The Free ChatGPT Ads Course: Everything I Learned Spending $10K
Henry Purchase
Co-Founder
There is no real ChatGPT Ads course on the internet right now.
I checked. Search the keyword and you get Udemy courses written before the platform launched, generic "use ChatGPT to write Facebook ads" content, and a few agency think pieces from teams that have never spent a dollar inside the Ads Manager.
This is the course I wish someone had handed me three months ago when I started running ChatGPT Ads. Account setup, the Ads Manager walkthrough, the conversion pixel bug that catches almost everyone out, how to write context hints that actually convert, and the five mistakes I see people making after watching them inside our community.
It is free. It updates every time OpenAI ships something. And it is based on the campaigns I am running right now while spending $10,000 of my own money to figure out what works.
What you will actually learn
By the end of this guide you will be able to:
- Apply for and get approved on the ChatGPT Ads Manager
- Walk the seven sections of the dashboard without getting lost
- Set up conversion tracking properly (not the broken default version)
- Build your first campaign with the right naming, budget, and bid
- Write context hints that match conversational intent, not Google-style keywords
- Avoid the five mistakes killing most early ChatGPT Ads campaigns
If you finish this article and apply it, you are weeks ahead of every other advertiser in your category. The window for first-mover advantage is real, and it is closing.
Why ChatGPT Ads matter right now
A bit of context before we get into the steps.
OpenAI launched ChatGPT Ads on 9 February 2026. On 5 May 2026, they dropped the $50,000 minimum spend and opened the self-serve Ads Manager to all US advertisers. Two days later they announced the UK, Japan, South Korea, Brazil and Mexico are launching in the coming weeks.
CPMs at launch were $60. They have dropped to around $25 in some categories. CPC bidding is now live at $3 to $5 per click. ChatGPT has roughly 800 million weekly active users, and 85% of them are on the Free or Go tiers where ads appear.
Neil Patel shared a study showing affiliates already account for roughly 10% of ChatGPT Ads on commercial terms. As he put it, if affiliates are buying, the channel is profitable. Affiliates do not waste money on losing platforms.
This is the early-mover moment. Two months from now there will be hundreds of agencies offering ChatGPT Ads as a service. Right now there are barely a dozen who have actually run a campaign. That gap is the opportunity.
Step 1: Get into the Ads Manager
The whole platform lives at ads.openai.com. Click Start Now and you are into the flow.
OpenAI asks for your legal business name, business website, favicon and industry. Have these ready. The sign-up does not have a save-and-continue button, and if you make a mistake on a permanent field, you cannot go back.
Three things you cannot change later: your currency, your time zone, and whether you are signing up as a Business or an Individual. Pick carefully.
Country eligibility right now is limited to the US, Canada, Australia and New Zealand. The UK and four other markets are launching in the coming weeks. If you are outside these markets, you cannot create an account today, but you can use this article to be ready on day one of your country's rollout.
The agency trapdoor
The single most common mistake I see in sign-up is this question:
If you click "I act on behalf of other organizations and clients," OpenAI shows you a blocker. You cannot create an account for a client. The client has to create their own account first, then invite your team in.
If you are an agency, the workaround is to create an account for your own business first. Run a small test on your own brand. Get fluent on the platform. Then when a client invites you in, you already know what you are doing. I wrote a full walkthrough of the account creation flow covering every screen if you want the screenshot-level detail.
Once you submit your business details (EIN for US businesses, equivalent tax ID for other markets when they open), your application goes to review. Mine took three days. Some are reporting up to a week. There is a support email on the review screen if your application sits longer.
Step 2: Walk the dashboard before you launch anything
You will be tempted to skip this. Don't. The Ads Manager has seven sections in the left nav, and knowing where they live saves you time later when you are troubleshooting at 11pm.
Here is what each one does:
Ad Manager is your home screen. Live campaigns, performance overview, the create button. You can spin up campaigns manually or import them in bulk via CSV. The bulk import is particularly useful for agencies because you can export your Google Ads campaigns and reformat them into ChatGPT Ads imports. Clunky today, but workable.
