ChatGPT Ads Manager: Complete Self-Serve Setup Guide (2026)

What Changed on May 5, 2026

For most of 2026, advertising on ChatGPT was a large-enterprise game. When OpenAI launched ads on February 9, only brands willing to commit $200,000 or more, working through managed agency partners like Criteo, DEPT, Jellyfish, and WPP, could participate. The platform was powerful but gated behind a barrier that locked out small businesses, solo founders, and mid-market companies entirely.

That changed in two steps. In April, OpenAI dropped the minimum to $50,000 and began onboarding a second wave of advertisers through a semi-managed process. Then on May 5, 2026, OpenAI removed all spending minimums and opened the self-serve Ads Manager at ads.openai.com to every US advertiser. No agency required. No minimum daily budget. No onboarding queue.

Alongside the self-serve launch, OpenAI shipped several features that make the platform competitive with Google and Meta for performance advertisers:

• CPC bidding – Advertisers can now choose cost-per-click alongside the original CPM (cost-per-thousand-impressions) model, paying only when users actually click through to a landing page.
• Conversion pixel – A lightweight JavaScript tag you place on your site to track post-click conversions (purchases, signups, form submissions).
• Conversions API – A server-side integration for advertisers who need to report conversions without relying on browser-side pixels, similar to Meta’s CAPI.
• Bulk campaign upload – A JSON-based schema for launching dozens or hundreds of campaigns simultaneously, designed for agencies and large advertisers managing multiple brands.

The scale of the opportunity is enormous. ChatGPT now reaches over 900 million weekly active users globally, with the ad-eligible audience (Free and Go tier users in the US) representing hundreds of millions of sessions per day. Over 1,000 brands have already run campaigns through the managed pilot and Criteo’s ad network, and OpenAI has crossed $100 million in annualized ad revenue within its first six weeks of advertising (The Information, March 2026; Digiday, April 2026).

For advertisers who have been waiting for the right moment, this is it. The platform is live, the barrier is gone, and early movers still have an advantage before the flood of new competition arrives. The rest of this guide walks you through every step of the setup process.

Before You Start: What You Need

Before you open the Ads Manager, gather everything you need so you can move through setup without stopping. The most common reason campaigns stall in the first week is that advertisers start building before their assets are ready and end up with half-finished campaigns sitting in draft.

Account requirements

• OpenAI account – A work email is recommended over a personal Gmail or Yahoo address. If your company already uses ChatGPT Team, Business, or Enterprise, you can use the same organizational account.
• Payment method – Credit card, debit card, or company credit line. OpenAI supports Visa, Mastercard, and American Express. Invoicing is available for accounts spending over $10,000/month.
• Business verification – Basic company information including legal name, website URL, and industry category. This is used for ad review and brand safety classification.

Ad asset checklist

Every ad unit in the ChatGPT Ads Manager requires these components. Having them ready before you start saves significant time:

• Headlines – 3 to 50 characters. These are shorter than Google RSA headlines (30 characters) but can stretch longer. Write at least 3 to 5 variations per ad group.
• Descriptions – Up to 100 characters. This is your value proposition in one line. Think of it as the bridge between the headline and the landing page.
• Images – Minimum 256×256 pixels, recommended 512×512. Square format. PNG or JPG. These appear as thumbnails alongside your text, so clarity at small sizes matters more than visual complexity.
• Landing page URLs – The destination users reach when they click your ad. Append UTM parameters before you launch: utm_source=chatgpt, utm_medium=cpc (or cpm), utm_campaign=your-campaign-name.
• Brand favicon – A small icon (at least 64×64) that appears next to your brand name in the ad unit. Use your standard favicon or a simplified version of your logo.

Context hints

Context hints are the ChatGPT equivalent of keywords. Each ad group gets one plain-language description (up to 280 characters) telling the system what kinds of conversations your ad is relevant to. More on how to write effective context hints in the ad group section below.

Budget decision: CPM vs. CPC

You need to decide before you start whether you are optimizing for visibility (CPM) or clicks (CPC). This choice is set at the campaign level and cannot be changed after launch. CPM works best for brand awareness campaigns where impressions and reach are the primary goals. CPC works best when you need measurable clicks and downstream conversions.

