8 Facts About ChatGPT Ads Cost Per Lead for B2B

Published 6 November 2026 · Last updated and fact-checked November 2026

Short answer: Published ChatGPT ads cost per lead runs from $92.21 (Stub Group, 2026) to $2,550, and two named B2B tests produced no qualified lead at all. I found no report with a lead-to-customer rate. Whether it pays for B2B is unproven, so test with a kill rule at 3 times your target.

Key takeaways

• Stub Group reports $92.21 per lead against $157.57 on Google, but its clicks cost $0.76, and it says it usually sees $3 to $5 (Stub Group, 2026).

• Workshop Digital spent $2,319 on 739 clicks and logged 0 MQLs, against 2 from Google Search (Workshop Digital, 2026).

• Only 5 of 146 identified organizations behind one B2B campaign matched the advertiser's ideal customer profile (Search Engine Journal, 2026).

• By my calculation, a ChatGPT lead at $92.21 must close at 58.5% of the rate of a $157.57 Google lead just to tie on customer cost.

• OpenAI says it has no performance benchmarks yet, and each figure here comes from a single advertiser's own report (Search Engine Journal, 2026).

What does every published B2B ChatGPT ads result show?

Seven named reports give usable B2B or near-B2B numbers as of 6 November 2026. None gives a lead-to-customer rate. OpenAI opened self-serve ads in May 2026, so the oldest of these reports is only months old, and each comes from a single advertiser, usually an agency with services to sell.

All of these are vendor, agency or trade sources, not peer reviewed. Two figures in circulation need a correction. Localogy prints a 0.03% conversion rate for GrowthLine, while 1 lead from 314 clicks is about 0.32%. An August Search Engine Land piece cites Stub Group GA4 data at 0.21% against 3.71%, a GA4 sample of around 1,500 users. The case study page I read gives 0.8% against 6.64%.

1. What cost per lead do published tests report?

The published dollar figures run from $92.21 at Stub Group to $2,550 at GrowthLine, a spread of about 28 times, and two named B2B tests logged no qualified lead or form fill. Only 2 of the 7 reports give an actual dollar cost per lead, so any single average would be invented.

• Lowest dollar figure: $92.21, Stub Group, one B2B compliance client

• Highest dollar figure: $2,550, GrowthLine, from 1 lead

• Zero results: Workshop Digital ($2,319) and Windmill Strategy ($1,470)

• Relative claim only: InterTeam, about 60% below Google

• OpenAI benchmark: none published as of September 2026

"OpenAI says it does not yet have performance benchmarks across advertisers, industries or campaign types," Brooke Osmundson reported in Search Engine Journal, 2026.

I would not average these. They count different events: a lead at Stub Group, an MQL at Workshop Digital, a form at Windmill. They also sell different things. Workshop Digital and Windmill Strategy were advertising their own agency services to B2B buyers, and Stub Group ran a compliance services client. An average of $92.21, $2,550 and two blanks would mean nothing for your account or anyone else's.

What the range does tell you is that nobody has a stable answer yet, and that includes the platform. Search Engine Journal puts the click range alone at $3 to $22 across advertisers, and that sits underneath every lead figure. When a page quotes one "typical ChatGPT cost per lead," ask which of these tests it came from, then read the original. In my reading, most of the ranking pages repeat the same two or three reports, so the number of sources looks larger than it is.

2. Why is the $92.21 result so low?

Stub Group's $92.21 comes from a $0.76 average click, and the agency says that click cost is at the low end of what it sees. Keep its 0.8% conversion rate and move the click to $3 or $5, and the same math gives a cost per lead of $375 to $625.

• Stub Group conversion rate: 0.8%, or 125 clicks per lead

• Stub Group CPC: $0.76, including one day at $0.06

• OpenAI starting bid advice: $3 to $5 maximum CPC

• Other reported CPCs: $3.14, $5.68 and $8.12

• Same rate at $3.14: $392.50 per lead, my calculation

"We usually see CPCs more in the $3-$5 range, and sometimes significantly higher depending on the industry," Stub Group wrote in its case study, 2026.