Settings is where you add billing, edit account info, and generate API keys. The API access is one of the most interesting parts of the platform — anything you can do in the dashboard, you can do through the API. This is the layer Focal is built on.
Users is where you invite team members or, if you are a client, where you invite your agency. You can give different team members different access levels.
Change history is the audit log. Every change made on the account, by user, by timestamp. Use this when something breaks and you need to figure out what changed.
Conversions is the most important section in the entire dashboard. We will spend an entire step on this in a minute because the default pixel code has a bug that breaks reporting for most people who do not catch it.
Billing is self-explanatory. Add a card, see your invoices, set your spend limits.
Documentation is the full OpenAI API and tracking docs. Bookmark it. Read the tracking section twice before you set up your pixel.
How the auction actually works
One thing the docs do not explain clearly: ChatGPT Ads run a relevance-weighted auction, not a pure highest-bid auction.
Your bid score is your bid plus your relevance score. A $5 CPC bid with high relevance can beat a $7 CPC bid with low relevance.
Relevance is a function of how well your creative, your headline, your context hints, and your landing page match the conversation the user is having. Tight match wins. Generic match loses, even if you outbid the room.
This is why context hints matter more than your bid. Most people get this backwards.
Step 3: Fix the conversion pixel (this is where 80% of people fail)
This part deserves its own section because the default OpenAI documentation has a problem that will quietly destroy your reporting if you copy and paste without thinking.
Here is what happens. You go to Create > Data source. You name it (something like "Main website"), pick Web, click create. OpenAI generates a block of code.
The code has two parts. The top half is your base pixel — the bit that goes on every page of your site and tracks visitors. The bottom half is a sample conversion event called registration_completed.
OpenAI shows you a single block of code with a Copy button. If you click Copy and paste the whole thing across every page of your site, OpenAI will count every page view as a registration_completed conversion. Your conversion data will be inflated by a factor of 100x. Your bids will move toward the wrong actions. Your campaigns will look like they are winning when they are quietly burning money.
The fix is simple once you know about it. Copy only the top half — the base pixel — and put that in the <head> of every page on your site. Leave the conversion event code behind.
Then go to OpenAI's documentation and click Supported Events. There is a long list of pre-built events: lead_created, appointment_scheduled, purchase_completed, subscription_started, and more. Pick the one that matches your actual conversion goal.
For lead gen, use lead_created. Put that snippet only on your booking-confirmed page or thank-you page — the page someone lands on after they convert. Not your homepage. Not your landing page. The page that fires after the form submission.
If you are unsure how to format the snippet for your specific event, take the three OpenAI documentation pages (data source, supported events, conversion events), paste them into Claude or ChatGPT, and ask it to write the exact pixel for your conversion. That is what I do.
The final step is creating the conversion event inside the Ads Manager. Go to Conversions > Conversion events > Create, name it something like "Form submitted," select your base event, pick your data source, and click create. This is the step that tells the platform "when this fires, count it as a conversion in my campaign."
Skip this step and your pixel is firing into the void. Your campaigns cannot optimise toward something they cannot see.
There is also a Conversions API for server-side event tracking. It is more reliable than the pixel alone, especially with browser tracking restrictions tightening every year. If you are running ads at scale, use both. For your first campaign, the pixel is enough.
Want the one-page version? Get the free ChatGPT Ads Pre-Launch Checklist — every step in this article on a single PDF you can stick on your wall.
Step 4: Build the campaign
The hierarchy is identical to Google or Meta. Campaigns sit at the top with one objective each. Ad groups sit inside campaigns, with their own targeting and landing page. Ads sit inside ad groups, with their own creative.
Naming conventions matter
Use this structure from day one and you will thank yourself when you have 30 campaigns running:
[Brand] - [Objective] - [Geography] - [Launch date]
Example: Focal - Lead gen - US - 2026-05-14
For ad groups: [Audience] - [Theme]. Example: Business owners - broad.
For ads: [Audience] - [Angle] - [Format]. Example: Business owners - FOMO - text+image.