Generate all of these assets in under 3 minutes. Lapis creates ChatGPT-format ad creatives (headlines, descriptions, and images) from a single product URL or brief. Instead of starting from scratch, generate 10+ variations and pick the best before you even open the Ads Manager.

Step 1: Create Your Account

Account creation takes five to ten minutes if you have your information ready. Here is the step-by-step process.

Navigate to the Ads Manager

Go to ads.openai.com in your browser. This is the dedicated advertising portal, separate from chat.openai.com and the API dashboard. If you have an existing OpenAI account, click Sign In. If not, click Create Account and follow the standard email verification flow.

Complete onboarding verification

After signing in, the Ads Manager walks you through a four-step onboarding process:

1. Business information – Enter your company’s legal name, website URL, and industry vertical. Select the category that best matches your business from OpenAI’s dropdown (Software, E-commerce, Professional Services, Education, etc.).
2. Advertiser profile – Add your brand name (as it will appear in ads) and upload your brand favicon. The favicon renders at 32×32 in the ad unit, so make sure it is recognizable at that size.
3. Billing setup – Add a payment method and set your billing address. You can optionally set a monthly spending cap as a safety net. This cap pauses all campaigns once reached.
4. Compliance review – Accept OpenAI’s advertising policies. Your account is typically approved within minutes for standard industries. Regulated industries (finance, healthcare, alcohol) may require additional documentation and take 24 to 48 hours.

Invite team members

Once your account is active, you can invite teammates with role-based permissions. The Ads Manager supports three roles: Admin (full access including billing), Campaign Manager (create and edit campaigns, no billing access), and Viewer (read-only reporting access). Add anyone who will be building or monitoring campaigns before you proceed.

At this point your account is live and you are ready to build your first campaign.

Step 2: Build Your First Campaign

The Ads Manager offers two workflows for campaign creation: Guided creation (a step-by-step wizard ideal for your first campaign) and Bulk upload (a JSON-based schema for launching multiple campaigns at once). Start with guided creation for your first campaign, and graduate to bulk upload once you are managing more than five campaigns simultaneously.

Campaign settings

Click Create Campaign and configure the following settings:

• Campaign title – Use a clear naming convention from day one. Recommended format: [Brand]-[Objective]-[Audience]-[Date]. Example: Lapis-CPC-SaaS-Marketers-May2026.
• Objective – Choose Views (CPM) or Clicks (CPC). This determines your bidding model and the metrics the system optimizes for.
• Daily budget – There is no minimum, but OpenAI recommends at least $500/day for campaigns to collect enough data for optimization. Campaigns under $100/day may run inconsistently due to limited bid competitiveness.
• Schedule – Set start and end dates, or choose “Run continuously” for evergreen campaigns. You can pause at any time.
• Country targeting – Currently US only. Canada, Australia, and New Zealand are on OpenAI’s near-term roadmap.

Campaign objective decision matrix

The objective you choose determines how OpenAI’s system optimizes your delivery. Here is how to decide:

If you are unsure, start with CPC. You get direct feedback on what users actually click, and you can measure downstream conversion rates from day one. CPM campaigns require more volume and time to assess impact.

Step 3: Set Up Ad Groups with Context Hints

Ad groups are where ChatGPT advertising gets different from every other platform. On Google, you target keywords. On Meta, you target audience segments. On ChatGPT, you write context hints – plain-language descriptions of the conversations where your ad is relevant.

Each ad group gets one context hint, and that hint has a 280-character limit. This is the single most important piece of text in your entire campaign. A well-written context hint puts your ad in front of high-intent users at exactly the right moment. A poorly written one wastes your budget on irrelevant impressions.

