> ## Content Index > Fetch the complete content index at: https://cascader.io/blog/llms.txt > Use this file to discover other available public pages before exploring further. # Should a lead-gen client test ChatGPT ads, and how do you judge the result against Google Ads? - URL: https://cascader.io/blog/chatgpt-ads-lead-gen-test-vs-google-ads/ - Published: 2026-11-04T15:00:04.000Z - Updated: 2026-11-04T15:00:04.000Z - Description: Other advertisers' results with ChatGPT ads can't set the bar for your client's test. The client's own Google cost per qualified lead, counted the same way, can. - Author: Rick Hecker - Tags: Cascader Answers, ChatGPT Ads, Lead Generation, Conversion Tracking, Google Ads, PPC Testing If a lead-gen client asks whether to try ChatGPT ads, most of the answer can be settled before anyone spends a dollar. A test is worth recommending when three things hold: - OpenAI will take the client's ads: its category is accepted, and it has a way to buy them from where it's billed, through self-serve Ads Manager or, in some European markets, OpenAI's Ads Solutions team or partners. - The client's buyers use ChatGPT on plans that show ads. - You can follow every ChatGPT lead into the CRM and qualify it by the same rule as Google's. Then the test is judged on cost per qualified lead, counted the way you count Google Search leads, not on the Conversions column in Ads Manager. Set the budget and the point at which you stop before launch, not after a disappointing fortnight. OpenAI says it has no performance benchmarks yet across advertisers, industries or campaign types ([OpenAI Ads FAQ](https://help.openai.com/en/articles/20001220-frequently-asked-questions?ref=cascader.io)). So the glowing and dismal results you hear from other advertisers can't set the bar for your client. The client's own Google cost per qualified lead is that bar. It is the line the test is judged against, not a forecast of where the test will land. *Last verified October 2026\. ChatGPT Ads is in beta, and OpenAI updates its help pages often.* ## Settle three questions before you recommend a test ### Will OpenAI take the client's ads? Start with the category. OpenAI's [ad policies](https://openai.com/policies/ad-policies/?ref=cascader.io) say that during the initial test period, ads are mainly limited to consumer categories such as household goods, local services, travel and experiences, and digital products or education. Regulated services depend on the country: - **In the US:** financial and health services may be allowed case by case for approved advertisers. Legal services are permitted when the advertiser is licensed to practice law in the jurisdiction where the ad is shown. - **Outside the US:** financial and health services ads are generally prohibited, and legal services ads currently are. Ads for individual housing or job listings are prohibited. OpenAI's [eligibility table](https://help.openai.com/en/articles/20001534-troubleshooting-common-onboarding-and-policy-issues?ref=cascader.io) marks several lead-gen staples, among them mortgages, credit cards and personal loans, as restricted to select approved advertisers in the US. It also says OpenAI isn't currently accepting approval applications in certain categories. If your client's category appears in that table, read its row first. Then check where the client is billed. Self-serve Ads Manager requires the legal entity that advertises and is billed to be based in a country on OpenAI's [availability list](https://help.openai.com/en/articles/20001245-ads-manager-availability?ref=cascader.io). In some European markets, OpenAI says advertisers can also get started through its Ads Solutions team, agency partners or technology partners. A new self-serve account may at first advertise only in its home country, until its identity verification is approved and it has reached a required spend there ([Creating campaigns](https://help.openai.com/en/articles/20001210-create-campaigns-for-chatgpt-ads?ref=cascader.io)). A client billed in one country and selling into another may not be able to test where its buyers are. ### Can the buyers see the ads? Ads appear below ChatGPT's responses for users on the Free and Go plans. OpenAI says it is rolling them out to those users gradually, in select regions. Plus, Pro, Business, Enterprise and Edu accounts don't get ads, and neither do accounts identified as belonging to under-18s ([Ads in ChatGPT FAQ](https://help.openai.com/en/articles/20001047-ads-in-chatgpt?ref=cascader.io)). For a home-services or education offer, that may not matter. For a B2B offer whose buyers use ChatGPT through their company's Business or Enterprise workspace, it does: the ads don't appear there. On the buyers' own country, OpenAI's [October 21, 2026 announcement](https://openai.com/index/chatgpt-ads-expands-26-new-countries/?ref=cascader.io) says ChatGPT Ads is now available in 89 countries, where advertisers can access it through technology partners or self-service, and it names only some of them. The user FAQ cited above puts the rollout to Free and Go users in "select regions". Neither page lists the countries where those users are shown ads. What you can confirm is the account side. Country targeting decides where your ads are eligible to appear, and Ads Manager may refuse to publish a campaign when none of its selected countries is one where your account can currently advertise. That shows where your account can advertise, not where buyers are shown ads. Saving a draft doesn't establish that a campaign can deliver, and normal delivery