If you are an agency, swap the brand for the client name. The point is being able to find what you need three months from now.
Objective: CPM or CPC
Two options today. Reach (CPM) means you pay per 1,000 impressions. Good for awareness and brand campaigns, and a natural fit for anyone coming from Meta Ads. Clicks (CPC) means you pay per click. Good for direct response, lead gen, and anyone used to Google Ads.
I am testing CPC on most campaigns because I came from Google Ads. Performance teams from Meta backgrounds tend to prefer CPM. Run a $500 test on each and let the data tell you.
Setting your budget and bid
Start small. I recommend $1,000 to $2,000 for a first test if you are spending out of pocket. I am running $10,000 across multiple campaigns because I am stress-testing the whole platform and posting the results inside our community.
For your max CPC bid, $3 to $5 is the current sweet spot for most categories. Premium categories (finance, legal, SaaS) bid $8 to $18. E-commerce and travel sit at $3 to $6. OpenAI shows a "bid competitiveness" indicator while you set this, but in my testing it does not move meaningfully no matter what number you type, so treat it as a rough signal not a rule.
Pick a landing page (and please, not your homepage)
Add a destination URL. HTTPS only.
The biggest mistake I see at this stage is sending every campaign to the homepage. Most ChatGPT users are on mobile. They clicked because the ad spoke to a specific need in a specific conversation. They want a page that matches that intent.
Sending them to your homepage drops conversion rate by 60-80% in our testing. Build a dedicated landing page per campaign. Even a stripped-down one with one offer and one form converts better than a homepage with five competing CTAs.
Context hints: the most important field in the entire platform
Context hints are how ChatGPT decides when to show your ad. They are not keywords. They are descriptions of the conversations, topics, and queries where your product is relevant.
Most people see this field, paste in their Google Ads keyword list, and wonder why their ads do not perform. The platform is not keyword-matched. It is conversation-matched.
The OpenAI helper text says hints "guide matching but aren't exact match targeting rules." That second sentence is the whole game. Your hint is a hint, not a filter. Write it like a description, not a list.
Here is the three-part structure that is working for our campaigns:
1. Open with who you want to target. Be specific. Not "marketers." Try: "Marketing agency owners in the US who run 5 to 10 clients on monthly retainers and want to add ChatGPT Ads as a new service line."
2. List the keywords your target customer actually uses. Pull these from your existing Google Ads or Meta Ads data if you have any. If not, think about what your buyer types into search bars. Things like "how to run ChatGPT Ads," "ChatGPT Ads agency," "what does ChatGPT advertising cost."
3. Add the conversational intent. This is the part nobody else covers. ChatGPT users do not search "best CRM." They ask "what CRM should a 5-person agency use to manage client work without paying for stuff we won't use." Your context hint should describe these kinds of full-conversation moments.
Combine all three sections into one paragraph. Submit. Read your data after seven days and adjust.
Writing the ads
Run at least three ads per ad group. Five is better. With one ad you are not testing anything.
The ad itself has four parts:
Headline — under 30 characters. Past that it gets truncated on mobile. Generic headlines die. Numbers, specific outcomes, and category positioning win.
Description — 15 to 19 words. Two short sentences. Include numbers where you can. Make the offer obvious.
Image — required. Square works best because of how the card renders inside ChatGPT.
Destination URL — add your UTM parameters here. Use this convention: utm_source=chatgpt&utm_medium=cpc&utm_campaign=[campaign name]&utm_content=[ad variant].
A working example from one of our own campaigns:
Headline: ChatGPT Ads Agency Description: First-mover advantage. Free discovery call for businesses ready to launch ChatGPT Ads in the next 30 days. Image: Square brand image, light background
Notice the headline matches the context hints exactly. That improves relevance, which improves the auction score, which lowers our effective CPC. That is the lever.
Why your offer matters more than usual
ChatGPT Ads only show to Free and Go tier users. These users have lower buying intent on average than Plus or Pro users (who do not see ads at all). The conversion model that works on this audience is value-first, contact-capture-second.