The 4-component formula

The most effective context hints follow a four-part structure: Persona + Intent + Scope + Disqualifier. Here is what each component does:

• Persona – Who is having this conversation? (e.g., “small business owner,” “marketing manager,” “first-time homebuyer”)
• Intent – What are they trying to accomplish? (e.g., “comparing CRM tools,” “planning a kitchen renovation,” “learning to code”)
• Scope – What constraints or context narrow the match? (e.g., “for a team under 20,” “budget under $10,000,” “in Python”)
• Disqualifier – What conversations should be excluded? (e.g., “not enterprise-scale,” “not academic research,” “not DIY”)

5 example context hints

Here are context hints across different industries that follow the formula:

1. SaaS (project management): “A startup founder or team lead comparing project management tools for a remote team of 5–50 people, looking for something that integrates with Slack and has a free tier. Not enterprise procurement.”
2. E-commerce (home office): “Someone furnishing a home office, researching ergonomic desks or chairs in the $300–$800 range, looking for specific product recommendations rather than general ergonomics advice.”
3. Education (online courses): “A working professional exploring online certifications in data science, UX design, or product management, comparing programs by cost and time commitment. Not a student looking for free tutorials.”
4. Financial services (small business): “A freelancer or small business owner asking about bookkeeping software, invoicing tools, or quarterly tax filing. Looking for affordable solutions, not enterprise accounting systems.”
5. Real estate (first-time buyers): “A first-time homebuyer asking about mortgage rates, down payment requirements, pre-approval processes, or comparing lenders. Not an investor or someone looking to refinance.”

The most common mistake

Advertisers coming from Google Ads instinctively write context hints that read like keyword lists: “CRM software, best CRM, CRM comparison, CRM for small business.” This does not work. ChatGPT’s targeting system is semantic, not keyword-based. It analyzes the meaning of conversations, not individual words.

Write your context hint as if you are describing your ideal customer conversation to a colleague. “Someone who runs a 10-person marketing agency and is asking ChatGPT to help them choose a CRM that integrates with HubSpot” is far more effective than “CRM, HubSpot integration, marketing agency CRM.”

For a deeper dive into contextual targeting strategy, including how to build topic clusters and map intent to creative, read our ChatGPT ads targeting guide.

Step 4: Create Your Ads

Each ad group contains one or more ad units. The Ads Manager supports text-plus-image ads, which are the native format for ChatGPT. Ads appear below the AI’s response, clearly labeled as “Sponsored,” and include your brand favicon, headline, description, image, and a click-through link.

ChatGPT ad specs

How ChatGPT ad specs compare to other platforms

If you are migrating creatives from Google, Meta, or LinkedIn, you need to adapt them. The character limits and image specs are different enough that copy-pasting will produce suboptimal results.

The key difference is constraint. ChatGPT gives you less text to work with than almost any other platform, which means every word matters. You cannot compensate with a long primary text block (Meta) or 15 headline variations (Google). You get one headline at 50 characters and one description at 100 characters. Make them count.

3 example ad units

Here is what strong ChatGPT ads look like across different verticals:

SaaS – project management tool:

• Headline: “Ship Faster with Linear” (22 chars)
• Description: “The project tracker built for product teams. 5,000+ teams use Linear to plan, build, and ship.” (95 chars)
• Context hint: “Product managers or engineering leads evaluating project tracking tools for software teams, comparing Jira alternatives, looking for streamlined issue tracking.”

E-commerce – standing desks:

• Headline: “Uplift V2 Standing Desk” (23 chars)
• Description: “Rated #1 by Wirecutter. Adjusts in 12 seconds. Free shipping, 15-year warranty.” (80 chars)
• Context hint: “Someone researching standing desks, sit-stand desks, or ergonomic home office setups in the $400–$1,200 range, looking for product recommendations not general health advice.”

Education – data science bootcamp:

• Headline: “Learn Data Science Online” (25 chars)
• Description: “Go from zero to job-ready in 16 weeks. 92% employment rate. Flexible schedule.” (79 chars)
• Context hint: “A working professional exploring career changes into data science, asking about bootcamps, online programs, or self-study paths. Comparing cost and time commitment, not a CS student.”

For a deep dive on writing high-converting ad copy for ChatGPT, read our ChatGPT ad copywriting guide. For generating image assets at scale, see our guide on creating ChatGPT ad images at scale.

Generate 10+ ad variations in minutes. Instead of writing each headline and description from scratch, use Lapis to generate dozens of on-spec variations from a single product URL or brief. Paste your landing page, select ChatGPT as the target platform, and get production-ready creatives instantly.

Step 5: Launch, Monitor, and Iterate

Once your campaign, ad groups, and ads are configured, click Submit for Review. Here is what happens next.

