How ads in ChatGPT are placed: formats, who sees them, and context hints instead of keywords
Ads in ChatGPT appear below ChatGPT responses, not inside the answer itself, and OpenAI says they are clearly identified in the experience and remain separate from ChatGPT's responses. Each unit carries the advertiser name, a favicon, a title, copy, a landing page, and an image asset. Since October 26, 2026, Ads Manager also offers a multi-image format as an open beta — 4–6 square images at 1:1 aspect ratio sharing one destination URL — alongside the original single-image ad. Practitioner reporting additionally describes product-feed campaigns that serve ads from a submitted catalog.
OpenAI does not show ads to users on Plus, Pro, or any Business plan, or to accounts identified as belonging to users under 18. Practitioner reporting from early campaigns adds the flip side: delivery effectively reaches users OpenAI believes are 18 or older on the Free and Go tiers, including logged-out sessions. If your buyers skew toward paid ChatGPT tiers, they will not see your ads at all — size the opportunity accordingly.
There is no keyword auction and no query-level report. The system selects ads based on expected relevance and outcomes, weighing the context and intent of the current conversation, the ad's landing page, title, and copy, advertiser-provided context hints, and — when ads personalization is enabled — select signals from the user's broader ChatGPT experience. At the ad-group level, advertisers write context hints describing the conversations, topics, or keywords where their products or services may be relevant. OpenAI is explicit that these hints guide matching but are not exact-match keywords and do not guarantee delivery in specific conversations.
Practitioners who have spent on the platform note there are no keywords, no demographics, and no in-market audiences, and that you cannot see which conversations or themes triggered your ads — which makes ChatGPT Ads feel closer to the early days of Google's Performance Max than to Search. In one head-to-head test documented by Search Engine Land, a keyword-style hint list delivered more often, with slightly higher CTR and lower average CPC, than a list of complete questions — a single test, not a platform rule.
Because matching is contextual, OpenAI recommends building for coverage: a high volume of diverse title and copy variations per offering, each introducing a different angle, with benefit-focused copy that explains the practical value of the offer, who it is for, and when it helps. Two structural notes for media leads: inclusion targeting with custom audiences requires at least 25,000 matched users (exclusions can use smaller audiences), and for campaigns targeting the EEA or Switzerland OpenAI says to avoid custom audiences entirely while ads personalization is unavailable there.
Setting up an OpenAI Ads Manager account: availability by country, verification and account approval
Self-serve buying happens in Ads Manager Beta at ads.openai.com, with two signup paths: enter your website URL to generate a first ad during signup, or skip straight to Ads Manager and build campaigns later. OpenAI maintains a country-by-country availability page, and its FAQ says that if Ads Manager is not available in your country yet you can create an account and be notified when access opens — creating an account does not grant early access. After a US-led rollout, trade coverage tracked UK access in June 2026 and a September 2026 expansion to select countries across Europe, India, the Middle East and North Africa. Expect the list to keep moving during beta.
Ownership is single-creator by design: one person creates the advertiser account, then invites everyone else who needs access. That matters for agencies — practitioner reporting says OpenAI advises agencies not to create accounts on behalf of clients, and that there is not yet an equivalent of Google Ads Manager Accounts or Meta Business Managers. The client creates the account and invites the agency with an Admin (full access), Member (create and manage campaigns), or Viewer (view only) role. A single OpenAI login can belong to multiple advertiser accounts and switch between them, but there is no centralized multi-account view — each account is accessed individually. Once you belong to 10 or more ad accounts, you cannot create another yourself, though you can still be invited to existing accounts.
Three settings are locked at creation: country or region, currency, and time zone. Ads Manager also maintains three names — Account Name (internal), Brand Name (public, shown on ads), Legal Name (billing and verification) — which start identical and can be edited separately in Settings > General after a Sep 25, 2026 update. Changes have teeth: Brand Name edits trigger brand review and pause ad serving until the new name is approved, and Legal Name edits restart business review and may pause serving. OpenAI also notes that billing country, currency, and legal entity details may not be self-serve editable after setup — so get the entity right on day one, because there is no documented portability path for moving an account to a different legal entity or billing country later.
Before anything delivers, billing must be fully set up: a billing profile (business name, billing address, invoice delivery email) plus a payment method. Adding a card may trigger a temporary authorization hold, which is a verification step, not an advertising charge. Account verification is surfaced as a banner in Ads Manager when required, and ads may not serve until it is complete — as with review, budget limits, campaign end dates, and payment issues. Some new self-serve accounts can initially advertise only in their home country; eligibility for other supported countries opens once identity verification is approved and required home-country spend is reached. Ads created during signup go live only after passing review, once billing, account readiness, and campaign settings allow delivery. OpenAI does not publish review turnaround times, so build lead time into launch dates rather than assuming same-day approval. For larger teams, campaign creation supports both a guided flow and bulk upload via a schema template (bulk upload does not currently support product-feed campaigns).
