openai.com/policies/ad-policies/.
Update 2026-09-29 — $1B run rate, 40 countries, wider self-serve
Secondary reporting from podcasts; none of it is from an OpenAI page in the vault yet.
- Scale. Andrey Kurenkov, relaying reports on Last Week in AI #256: OpenAI’s ad business “hits $1 billion in annualized revenue run rate” and is “roughly 200 days old.” Mike Kaput on The Artificial Intelligence Show Ep 237 gives the same figure, “less than 200 days after launch.”
- Geography (viewers). Kurenkov: Free and Go users “are seeing ads now in most countries, 40 countries,” which is “a majority of ChatGPT’s one billion weekly active users.” This replaces the four-country list at launch. No country list was given.
- Geography (advertisers). Kaput: OpenAI “expanded their self-service ads manager access across India, Europe, the Middle East, and North Africa.” Podcast relay of an OpenAI statement; confirm against OpenAI’s announcement before relying on it.
- Open profitability question. Kurenkov is unsure the ads cover the cost of serving free users, since tokens are expensive relative to Google search.
- Where the format competes (opinion). AppLovin CEO Adam Foroughi on All-In: chatbot ads serve bottom-of-funnel intent, so “that ads model is almost going to exclusively compete with the Google Search business.” Discovery advertising (Meta, AppLovin) is what he says creates “economic expansion.” He is also skeptical of mass agentic shopping: “the typical shopper is not the person who’s deep into agents.” He runs a competing ad network.
What to do with it. If you excluded ChatGPT Ads because your market was outside US/CA/AU/NZ, re-check eligibility in Ads Manager now. Foroughi’s framing suggests funding a test from search budget rather than from discovery/social budget, since the two surfaces serve different intent.
Performance Data (Similarweb panel, Mar-Apr 2026)
Similarweb’s Ad Intelligence panel data (Mar 30 – Apr 13, 2026, ~1K active brands on ChatGPT) gives the first measured performance picture of the surface:
- CTR: 0.68% overall, 0.50% median brand, 1.00% top-quartile, 1.57% top-decile, 5.4% peak. Slots between display (0.35%) and podcast (0.5-1%) on raw CTR — well below search (3-5%), but the format is brand-not-direct-response.
- **Implied CPC: 3-5 starting max bid — implied CPC reflects actual auction clearing). 15-25), 3× search ($2-4).
- Ad penetration: 1.5% of conversations on ChatGPT (2.0% weekday peak, 0.8% weekend low). Clear Mon-Thu peak / weekend dip / Monday recovery cycle — consistent with B2B-skewed top advertisers.
- Engagement is 6× deeper than Google AI Mode. ChatGPT median 6 turns / avg 17 turns vs Google AI Mode 1 / 2.6. After an ad fires, ChatGPT users stay 4 more turns vs 1 (73% vs 41% post-ad continuation).
- Top ChatGPT advertisers skew B2B SaaS: HubSpot, Fiverr, Preply, Cursor — fundamentally different from Google AI Mode (eBay, Walmart, Best Buy, Amazon) or AI Overview (eBay, Walmart, Temu, Etsy). The conversational format favors brands selling to people doing research/comparison/planning, not transactional shoppers.
- 44% of ChatGPT ads fire on Turn 1; 56% later in conversation — most ads serve mid-session. Compare Google AI Mode at 98.5% Turn 1 (single-query model). Targeting must account for conversation context, not just opening query.
- 83% of ChatGPT ad-triggering queries would never have triggered a Google Shopping ad. Only 2% of headlines say “buy now.” This is brand advertising wrapped in conversation.
- Pharma, Auto, Insurance verticals are still absent on ChatGPT per Similarweb — wide-open category opportunity for early movers.
Access — Self-Serve vs Managed
The public Help Center positions Ads Manager Beta as self-serve. Similarweb’s earlier panel-window data (Mar-Apr 2026) reported “Apply via OpenAI” with a $50K minimum. These are likely sequential, not contradictory:
- Today (May 2026): ads.openai.com is the self-serve entry point with no published minimum. CPC starting at $3-5 max bid.
- Mar-Apr 2026 (Similarweb panel window): Invitation-only access via OpenAI sales contact, $50K minimum spend per pilot.
- Open question: does the $50K-minimum managed tier persist alongside self-serve for larger campaigns, or has it been retired? Help Center is silent.
For brand teams: assume self-serve is the default today. If you want a managed pilot or higher access, contact OpenAI directly — the managed tier may still be available.