Submission and approval

OpenAI reviews all ads before they go live. The review process checks for compliance with OpenAI’s advertising policies, including restrictions on misleading claims, prohibited categories (weapons, drugs, adult content), and landing page quality. Most ads are approved within 2 to 6 hours. Ads flagged for manual review may take up to 24 hours.

Once approved, your ad begins serving immediately if your campaign is scheduled to start. You will see the status change from “In Review” to “Active” in the Ads Manager dashboard.

Available metrics

The Ads Manager dashboard provides these metrics in real time:

• Impressions – Number of times your ad was shown
• Clicks – Number of times users clicked through to your landing page
• Spend – Total amount spent to date
• CTR (click-through rate) – Clicks divided by impressions
• Avg. CPC – Average cost per click (for CPC campaigns)
• Avg. CPM – Average cost per thousand impressions (for CPM campaigns)
• Conversions – Post-click conversion events reported by the pixel or Conversions API

Week 1 benchmarks

Based on data from early advertisers in the managed pilot, here are the benchmarks to target in your first week:

• CTR below 0.3% – Your context hints or ad creative are mismatched. Review your context hints for specificity and test new headline/description combinations.
• CTR 0.3%–0.5% – Acceptable for the first week. Focus on testing new creative variations to push toward the 0.5% threshold.
• CTR 0.5%–1.0% – Strong performance. Your targeting and creative are aligned. Focus on scaling budget and expanding ad groups.
• CTR above 1.0% – Top quartile performance. Maintain the winning combinations and duplicate the ad group structure for adjacent contexts.

When to pause vs. adjust vs. scale

Pause when an ad group has spent more than 2× your target CPA without a single conversion, or when CTR drops below 0.1% after 10,000+ impressions. These are signals that targeting or creative fundamentally miss the mark.

Adjust when CTR is in the 0.2%–0.5% range and you believe the targeting is correct but the creative is underperforming. Swap headlines and descriptions first (fastest to test), then images, then context hints.

Scale when an ad group maintains 0.5%+ CTR with positive ROI for 7+ consecutive days. Increase daily budget by 20–30% increments rather than doubling, which can destabilize the algorithm’s optimization.

For a comprehensive optimization framework including bid management, creative rotation, and A/B testing strategy, read our ChatGPT ads optimization playbook.

Bulk Upload: Campaigns at Scale

The guided creation workflow works well for your first 5 to 10 campaigns. Beyond that, the Ads Manager’s bulk upload feature lets you launch dozens or hundreds of campaigns from a single JSON schema.

When bulk upload makes sense

• 10+ campaigns – Managing more than 10 campaigns through the guided wizard becomes tedious and error-prone.
• Agency management – Agencies managing multiple client accounts can template campaign structures and swap in client-specific creatives and context hints.
• Product catalog campaigns – E-commerce brands with large product catalogs can auto-generate one campaign per product category with category-specific context hints.
• Multi-region testing – When OpenAI expands beyond the US, bulk upload will let you launch identical campaign structures across countries simultaneously.

Campaign schema template

The bulk upload schema follows a nested JSON structure: Campaign → Ad Groups → Ads. Each level inherits settings from the parent unless explicitly overridden. The key fields at each level:

• Campaign level: title, objective (views/clicks), daily_budget, start_date, end_date, country
• Ad group level: name, context_hint (280 chars), bid_amount (optional override)
• Ad level: headline, description, image_url, landing_page_url

You can upload the schema via the Ads Manager UI or through the Ads API (currently in beta). The system validates the entire schema before creating any campaigns, so errors in one campaign do not block the others.

Export directly to bulk upload format. Lapis can generate ad creative variations and export them in the ChatGPT Ads Manager bulk upload schema, letting you go from a product catalog to live campaigns in minutes instead of hours. For a step-by-step walkthrough, see our guide on turning a product catalog into ChatGPT ads.

5 Setup Mistakes That Waste Budget

After reviewing hundreds of campaign setups from early advertisers, these are the five most common mistakes that lead to wasted spend. Avoid all of them before you launch.