Pricing, bidding and budgets: impression and click billing, maximum bids and budget pacing
A campaign picks one objective — Views (CPM), Clicks (CPC), or Conversions — and the objective sets both optimization and billing. ChatGPT Ads supports impression and valid-click billing: Views bills per 1,000 impressions served, Clicks bills per valid click, and conversion-optimized campaigns choose between click billing (oCPC, billed per valid click) and impression billing (oCPM, billed per 1,000 impressions served). You are never billed per conversion. In the API, the campaign carries the objective and budget while each ad group carries the billing event, bid strategy, and maximum bid.
Advertisers set a maximum bid at the ad-group level — a maximum CPM for Views campaigns or a maximum CPC for Clicks campaigns. OpenAI recommends starting at $3–$5 USD per click for CPC and may show bid-strength guidance when a bid is likely uncompetitive. Selection runs through a relevance-weighted, second-price auction. Practitioners running early campaigns report average CPCs in the $2-$5 range across industries and note that setting a max CPC below $3 draws a may-not-deliver warning, though OpenAI's own API guidance frames low-bid signals as guidance, not a serving restriction. For conversion campaigns, Bid Cap is a conversion-oriented bid used to compete in click auctions — it is not the price charged for a conversion or a guaranteed achieved cost per acquisition. The Ads API also offers Maximize Results strategies (maximize_clicks, maximize_conversions) that drop the manual bid and require a daily budget.
Budgets come in two types. A daily budget is an average per day over a Sunday–Saturday budget week: spend can reach twice the daily figure on a single day and seven times across the week, and unused amounts can roll forward within the week. If a campaign exhausts its weekly allowance early, it may stop delivering until the next budget week even while marked active. A campaign-total budget is the stricter control, capping spend over the campaign's full duration with pacing based on remaining budget and schedule. Minimum daily budgets vary by billing currency — $25 for USD accounts. Budget-type changes are one-way: campaign-total can become daily, but not back, and the API documentation likewise says you cannot switch a daily budget back to a lifetime budget and should create a new campaign instead. OpenAI recommends new advertisers start with a daily budget to monitor delivery before committing to a campaign-total cap. Objectives and conversion events also cannot be edited in place; clone or rebuild.
Self-serve accounts are postpay: spend accrues and the saved card is charged when the account reaches its assigned payment threshold or at month end. Thresholds start low — OpenAI's illustrative progression runs $25 to $50 to $100 to $200 — and may increase automatically with successful payment history; neither advertisers nor support can set them manually. Pausing is not instantaneous either: campaign updates take time to propagate, ads may continue to appear for up to 24 hours after a pause, and costs incurred in that window remain billable. Eligible sales-managed advertisers can move to postpaid invoice terms and set an account-wide daily spending limit in Billing > Account limits — a cap shared across campaigns that does not pace or allocate spend, and one that is not available to self-serve card-billed advertisers.
Conversion measurement and reporting delays
Ads Manager Beta reporting includes impressions, clicks, spend, CTR, average CPC, average CPM, and conversions when conversion measurement is set up — viewable as campaign, ad group, and ad tables, trend charts, and downloadable CSVs for daily or cumulative values.
Timing is the first thing to internalize: impressions, clicks, and CTR refresh approximately every 15 minutes, but spend insights may be delayed 7–8 hours, so a displayed spend of zero does not mean no charges have accrued. Two operational consequences: do not build same-day budget automation on top of Ads Manager spend figures, and do not judge a campaign's afternoon on a $0 readout. Charges can also post after the fact — campaign updates take time to propagate, ads may continue to appear for up to 24 hours after pausing, and reported spend can update after delivery stops without exceeding budget limits.
Conversion tracking is configured at the ad-account level via the Conversions API and/or a JavaScript pixel, and it must be live before conversion-optimized campaigns launch. Those campaigns optimize toward exactly one supported standard conversion event — custom conversion events are not currently supported — and the selected event cannot be changed after campaign creation. OpenAI recommends enabling automatic advanced matching on the pixel to improve conversion matching, and notes that impression-billed conversion campaigns should be evaluated with click-through and view-through conversion reporting where available. Measurement is aggregated by design: advertisers receive performance insights without access to individual conversations.
For your existing stack, add static tracking parameters (UTMs) to landing page URLs — they persist on ad clicks and flow into GA4 or any other analytics tool, which practitioners strongly recommend alongside in-platform reporting. Watch for two noise sources: OpenAI's ad-review crawlers may visit submitted landing pages carrying your tracking parameters, and third-party analytics may classify those visits as paid traffic even though crawler visits are not billable clicks; and card-level impressions in product-feed reporting are reported separately from billable ad impressions.
