Short answer
Measure AI B2B ads at five levels: delivery (spend share, CPM), engagement (CTR, CPC), visit quality, accounts (companies identified, share that fit your profile, target accounts reached, cost per fitting company) and pipeline (meetings, opportunities, pipeline per $1,000 of media). The headline KPI is cost per fitting company from vendor and decision questions. Early benchmarks are thin: the one published B2B ChatGPT pilot had a 0.66% CTR at $5.68 a click, so treat your first 4–6 weeks as your own baseline.
Why click-through rate misleads
A narrow ad group built on a decision-stage question may have a lower CTR than a broad one, and still bring the only companies worth calling. In B2B most visitors never fill in a form, and AI ads can’t target accounts before the click. So the value shows up after the click, at the account level. A KPI set that stops at CTR will push budget to the wrong questions.
The KPI tree
| Level | KPI | What it tells you |
|---|---|---|
| 1. Delivery | Spend vs budget share by ad group; CPM | Whether buyers ask the question often enough |
| 2. Engagement | CTR, CPC by ad group | Whether the ad answers the question |
| 3. Visit quality | Engaged visits, pages per visit, time on decision pages | Whether the page continues the conversation |
| 4. Accounts | Identification rate, companies identified, ICP fit rate, target list hits, cost per fitting company | Whether the right companies arrive |
| 5. Pipeline | Meetings, opportunities, pipeline per $1,000 of media, win rate, time to opportunity | Whether it pays |
Level 4 is where AI B2B ads differ from search: you choose the problem before the click and learn the company after it. More on identification match rates and ICP scoring.
Early benchmarks
| Metric | Figure | Source |
|---|---|---|
| ChatGPT CTR, published B2B pilot | 0.66% | Windmill Strategy, June–July 2026 |
| ChatGPT CPC, same pilot | $5.68 | Windmill Strategy |
| ChatGPT CPC, reported range | About $3–5 | Index Lab |
| ChatGPT CPM, reported range | About $25–60 | Index Lab |
| LinkedIn CTR and CPC | 0.4–1.0%; $6–15 | ClickMinded, July 2026 |
One pilot in one industry is a reference point, not a target. More in B2B AI ads benchmarks.
A weekly report template
| Section | Contents |
|---|---|
| Headline | Spend, fitting companies, meetings this week vs last |
| By question | Spend, clicks, CPC, companies, fitting companies, cost per fitting company |
| Named accounts | Top fitting companies, the question that brought them, status in CRM |
| Pipeline | New meetings and opportunities with an AI ads touch |
| Changes | Budget moved, ads replaced, questions added, and why |
| Next week | What we’ll test |
See it filled in: our sample ChatGPT ads report.
Attribution that works in long cycles
- Click identifiers and UTMs: keep ChatGPT’s click parameter through redirects and store the ad group in the CRM with every form fill.
- Pixel and Conversions API: deduplicated, so conversion bidding has clean signals. Set-up guide.
- Account-level touch: an opportunity counts as influenced if its account was identified from an AI ad before or during the deal.
- Self-reported attribution: add “AI assistant (ChatGPT, Copilot…)” to “How did you hear about us?”.
- Separate view-through and modelled conversions from click conversions. Why.
Deeper: attribution for AI ads in long B2B cycles.
Turning KPIs into decisions
| Pattern after 4–6 weeks | Action |
|---|---|
| High CTR, few fitting companies | The question attracts the wrong people: tighten context hints |
| Low spend share | Widen phrasings of the same question |
| Fitting companies, no meetings | Fix routing and follow-up, not the ads |
| Low cost per fitting company | Add budget and test neighbouring questions |
Questions
What’s the single most important AI B2B ads KPI?
Cost per fitting company from vendor and decision questions. It combines what you paid, the intent of the question and whether the account is one you want.
Can we measure AI B2B ads without company identification?
Partly: clicks and form conversions. But you’d miss most of the B2B value, because most visitors never fill in a form.
How do we compare AI B2B ads with LinkedIn?
On the same denominators: cost per target account engaged, cost per meeting and pipeline per $1,000 of media. How to compare channels fairly.
Why do Ads Manager and GA4 numbers differ?
They count differently: attribution windows, modelled conversions and consent all play a part. The details.
Facts checked against the sources below on 15 October 2026. AI ad platforms change quickly: if something here no longer matches what you see, tell us and we will update it.