Cost per lead is click cost divided by conversion rate, so a low figure needs one of the two to be unusually good. At $0.76 and 0.8%, the sum gives about $95, close to the reported $92.21. At Stub Group, the conversion rate was about one eighth of the client's Google rate, 0.8% against 6.64%. The click cost did all the work, at about one fourteenth of Google's $10.46.

Clicks that cheap may not last. OpenAI's help page recommends a starting maximum bid of $3 to $5, and the agency itself flags a day when clicks spiked for no clear reason at $0.06. Peer-reviewed work on search ads, such as Ghose and Yang in Management Science, 2009, models click rates, conversion rates and click costs as linked, so I would not assume one moves while the others stay still. My calculation at $3 to $5 is a what-if on one client's rate, not a forecast for yours.

3. What do zero-lead tests actually prove?

Zero leads does not mean a channel is worthless, and it does not mean nothing was learned. The Hanley and Lippman-Hand rule says that after zero events in n trials, the 95% upper limit on the true rate is about 3 divided by n. In cost terms, divide spend by 3 to find the likely floor on cost per event.

• The rule: zero events in n trials, 95% upper limit near 3 divided by n

• Workshop Digital: $2,319 divided by 3 gives $773 per MQL or more

• Windmill Strategy: $1,470 divided by 3 gives $490 per form or more

• Barn2: $477 divided by 3 gives $159 per sale or more

• Assumption: every click has a similar chance of converting

"Neither campaign generated a tracked website conversion during the first month," Windmill Strategy wrote in its pilot report, 2026.

The rule comes from Hanley and Lippman-Hand in JAMA, 1983, written for doctors asking what a clean run of cases proves. It carries over to clicks: 739 clicks with no MQL means the true rate is probably under about 0.41%, so at $3.14 a click the cost per MQL is probably $773 or more. That is my calculation and not a figure from Workshop Digital, which rightly says it is too early to write off the channel.

The floors give you something to compare. If your target cost per MQL is $300, Workshop Digital's result is bad news for that campaign setup. If your target is $2,000, it says almost nothing. Windmill also ran on clicks because, at the time, it could not optimize for conversions. OpenAI now lists conversion-optimized campaigns, so a retest could look different. These floors also apply to one setup in one period, and a different offer or landing page is a different test.

4. How far can one lead be trusted?

GrowthLine's $2,550 cost per lead is the highest published figure, and it comes from a single lead on 314 clicks. By my exact Poisson calculation, one observed lead fits a true cost per lead anywhere from about $458 to about $100,000, so this number is a warning sign and not a price.

• Impressions: 46K+, per Localogy

• Clicks: 314 at $8.12

• Leads: 1

• Printed conversion rate: 0.03%; the math gives about 0.32%

• Spend: about $2,550, inferred from clicks and CPC

• Client type: not stated, so B2B is an assumption

"They admit this is a limited sample," Mike Boland wrote of GrowthLine's principals in Localogy, 2026.

Credit to GrowthLine for saying so. The figure still travels without that caveat. Two leads instead of one would have halved it to $1,275, and three would have made it $850, with no change in the campaign at all. That is what very small counts do to a ratio.

Workshop Digital's comparison has the same weakness in the other direction. Its table shows 2 MQLs from Google Search and 0 from ChatGPT, a 100% gap on paper. Two events cannot carry that claim, and the agency itself cautions readers about the comparison. I would treat any cost per lead built on fewer than 5 events as a direction, then state the event count next to the number every time you report it. That threshold is my own working habit, not a statistical law. If your CEO asks for one number from a pilot, give the count first and the cost second.

5. What does a "60% lower than Google" claim leave out?

InterTeam reports over 150 qualified leads for three B2B software clients at about 60% below Google, the strongest positive claim I found. It publishes no spend, no dollar cost per lead, no definition of a qualified lead and no customer data, so a reader cannot rebuild the figure or compare it with any other test.

