ChatGPT Ads Return About a Sixth of Google Search's Revenue per Dollar. Here's How to Test That for Your Funnel.

A working paper by Maximilian Kaiser and Christian Schulze, covered by Search Engine Land, Yahoo Finance, and TechXplore in early November 2026, puts a number on what many paid-media operators suspected: revenue per website session from paid ChatGPT visits came in at $0.89 versus $2.45 for paid Google Search in the study's sample. Conversion rates told a similar story: 0.37% for ChatGPT versus 1.17% for Google Search. And when the researchers rolled pricing into the comparison, Google Search generated roughly 6.6 times more revenue per advertising dollar.

That "less than a sixth" framing has been circulating all week. Before you forward it to your CFO (or use it to kill a ChatGPT Ads pilot), it's worth understanding what the data can and can't tell a B2B SaaS team.

What the study actually measured

The dataset spans 1,898 online shops and nearly 10 billion website visits from June 2025 through August 2026. Big sample. But the paid-ad effectiveness slice used a much smaller subset: 102 advertisers with complete revenue data for the head-to-head comparison. Of those 102, 82% recorded zero purchases from ChatGPT ads. Roughly 97% earned less revenue per visit through ChatGPT than through Google Search.

This is observational, not a randomized experiment. The products, audiences, and creative varied across the two channels in ways the study can't fully control for. The sample skews ecommerce, where purchase intent in a Google Search session is high by design. B2B SaaS deals close over weeks or months; a session-level revenue comparison may undercount the value of a first touch that later converts through a sales process.

So the 6.6x gap is real in this dataset. Whether it's real for your pipeline is a different question.

Why cheap clicks can still be expensive pipeline

One case study from StubGroup illustrates the trap. ChatGPT ads delivered a $0.76 CPC and $92.21 cost per lead, while Google Ads came in at $10.46 CPC and $157.57 cost per lead over the same period. On a CPL dashboard, ChatGPT looks like the winner. But Google leads converted to customers at a higher rate (6.64% versus 0.8% on ChatGPT). If your reporting stops at cost per lead, you'd shift budget toward the channel that produces cheaper, weaker leads. That's the classic signal problem: top-of-funnel efficiency masking downstream waste.

For B2B SaaS, the KPI that matters isn't CPL. It's cost per sales-qualified opportunity, or cost per closed-won dollar.

How to structure a test that doesn't lie to you

ChatGPT ads are still in beta. OpenAI says over 100,000 advertisers have used the platform, and the Ads Manager Beta now supports impression-based and CPC bidding with conversion-optimized campaigns. The measurement stack is evolving (pixel, Conversions API, reporting updates as recently as October 2026). Your test design needs to account for a platform that's still changing underneath you.

Setup: Pick one offer, one landing page. Allocate 10-15% of your paid search budget to ChatGPT. Instrument both channels with the same downstream tracking: UTM parameters through CRM to opportunity stage. On ChatGPT, implement the Conversions API (not just the pixel) so you're capturing server-side events.

Success metric: Cost per sales-accepted opportunity (SAO). Guardrails: If ChatGPT CPL exceeds 3x Google Search CPL with zero SAOs after two weeks, pause and diagnose creative and targeting before continuing. Stop-loss: If total spend hits $3,000 with no MQL-to-SQL conversion, kill the test and document findings.

The study's parity estimate suggests ChatGPT CPC would need to be about $0.26 to match Google Search's revenue-per-ad-dollar under the paper's assumptions. Your B2B math will differ, but that number gives you a reference point for what "cheap enough" might mean.

ChatGPT ads probably aren't keyword-intent capture. They're closer to mid-funnel, contextual discovery. A user asking ChatGPT "what's the best way to automate lead routing" isn't typing a transactional query into Google; they're exploring. That distinction matters for creative, landing page design, and the qualification bar you set.

If you treat ChatGPT like a Google Search replacement, you'll be disappointed. If you treat it as a net-new consideration channel, you need a different attribution model and a longer measurement window. Neither approach is wrong, but conflating them will produce misleading readouts.

The Kaiser and Schulze paper gives us a useful baseline, especially the conversion rate and revenue-per-session gaps. But a baseline from nearly 1,900 ecommerce shops isn't a verdict on your B2B SaaS funnel. The only verdict that matters is the one your own data produces, measured at the stage where revenue actually happens.