Pricing & Auction Mechanics
| Lever | Value |
|---|---|
| CPC starting max bid | $3-5 USD per click |
| CPM default max bid | $60 |
| Auction type | Relevance-weighted, second-price |
| Buying objectives | Reach (CPM) or Clicks (CPC) |
| Conversion tracking | Native (in Ads Manager Beta) + UTM-style static parameters |
The 10-15) and ~2× LinkedIn Sponsored Content (3-5 CPC is in line with mid-tier Google Search competitive verticals. The auction is the same shape Google has run for two decades — bid × Quality Score (here: relevance) determines rank, second-price clears.
Ad Format
Ads render below the conversation. Components per ad:
- Advertiser name
- Favicon (brand logo, 1:1)
- Title (headline)
- Copy (description body)
- Landing page URL
- Image asset (creative)
Closer to a Google PLA / Meta single-image ad than a search text ad. Image is required, not optional.
Targeting: Context Hints
The single biggest mental-model shift from existing search ads. Advertisers don’t bid on keywords — they describe the conversation context where their product is relevant at the ad group level.
Example shape (extrapolated): an ad group for “kitchen remodel” might have context hints like “homeowner researching kitchen renovations,” “comparing IKEA vs. custom cabinetry,” “kitchen design questions for families with kids.” These hints guide matching but don’t guarantee delivery — actual targeting is the relevance match across ad title + copy + landing page + hints against the live conversation context.
This is the AEO/GEO framing applied to paid: write the ad copy and pick the context the way you’d write a citation-worthy page for organic AI search.
Account Hierarchy
Same shape as Google Ads / Meta Ads:
- Campaign — objective, budget, dates, targeting
- Ad Group — ads grouped around theme/intent; context hints attached
- Ad — creative (title + copy + image + landing page)
Resources require submission for review before launching.
Eligibility
| User tier | Sees ads? |
|---|---|
| ChatGPT Free | Yes |
| ChatGPT Go | Yes |
| ChatGPT Plus | No |
| ChatGPT Pro | No |
| Any Business plan (Team / Enterprise / Edu) | No |
| Anyone told to be or predicted to be under 18 | No |
| Geography | Available? |
|---|---|
| United States | Yes |
| Canada | Yes |
| Australia | Yes |
| New Zealand | Yes |
| All other geos | No (at launch, May 2026) |
| Update 2026-09-29 | 40 countries reported (Last Week in AI #256); list not given |
What This Changes for AI Search Strategy
OpenAI Ads turns ChatGPT into a two-surface marketing channel: organic citation visibility (which Similarweb’s 600K-event study mapped) PLUS paid ad placement. For brand teams, this mirrors the Google search reality of 20 years — ranking organically AND running paid is now table stakes.
Strategic consequences:
- Citation-side optimization remains central for Plus/Pro/Business audiences (no ads served).
- Paid + citation are complementary, not competitive — a brand that earns Wikipedia/Reddit/UGC citations and runs context-hint-targeted ads owns more of the ChatGPT conversation surface.
- Context hints are an ad-buyer’s FLUQs equivalent. Same exercise: enumerate the buyer-decision-moment conversations where you should appear. The output then forks: write a citable EchoBlock for organic, or write context hints + ad copy for paid.
- The Free/Go restriction means audience composition matters. B2B SaaS targeting CTOs has very few of them on Free or Go (most are on Plus/Team) — paid will be weak vs organic citations. Consumer brands targeting general-population buyers will see strong addressability on Free.
- Geographic gating limits multinationals. Until EU/UK ships, this is an Americas + ANZ play. Updated 2026-09-29: reportedly 40 countries for viewers and self-serve for advertisers in India, Europe, the Middle East and North Africa (see update above); confirm your target markets in Ads Manager.
Try It
- Audit your context-hint surface. Open ChatGPT (Free or Go account) and ask 10 prompts a typical buyer would ask. Where does an ad fit? What context language describes that decision moment? Capture that — it’s your ad-group context hints.
- **Stand up a 3-5 starting max bid → one ad group, one ad, your strongest landing page. Measure CTR + cost-per-conversion against your existing Meta/Google paid baseline.
- Mirror your highest-converting Google Ads ad group. Take the keywords that work and translate them into context hints + ad creative. The relevance matching is different but the underlying buyer-intent maps over.
- Don’t kill citation-side investment. Plus/Pro/Business users see only citations — and those are the higher-LTV segments for most B2B. Run both surfaces.
- Test bulk upload early. The CSV schema unlocks programmatic campaign management — a natural fit for AI-agent-driven media buying. Pair with the Meta Ads CLI pattern (resources created in PAUSED status, AI agents stage but don’t spend until human flips ACTIVE).
- Reserve EU/UK budget for the next launch wave. Geographic expansion is the obvious roadmap step; allocate paid-ad reserve accordingly. Updated 2026-09-29: the wave has reportedly happened (40 countries; self-serve in Europe per Ep 237). Check whether your EU/UK markets are live and move that reserve into a test.