• Clients: 3 B2B software companies

• Leads: 150+, described as qualified

• Average CPC: about $5

• Offers: ebooks, webinars, comparison pages and booked demos

• Not disclosed: spend, lead split by offer, pipeline, closed deals

"in total we've generated over 150 qualified leads with a cost per lead that's less than half of what we were paying on Google," InterTeam wrote in its case study, 2026.

The offer mix matters most here. An ebook download and a booked demo are not the same kind of lead, and InterTeam does not split its 150+ by offer. If most came from ebooks and webinars, a comparison with Google search leads may be comparing a content download with a sales request. I do not know that this happened, and the page does not say.

InterTeam's claim may well be true, and its FAQ does report real click costs. It just cannot be checked yet. So use four questions on any ChatGPT ads result, whoever publishes it. How much was spent? How many clicks? What event counted as a lead? How many of those leads became customers? Of the 6 lead-generation reports I graded, 2 state both spend and clicks, and none answers the fourth question with a rate. The agency may have the data and good reasons not to publish it. The buyer still has to treat the claim as unverified.

6. Are ChatGPT ad leads as good as Google leads?

Nobody has published the answer as a rate. The one direct statement goes against ChatGPT: Stub Group says its client's Google leads closed at a significantly higher rate, giving a much lower customer cost. A separate campaign found just 5 of 146 identified organizations matched the advertiser's ideal customer profile.

• Stub Group: Google leads closed better; no figures given

• Stub Group's caveat: Google campaigns had years of quality tuning

• Floyd Blaikie's campaign: 146 organizations from 336 clicks, 5 profile matches

• Fit rate: about 3.4% of identified organizations, my calculation

• Break-even: 58.5% of the Google close rate, my calculation

"the leads coming via Google Ads during this date range have converted into customers at a significantly higher rate than the ChatGPT Ads leads," Stub Group wrote in its case study, 2026.

Here is the sum that decides it. Customer cost is cost per lead divided by close rate. At $92.21 against $157.57, the ChatGPT lead is 58.5% of the price of the Google lead, so it needs to close at 58.5% of the Google rate to tie. If Google closes at 5%, ChatGPT has to reach about 2.9%; at 10%, about 5.85%; at 20%, about 11.7%. Stub Group says its client's Google leads gave a much lower customer cost, so on data so far that client missed the line, though the page publishes no rates and the sales cycle is long.

The Blaikie finding gives you a cheaper early check. Company fit shows up on day one, while closed deals can take months. That lines up with Wu, Andreev and Benyoucef in the Journal of Personal Selling and Sales Management, 2026, whose lead model uses demographic and company data before behavior. Visitor identification has its own error rate, so treat the 3.4% as a rough read. It also counts only the organizations the tool could identify, not every click.

7. Which Google number should you compare against?

Use your own Google Ads account from the same period. The WordStream and LocaliQ 2026 benchmark gives a median cost per lead of $66.69 overall and $93.69 for business services, while Stub Group's client paid $157.57. The right comparison could be any of them, so a generic benchmark can flip the verdict.

• All industries: $66.69, a median from 13,474 US campaigns

• Business services: $93.69

• Industrial and commercial: $75.19

• Finance and insurance: $74.44

• Stub Group's client: $157.57, same 30 days as its ChatGPT test

Stub Group's $92.21 looks cheap beside its client's $157.57, and about level with the WordStream median of $93.69 for business services. Both comparisons are true, and they point different ways. That is why your own account is the only baseline worth a budget decision.

Your Google number is not clean either. The eBay experiments by Blake, Nosko and Tadelis found brand keyword ads had no measurable short-term benefit, because many of those buyers would have arrived anyway. If your Google cost per lead includes brand terms, it may look better than the ads really are. Split out non-brand campaigns before you compare.

"Because search clicks and purchase intent are correlated, we show that returns from paid search are a fraction of non-experimental estimates," wrote Blake, Nosko and Tadelis in Econometrica, 2015.