1. Writing context hints like Google keywords

This is the number one mistake, and it is the most expensive. Advertisers who list keywords (“CRM, best CRM, CRM software, customer relationship management”) instead of writing natural-language descriptions waste budget on loosely matched conversations. ChatGPT’s semantic targeting system interprets your context hint as a sentence-level description, not a keyword list. When you feed it keywords, the system tries to match conversations that contain all of those words, which rarely matches the high-intent conversations you actually want.

Fix: Rewrite every context hint using the Persona + Intent + Scope + Disqualifier formula described above. Test by reading it aloud. If it does not sound like something you would say to a colleague, rewrite it.

2. Using Google/Meta ad copy without adapting to ChatGPT limits

Google RSA gives you 30 characters per headline and 90 per description. Meta lets you write 125-character primary text on top of longer body copy. ChatGPT gives you 50 characters for a headline and 100 for a description, and that is all you get. Advertisers who paste in their Google headlines (which often rely on multiple headlines displayed together) or Meta copy (which depends on the primary text/body split) end up with truncated, unclear messaging.

Fix: Write ChatGPT-native copy from scratch. Your headline should make sense in isolation. Your description should convey one clear value proposition, not a fragmented sentence that depends on supporting text elsewhere in the ad unit.

3. Setting CPM when your goal is clicks (and vice versa)

The objective choice (Views vs. Clicks) is locked after campaign creation. Advertisers who want website traffic but choose CPM end up paying for impressions that do not translate to clicks. Conversely, brand awareness campaigns on CPC pay premium rates for clicks they do not need.

Fix: Match the objective to your KPI. If you measure success by conversions or clicks, choose CPC. If you measure success by reach and frequency, choose CPM. When in doubt, start with CPC because it provides clearer performance signals.

4. Not setting up UTM parameters before launch

Without UTMs, your analytics platform (Google Analytics, Mixpanel, Amplitude) cannot attribute traffic to your ChatGPT campaigns. You end up with “direct” or “unattributed” traffic and no way to measure ROI. This is especially common with advertisers who launch quickly and plan to “add tracking later.”

Fix: Set up UTM parameters on every landing page URL before you create your first ad. Use a consistent structure: utm_source=chatgpt&utm_medium=paid&utm_campaign={campaign-name}&utm_content={ad-group-name}. Build these into your URL template so they are automatically appended to every new ad. For landing page optimization specific to ChatGPT traffic, see our ChatGPT ads landing page guide.

5. Launching with only 1 ad variation per ad group

A single ad per ad group gives you zero ability to test and learn. You cannot tell if your context hint is working because the creative might be the bottleneck, and you cannot tell if the creative is strong because the context hint might be off. You need at least 3 ad variations per ad group to gather statistically meaningful performance data.

Fix: Launch every ad group with 3 to 5 ad variations. Vary headlines and descriptions across variations. Keep the landing page and context hint constant within an ad group so you can isolate creative performance. After 7 days, pause the lowest-performing variation and introduce a new one.

Lapis makes this easy. Generate 10+ ad variations from a single prompt, pick the top 3 to 5, and launch with built-in variety from day one. You can also use the headline testing guide to structure your variation strategy.

Full Setup Checklist

Before you launch your first campaign, confirm you have completed every item on this list:

1. Created your Ads Manager account at ads.openai.com
2. Completed business verification and added billing
3. Uploaded your brand favicon
4. Invited team members with appropriate roles
5. Written 3+ context hints following the 4-component formula
6. Created 3–5 ad variations per ad group (headline, description, image)
7. Set UTM parameters on all landing page URLs
8. Installed the conversion pixel or configured the Conversions API
9. Chosen the correct objective (CPM or CPC) for each campaign
10. Set daily budgets of at least $250 (CPC) or $500 (CPM)
11. Established Week 1 benchmarks and a pause/adjust/scale framework

If you are starting from scratch and need to generate creatives, images, and headlines quickly, Lapis is the fastest way to get launch-ready. Generate your assets, export them in the right format, and go live in the Ads Manager the same day.

For a broader perspective on how ChatGPT ads compare to Google Ads and Meta Ads across pricing, targeting, and creative strategy, read our ChatGPT ads vs. Google Ads vs. Meta Ads comparison. And for the complete overview of ChatGPT advertising including ad formats, audience eligibility, and the Criteo partnership, see our complete guide to ChatGPT ads.