Lead magnets work. Free trials work. Free discovery calls work. Hard sells do not.
Build for someone willing to give you an email, not a credit card on the first click.
Step 5: Read the data and avoid the five mistakes that kill most campaigns
Most campaigns start delivering within 12 to 24 hours of approval. There is a 7-hour lag between when ads actually deliver and when data appears in the dashboard, so do not panic if your first morning looks quiet.
Open your campaign, drill into ad groups, then into individual ads. At each level you can see impressions, clicks, spend, CTR, average CPC or CPM, the URL, and your conversions (assuming you set up the pixel correctly in Step 3).
The dashboard's analytics are the platform's weakest area. You cannot pivot data, cannot pull granular daily breakdowns, cannot compare ad groups side-by-side without manual work. The most reliable approach right now is exporting a CSV and pasting it into Claude or ChatGPT and asking for the insights. That is exactly the workflow we are productising inside Focal — talk to your ChatGPT Ads campaigns through MCP from Claude or ChatGPT, get structured analysis back instead of clicking through 30 ad groups. Waitlist is here.
The five mistakes I see most often
After watching hundreds of operators inside our community spin up their first campaigns, these are the patterns that kill results:
Treating ChatGPT Ads like Google Ads. It is not keyword-matched. Stop thinking in keywords and start thinking in conversational intent.
Pasting the default pixel everywhere. We covered this in Step 3. Fix the pixel before you launch or your data will be unusable.
Killing campaigns in week one. ChatGPT Ads need two to four weeks to stabilise. The algorithm is learning who clicks and converts. Three days of data is not data, it is noise. Give every campaign 14 days minimum before you make a decision.
Sending all ads to the homepage. Build a landing page per campaign. Even a simple one beats your homepage.
Running one ad per ad group. You are not testing anything with one ad. Three minimum. Five better.
Where to ask questions
This course stays free and updates every time OpenAI ships meaningful changes to the platform. The Ads Manager is moving weekly. What was true in February is already outdated.
Three places to go if you get stuck:
The FutureProof community is the free Skool community where I post weekly updates and troubleshoot setup problems. 350+ operators inside, growing fast.
The Focal YouTube channel has weekly video breakdowns of every change OpenAI ships, plus tactical reviews of specific campaigns and creative.
The Focal waitlist is for the software side — campaign analysis, the competitor ad library, AI agents that draft creative variations. 1,500+ businesses signed up.
The fastest learners in our community are the people who set up an account, run a $500 test, and bring the data back to the group. Ship first, refine later.
Frequently asked questions
Is there a paid ChatGPT Ads certification?
Not yet. OpenAI has not launched an equivalent of Google Skillshop or Meta Blueprint. When they do, it will likely cover the basics. The deeper operator knowledge — context hints that work, creative angles that convert, the pixel pitfall — will still come from running real campaigns.
Can I take this course if I am outside the US, Canada, Australia or New Zealand?
Yes. OpenAI announced on 7 May 2026 that the UK, Japan, South Korea, Brazil and Mexico are launching in the coming weeks. Use this article to get your pixel, landing pages and creative ready so you can launch on day one of your country's rollout.
How much should I spend on my first test?
$500 to $2,000 over two weeks is the right range. Below $500 you do not have enough data to read. Above $2,000 you are spending into a platform you have not tested yet.
Do I need Google Ads or Meta Ads experience first?
No, but it helps. Existing accounts can be piped into Focal so converting audiences and creatives translate into ChatGPT Ads. If you have neither, the Focal Chrome extension helps you build ChatGPT Ads from scratch using competitor patterns.
What is the difference between this article and Focal?
This article is the free playbook. Focal is the software that runs it for you — competitor ad library, creative variations generated from your converting data, AI agents that draft and ship, dashboards that surface real insight. Waitlist here.
My question is not in the FAQ.
Drop it in the FutureProof community or email support@tryfocal.com.
Last updated: 14 May 2026. This page updates every time OpenAI ships meaningful changes to the Ads Manager. Bookmark it.