Workshop Digital's warning applies too: a person searching Google has declared a need, and a person in a ChatGPT conversation may not have. Gordon and colleagues in Marketing Science, 2019 also found that observational methods often miss the true ad effect, so a side-by-side report is a starting point, not proof of what the ads caused.

8. How much should you spend on a fair ChatGPT ads test?

Set three stop rules from your target cost per lead before launch: quit with zero leads at 3 times target, quit with fewer than 5 leads at 10 times target, and cap the test at 30 times target. At a $93.69 target, those are $281, $937 and $2,811 of spend.

• Minimum budget: $25 a day per campaign, per InterTeam

• Stop 1: 3 times target with 0 leads, $281

• Stop 2: 10 times target with fewer than 5 leads, $937

• Cap: 30 times target, $2,811, about $94 a day for 30 days

• Customer-level read: about 662 leads per arm, out of reach for most

"The median confidence interval on return on investment is over 100 percentage points wide," wrote Lewis and Rao of 25 ad experiments in The Quarterly Journal of Economics, 2015.

Stop 1 is the zero-event rule from Hanley and Lippman-Hand: at 3 times target with no leads, your true cost per lead is probably above target. Stops 2 and the cap are my own judgement. At the cap, 30 leads would put you at target, which is enough for a rough read on cost, not on customers.

I would not promise a customer-level answer from this test. To tell a 10% close rate from 5.85% at 5% significance and 80% power, my two-proportion calculation needs about 662 leads in each arm. At Stub Group's best-case $92.21 that is about $61,000 on ChatGPT, plus $104,000 on Google at $157.57, and the 10% is a placeholder.

Judge the pilot on company fit and sales-accepted leads, and keep following the deals. Before launch, send form events to OpenAI's conversion measurement and add a CRM source field. If you have the volume, a randomized design such as ghost ads, Johnson, Lewis and Nubbemeyer, Journal of Marketing Research, 2017, is stronger, but small B2B teams will rarely have that scale.

How do the published tests compare?

This matrix grades each report on the four things a buyer needs before trusting it. A report earns a yes only if the page states the item directly. Stub Group is the one lead-generation report that compares with Google in the same period and gives a dollar cost per lead on both sides, though it omits spend.

What does the test math look like at your target?

Your target cost per lead sets every stop rule. These three rows use real benchmark medians and the Stub Group client figure as sample targets. Swap in your own number from your Google account, then multiply by 3, 10 and 30 to get the spend limits for your pilot.

Methodology and limits

I read each report on 6 November 2026 and recorded only what its page states, marking my calculations as mine. Search covered Google results for this query and the reports named by a research pass on 5 November. Reports without B2B numbers, including ecommerce tests, were left out of the results table on purpose.

I found no peer-reviewed study of ChatGPT ads and lead cost. The academic sources here cover search ads, ad measurement, lead scoring and zero-event statistics, and I apply them by analogy. Every ChatGPT number comes from a vendor, an agency or the trade press, and many authors sell paid media services. Conversion tracking, bidding options and prices on this platform change quickly, so treat this page as dated. My calculations assume each click has a similar chance of converting, which real campaigns only roughly meet. I did not run a ChatGPT ads campaign for this article, and I have no first-party ChatGPT ads data to report.

Frequently asked questions

How much do ChatGPT ads cost per lead?

Published B2B figures run from $92.21 at Stub Group to $2,550 at GrowthLine, and two named tests logged no qualified lead or form. OpenAI has published no benchmark. Treat these as single-advertiser reports, and calculate your own number from your click cost and conversion rate.

Are ChatGPT ads worth it for B2B?

Not proven yet. One agency reports leads about 60% below Google, while another found Google leads closed significantly better. I would run a capped pilot only if you already have working Google campaigns, a sales team that grades leads and budget to lose.

What is the minimum budget for ChatGPT ads?

InterTeam and Search Engine Land put the self-serve minimum at $25 a day in the United States. A useful test needs more than the minimum. At a $93.69 target cost per lead, a 30-lead read means about $94 a day for 30 days.

Do ChatGPT ads convert better than Google Ads?

