ChatGPT Ads measurement for B2B: from clicks to companies and pipeline
How to measure ChatGPT Ads in B2B: the OpenAI Pixel and Conversions API, click references and UTMs, company identification, CRM pipeline and incrementality tests.
Measure ChatGPT Ads in B2B on three layers: platform results from the OpenAI Pixel and Conversions API, company identification of the visitors your ads bring, and qualified pipeline in your CRM traced through click references and UTM tags. Judge the channel on cost per engaged account and opportunities, not CTR, because OpenAI reports aggregated results only.
What OpenAI gives you
ChatGPT Ads reporting covers the familiar metrics: spend, impressions, clicks, CTR, CPC, CPM and conversions. OpenAI states that advertisers receive aggregated performance insights and never see conversations or personal data (OpenAI).
For conversions, OpenAI offers two tools, plus third-party measurement integrations:
- The OpenAI Pixel, a browser tag on your site that records actions such as sign-ups, leads and purchases after an ad interaction.
- The Conversions API, which sends conversion events from your server. It is more resilient to browser blocking and lets you send events that happen later, such as a qualified lead in your CRM.
Using both, with deduplication, gives the most complete picture. OpenAI has also added conversion optimisation to bidding, so the events you send shape who the system finds (OpenAI).
What you won't get: the prompts that triggered your ad, the identity of anyone who saw it, or impressions by company. That's by design, and it shapes the rest of your measurement.
The three layers of B2B measurement
| Layer | Question it answers | Source |
|---|---|---|
| Platform | Did the ad deliver and get clicks at a sensible cost? | ChatGPT Ads Manager, OpenAI Pixel, Conversions API |
| Account | Which companies did the ads bring, and did they engage? | Company identification on your website |
| Pipeline | Did those accounts become opportunities and revenue? | CRM, joined by click reference and UTM |
Most B2B teams stop at the first layer. That is the least useful one, because a cheap click from a student and an expensive click from a target account count the same.
Layer 1: set up platform tracking properly
- Install the OpenAI Pixel on every page, or through your tag manager.
- Define a small number of conversions. Demo request, trial signup, contact form. Avoid counting scroll depth or time on page as conversions; they teach the bidding system the wrong lesson.
- Add the Conversions API if you can. Send the same front-end events with an event ID for deduplication, and add later events from your CRM, such as sales-qualified lead.
- Respect consent. Fire tags according to your consent setup and privacy notice.
Layer 2: keep the click reference and identify the company
Preserve the click reference
When someone clicks a ChatGPT ad, the landing URL carries OpenAI's click reference parameter. Add UTM tags to your landing URLs as well, for example utm_source=chatgpt&utm_medium=paid&utm_campaign=ap-automation&utm_content=integration-netsuite.
- Keep the parameters through redirects. Many sites strip query strings on redirect.
- Store them in hidden form fields, so a lead record in your CRM carries the source, campaign and question cluster.
- Store them in first-party storage for a sensible window, so a visitor who converts on a later visit can still be traced.
Identify the company
Most B2B visitors never fill in a form. Company identification on your website matches visits to companies, usually from business network data. For ChatGPT Ads this turns "312 clicks" into "41 companies, 18 in our ICP, 6 on our target list" (an illustrative example).
Be realistic about coverage. Visits from home connections, mobile data and VPNs often can't be matched to a company by any tool. Report an identification rate alongside the company count, so the number is read correctly.
Account metrics worth tracking
- Companies reached: identified companies that arrived from ChatGPT Ads.
- ICP fit: how many of them match your ideal customer profile.
- Target list coverage: how many are on your target account list.
- Engagement: return visits, pages per account, high-intent pages such as pricing.
- Cost per company and cost per engaged account: spend divided by each.
Note one asymmetry with Microsoft: Microsoft Advertising reports impressions by company, while ChatGPT Ads doesn't. For ChatGPT you only see the companies that clicked. See impressions per account for why that matters.
Layer 3: connect to pipeline
- Send the source into the CRM. The hidden fields from layer 2 should land on the lead and the account.
- Attribute at account level. In B2B, the person who clicked may not be the one who signs. Credit the account, then look at all touches on that account.
- Use a window that matches your cycle. A 30-day window understates a channel in a 90-day sales cycle.
- Send qualified stages back through the Conversions API, so bidding learns which clicks become opportunities.
| Metric | Formula | Why it matters |
|---|---|---|
| Cost per engaged account | Spend ÷ engaged accounts from ChatGPT Ads | Compares channels on account value |
| Opportunity rate | Opportunities ÷ companies reached | Shows whether questions attract real buyers |
| Pipeline per spend | Pipeline value ÷ spend | Board-level view, slower to read |
| Assisted rate | Accounts with a ChatGPT ad touch that converted elsewhere ÷ all such accounts | Captures influence beyond last click |
Test for incrementality
Attribution tells you what happened after a click. It doesn't tell you what would have happened without the ad. Two practical tests:
- Geo holdout: run the campaign in some regions and not others with similar profiles, then compare engaged accounts and pipeline.
- On-off periods: pause a stable campaign for a set period and compare branded search, direct visits and identified companies before and after.
Both need several weeks and stable budgets. Neither is perfect, but each is better than trusting last click alone. Keep in mind that the channel is new and OpenAI publishes no B2B benchmarks yet, so your own baseline is the comparison that counts.
Reporting rhythm
- Weekly: spend, clicks, companies reached, ICP fit by question cluster. Fix broken tracking fast.
- Every two weeks: engaged accounts and cost per engaged account. Rewrite or cut clusters.
- Monthly: opportunities and pipeline, assisted influence, and a channel comparison.
Put ChatGPT Ads next to your other channels in one account-level view. An account engagement score helps compare accounts touched by different channels.
How IntentBid helps
IntentBid reads your ChatGPT Ads results from your own account every hour, with 0% fee on media spend. The pixel recognises ChatGPT ad clicks from the click reference and UTM tags and names the company behind each visit where it can. Reporting shows companies reached, cost per company and the question each account came from, and company engagement reporting follows those accounts across every channel. Campaigns are always created paused with compliance checks.
FAQ
Can I see which companies saw my ChatGPT ads?
No. ChatGPT Ads reports impressions per campaign, not per company. You can see which companies clicked, using company identification on your website.
Do I need both the OpenAI Pixel and the Conversions API?
You can start with the pixel. Adding the Conversions API makes tracking more complete and lets you send CRM events such as qualified leads.
Why don't my clicks match my website visits?
Some people leave before the page loads, some block scripts, and redirects can drop parameters. A gap is normal; a large gap usually means a tracking problem.
What's a good CTR for B2B ChatGPT Ads?
OpenAI hasn't published official benchmarks. Compare against your own campaigns and focus on engaged accounts rather than CTR.
Does OpenAI share conversation data for attribution?
No. OpenAI states that advertisers don't receive conversations. Attribution relies on your own tracking and aggregated platform reporting.
Measure ChatGPT Ads by company and pipeline. Start a 14-day trial, no card needed.