Implementation
Tool/Service: OpenAI Ads Manager Beta (ads.openai.com) Setup:
- Sign in with existing OpenAI account (work email recommended) or create new
- Complete onboarding + verification
- Add advertiser name + favicon, billing profile + payment method
- Invite team members (members + permissions)
- Create Campaign → Ad Group (with context hints) → Ad (title, copy, image, landing) → submit for review
- Monitor “Serving” / “Not serving” status; iterate via in-line or bulk edits
Cost:
- No platform fee disclosed; pricing is media spend only
- CPC starting max bid: $3-5 USD/click recommended
- CPM default max bid: $60
- Second-price auction means actual clearing price is below max bid
Integration notes:
- Ads API exists at developers.openai.com/ads (API keys managed in Account Settings)
- UTM parameters persist on ad clicks → existing GA4 / Adobe Analytics pipelines work
- Conversion tracking native to Ads Manager Beta (separate setup flow)
- CSV bulk-upload schema documented in Help Center; automation-friendly
- Brand-safety policy at openai.com/policies/ad-policies/
- Submitting campaigns for review = approval gate before launch (similar to Meta’s review queue)
Comparison to Sister Surfaces
| Surface | Buying model | Targeting unit | Audience | Auction |
|---|---|---|---|---|
| OpenAI Ads (ChatGPT) | CPM / CPC | Context hints | Free + Go users; 40 countries as of Sept 2026 (US/CA/AU/NZ at launch) | Relevance-weighted second-price |
| Google Ads (Search) | CPC | Keywords + audiences | All Google searchers | Quality-Score-weighted second-price |
| Google Ads (AI Overviews) | Currently ad-free per Google | n/a | n/a | n/a |
| Meta Ads | CPM / CPC | Interests + lookalikes + custom audiences | All Meta users | First-price (since 2019) |
| CPM / CPC | Job title + company + skills | All LinkedIn users | Second-price | |
| Perplexity | ”Sponsored questions” (limited launch) | n/a (limited disclosure) | All Perplexity users | n/a |
ChatGPT Ads’ closest analog is Google’s text-ads-on-Search surface, with two key differences: (1) the targeting unit is conversational context not keywords, (2) the audience is gated to Free + Go tiers.
Open Questions
- Ad load and frequency. Density of ads per conversation, frequency capping, repeat-exposure controls — none disclosed in landing or basics articles.
- Quality Score equivalent. The auction is “relevance-weighted” but the relevance scoring inputs and how to optimize them are not documented in the public help articles.
- Inventory available. No disclosure of how many conversations/sessions on Free + Go are ad-eligible vs the brand-safety filter cutting them. Inventory shape determines reach modeling.
- Geography expansion timeline. No EU/UK roadmap published. Partly answered 2026-09-29: 40 countries reported for viewers and self-serve expanded to India, Europe, the Middle East and North Africa (both secondary relays). Which 40 countries, and whether UK is among them, is not stated.
- Plus/Pro ad-free durability. OpenAI’s commitment that paid users stay ad-free is the right move for trust, but is it a contractual guarantee or a current product choice?
- Measurement integrity. Conversion tracking is native to Ads Manager — third-party verification (DoubleVerify, IAS, Moat) availability is not mentioned. Ad-fraud and viewability metrics not exposed.
- Creative policy details. “Brand safety” policies are linked but not summarized — what verticals/claims/creatives are blocked? (Crypto? Political? Pharma?)
- API maturity. Ads API exists but the public help center pages don’t yet expose the schema. Programmatic media buying via Claude Code / agents is the natural next-step pattern but feasibility today is unclear.
- Multi-touch attribution. UTM parameters persist on ad clicks — but what about ad views that influence later conversations or organic citation queries? The cross-surface attribution problem is harder in conversational AI than it was on Search.
Related
- Similarweb — Ads in AI: Insights from Real User Behaviour — panel-measured performance data on this product (CTR, implied CPC, engagement vs Google AI Mode, top advertiser verticals, conversational-intent classification).
- Meta Ads CLI — Meta’s official paid-ads CLI; same agent-friendly shape (CSV bulk + API + PAUSED-by-default safety) — best precedent for how to operate OpenAI Ads programmatically.
- Outcome Kit ads agent — outcome-first ad attribution methodology that should generalize across paid surfaces, including ChatGPT.
- Similarweb most-cited-domains LLM citation study — the organic-citation companion to this paid surface; Wikipedia/Reddit/Fandom dominate organic citations, OpenAI now monetizes the same conversation surface for advertisers.
- FLUQs — the FLUQ-identification exercise (enumerate buyer-decision-moment questions) is structurally identical to the “context hints” exercise for paid OpenAI Ads.
- Reddit + AI-citation playbook — paired-channel thesis (one Reddit post earns both Google rank + AI citations) now joined by a third channel: paid placement on the same conversation.
- AI Marketing topic