Finally, calibrate expectations: OpenAI says the platform does not yet have performance benchmarks across advertisers, industries, or campaign types. Practitioner data is early too — one GA4 comparison covering around 1,500 users found ChatGPT Ads traffic less engaged and converting at a lower rate (0.21% vs 3.71%) than Google Ads traffic to the same site. Treat that as one data point from a beta channel, not a ceiling: judge the channel on blended contribution and incrementality rather than head-to-head CVR against a mature auction.
Evaluating agency ad account infrastructure?
Review dedicated agency ad accounts across Meta, TikTok, and Google, with compartmentalized billing, access custody, and human-led escalation.
Explore Agency AccountsChatGPT ad policies: eligible categories, regulated verticals and review
The baseline: all ads — creative, image, and landing page — must comply with OpenAI's Ads Policies, and ads may be rejected if they are misleading, inconsistent, or include disallowed content. Landing pages must use valid, reachable links and must not block OpenAI's user agents (OAI-AdsBot and OAI-SearchBot), so audit bot rules and firewall exclusions before launch.
OpenAI's advertiser-facing basics documentation does not enumerate eligible verticals; it points to the Ads Policies. Practitioner reporting says OpenAI restricts advertising in categories including adult content, alcohol, tobacco, financial services, and gambling — and notes the policy pages and changelog have not always been internally consistent, with legal services as a live example: legal ads were observed running even while a policy page said they were not permitted. Treat any third-party category list as a snapshot: confirm current eligibility in OpenAI's own policy documentation and expect the rules to keep evolving during beta.
On brand safety, OpenAI's stated policy is to place ads only near chats that are safe, appropriate, and aligned with user trust and brand safety, with safeguards designed to prevent ads from appearing in sensitive contexts. Placement control beyond that is limited: qualifying advertisers — not necessarily every advertiser — can use Negative Phrases for narrow placement requirements tied to their own brand policies. There is no keyword-level placement report, so brand-safety verification works differently than the workflows you have built for Google or Meta.
Review gates delivery. Ads go live only after passing review, and several routine actions restart it: Brand Name changes trigger brand review and pause ad serving until approved, Legal Name changes restart business review, and edited ads go back through review. Account verification, billing issues, exhausted budgets, and passed end dates can all stop serving, with status tooltips in Ads Manager surfacing the reason. OpenAI does not publish review turnaround times or approval outcomes, so regulated and sensitive verticals should plan extra lead time and multiple review cycles — and treat each ad on its own merits.
Testing ChatGPT Ads alongside an existing Meta and Google media plan
Placement in the mix. ChatGPT Ads reaches people while they explore options, compare choices, and make decisions, which makes it conceptually closer to Google Search — and specifically to evaluation-stage, mid-to-bottom-funnel media — than to Meta's feeds. Trade reporting frames it the same way. The inventory is real but young: OpenAI announced a $1 billion annualized ads run rate in September 2026, a figure OpenAI said excludes promotional ad credits, while also stating the platform is in beta and early in scaling, with delivery and spend expected to fluctuate as delivery systems improve.
Sizing the test. Entry costs are low — a $25/day minimum on USD accounts and card billing — so the constraint is measurement discipline, not minimums. OpenAI recommends daily budgets for new advertisers so delivery can be monitored before committing to a campaign-total cap, and suggests distributing planned spend across campaigns by priority while leaving enough room to generate meaningful signals. For a team spending $100K+/month on Meta and Google, a defensible starting posture is a low-single-digit percentage of media, ring-fenced from short-term ROI judgment: measurement voices in trade press say dollars stay in test budgets rather than core allocations until cross-channel lift can be modeled, with one measurement executive estimating roughly six months of data to model lift accurately.
Reading results. There are no official performance benchmarks yet; OpenAI says results depend on objective, creative, landing page, relevance, and setup. Benchmark against your search accounts with caveats: in one practitioner test, ChatGPT traffic engaged less and converted at a lower rate than Google Ads traffic to the same site — but it also reached people who were not necessarily searching on Google, which is the actual incrementality question. Keep two facts in mind: running ChatGPT Ads does not impact whether your brand is mentioned organically in ChatGPT, and vice versa, so paid and organic are separate conversations; and measurement is aggregated, with no conversation-level data shared with advertisers.
Operational realities. If your Meta and Google operations assume manager-account hierarchies, shared billing entities, and instant cross-account budget moves, expect friction. Clients own ChatGPT ad accounts and invite agencies; there is no centralized multi-account management; currency, country, and time zone lock at creation; and account-level spend controls are limited. Bulk upload and CSV exports cover part of the workflow gap, but no manager layer matches MCC or Business Manager yet — and OpenAI's public documentation does not describe a way to transfer ownership of an advertiser account, so decide who creates and owns each account before launch, not after. Given the pace of change, re-check OpenAI's documentation before each planning cycle.