Not in the one published same-period comparison with rates. Stub Group reports 0.8% on ChatGPT against 6.64% on Google for the same client. Its ChatGPT clicks cost far less, which produced a lower cost per lead despite the weaker conversion rate.

How do I track leads from ChatGPT ads?

Set up conversion measurement in OpenAI's Ads Manager, send your form submission as the conversion event, and add a source field in your CRM. Then grade every lead on company fit and sales acceptance, so you can compare quality with Google leads.

Who sets this test up

The test is cheap to run and easy to read badly. I would start with the landing page and tracking, because a test with no CRM source field cannot answer the question that matters. My website lead tracking guide covers that setup in detail.

If a campaign is live and you need to decide what to cut, my free Kill Sheet gives thresholds for that call. For post-click conversion, compare against my B2B conversion benchmarks and the landing page versus website test. I also make ad creative on Conversion Creatives plans at $999 and $2,449 a month, and I have no ChatGPT ads results of my own to sell you yet.

Sources

Peer-reviewed research

• Blake, T., Nosko, C. and Tadelis, S. (2015). Consumer Heterogeneity and Paid Search Effectiveness: A Large-Scale Field Experiment. Econometrica. https://doi.org/10.3982/ECTA12423

• Ghose, A. and Yang, S. (2009). An Empirical Analysis of Search Engine Advertising: Sponsored Search in Electronic Markets. Management Science. https://doi.org/10.1287/mnsc.1090.1054

• Gordon, B., Zettelmeyer, F., Bhargava, N. and Chapsky, D. (2019). A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook. Marketing Science. https://doi.org/10.1287/mksc.2018.1135

• Hanley, J. and Lippman-Hand, A. (1983). If Nothing Goes Wrong, Is Everything All Right? JAMA. https://doi.org/10.1001/jama.1983.03330370053031

• Johnson, G., Lewis, R. and Nubbemeyer, E. (2017). Ghost Ads: Improving the Economics of Measuring Online Ad Effectiveness. Journal of Marketing Research. https://doi.org/10.1509/jmr.15.0297

• Lewis, R. and Rao, J. (2015). The Unfavorable Economics of Measuring the Returns to Advertising. The Quarterly Journal of Economics. https://doi.org/10.1093/qje/qjv023

• Wu, M., Andreev, P. and Benyoucef, M. (2026). Profiling before scoring: a two-stage predictive model for B2B lead prioritization. Journal of Personal Selling and Sales Management. https://doi.org/10.1080/08853134.2026.2633231

Industry and vendor sources, not peer reviewed (read 6 November 2026)

• Stub Group, ChatGPT Ads case study. https://stubgroup.com/case-study/chatgpt-ads-case-study/

• InterTeam Marketing, ChatGPT Ads case study. https://www.interteammarketing.com/case-study/chatgpt-ads-case-study

• Workshop Digital, ChatGPT Ads for B2B, 26 August 2026. https://www.workshopdigital.com/blog/chatgpt-ads-for-b2b/

• Windmill Strategy, first 30 days pilot, 2 September 2026. https://www.windmillstrategy.com/chatgpt-ads-b2b-industrial-manufacturing-pilot/

• Localogy, Mike Boland, 25 August 2026. https://www.localogy.com/2026/08/chatgpt-ads-high-potential-low-performance/

• Search Engine Journal, Brooke Osmundson, 8 September 2026. https://www.searchenginejournal.com/6-months-into-chatgpt-ads-advertisers-still-dont-know-what-good-looks-like/587583/

• Search Engine Land, 10 August 2026. https://searchengineland.com/run-chatgpt-ads-484513

• Barn2, 6 October 2026. https://barn2.com/blog/chatgpt-ads-results/

• WordStream and LocaliQ, 2026 Google Ads benchmarks, 13,474 US campaigns. https://www.wordstream.com/blog/2026-google-ads-benchmarks

• OpenAI Help Center, Ads in ChatGPT: the basics. https://help.openai.com/en/articles/20001207-ads-in-chatgpt-the-basics