requirements still apply ([Creating campaigns](https://help.openai.com/en/articles/20001210-create-campaigns-for-chatgpt-ads?ref=cascader.io)). ### Can you count qualified leads from it? This gate decides whether the test can answer anything. The ads platform counts the event your site sends, such as a form fill or a booking. It can't see which of those leads turn out to be worth anything; your CRM can. OpenAI says static tracking parameters such as UTMs persist on ad clicks ([the basics](https://help.openai.com/en/articles/20001207-ads-in-chatgpt-the-basics?ref=cascader.io)). Ads Manager can also append macros such as `{campaign_id}` and `{ad_id}` to landing-page URLs ([Measure Results](https://help.openai.com/en/articles/20001214-launch-your-campaign-and-monitor-performance?ref=cascader.io)). If the client's forms or CRM lose those parameters today, fix that before the test. The CRM count is the only scale that weighs ChatGPT against Google, or against Meta, the same way. Here is how much can hide behind ordinary platform numbers. In a B2B test reported by Brooke Osmundson in [Search Engine Journal](https://www.searchenginejournal.com/6-months-into-chatgpt-ads-advertisers-still-dont-know-what-good-looks-like/587583/?ref=cascader.io), Floyd Blaikie's team used visitor deanonymization to identify 146 organizations behind 336 paid clicks. Only five of those organizations matched the advertiser's ideal customer profile. SEJ notes that Ads Manager offered no way to uncover that on its own. If any of the three questions gets a no, your recommendation is "not yet", and you can say exactly what would change it. ## Why one test can look cheaper than Google, equal to it, or 2.5 times the cost Take a hypothetical client whose Google Search campaigns produce qualified leads at $200 each. The client runs a ChatGPT test that spends $1,000, all on clicks old enough for their leads to have been qualified. Ads Manager's Conversions column shows 20\. Google's Search campaigns report $100 per conversion. So ChatGPT's leads cost $50, half of Google's? Here is the same test, with Google counted the same way at each step: If columns extend beyond the page, scroll sideways. With a keyboard, focus the table and use the arrow keys. | | ChatGPT test | Google Search, same period | | --------------------------------------------------------- | ------------ | -------------------------------------------------- | | Spend | $1,000 | $6,000 | | Conversions column in the ad platform | 20 | 60 | | …of which view-through | 15 | 0 (view-throughs aren't in this column for Search) | | Click-through conversions | 5 | 60 | | Leads in the CRM tagged to the channel (assumed to match) | 5 | 60 | | Leads qualified in the CRM, same rule for both | 2 | 30 | | Spend ÷ Conversions column | $50 | $100 | | Spend ÷ click-through conversions | $200 | $100 | | Spend ÷ qualified leads | $500 | $200 | Every figure in the bottom three rows is correct arithmetic, and each one supports a different verdict in the meeting: - **$50 against $100** says ChatGPT is cheaper. - **$200 against the client's familiar $200** says it matches Google. That comparison sets a platform lead against a qualified lead, an easy slip when the client quotes cost per qualified lead from memory. - **$500 against $200** is the only pair counted the same way on both sides. It says 2.5 times the cost. Two steps open the gaps between those rows. ### From $50 to $200: view-throughs leave the count The ChatGPT total includes view-throughs, according to OpenAI's [Measure Results](https://help.openai.com/en/articles/20001214-launch-your-campaign-and-monitor-performance?ref=cascader.io) page, when the view-through reporting window is set to 1 Day. A view-through is counted when someone converts within a day of seeing an eligible ad, and no qualifying ad click gets credit for the same conversion. OpenAI's developer pages for the [Pixel](https://developers.openai.com/ads/measurement-pixel?ref=cascader.io) and the [Conversions API](https://developers.openai.com/ads/conversions-api?ref=cascader.io) describe Conversions as the click-through total instead, with view-throughs as a separate metric. Don't settle that disagreement by assumption. In Edit columns, add the Click-through conversions and View-through conversions columns and see what your total is made of. Google keeps view-throughs out of the Conversions column for Search campaigns. They appear only in the View-through conversions and All conversions columns, apart from Performance Max store goals, App and Demand Gen ([About view-through conversions](https://support.google.com/google-ads/answer/16542520?hl=en&ref=cascader.io)). Why not count ChatGPT's view-throughs anyway? No ChatGPT click got the credit for them, so such a lead usually reached the site through some other door, without your ChatGPT UTM. That might be a direct visit, an organic result, or a Google ad clicked later that Google also counts. OpenAI's own troubleshooting guide notes that [different platforms can credit the same action differently](https://help.openai.com/en/articles/20001604-troubleshooting-conversion-tracking-for-chatgpt-ads?ref=cascader.io). Whether the ChatGPT impression caused any of those leads is a fair question, but a click-based comparison can't answer it. Keep view-throughs visible in the report and out of the verdict. ### From $200 to $500: the CRM decides which leads count Ads Manager counted five click-through leads. The CRM says two of them qualified. This is the step the verdict rests on, and no platform column can take it for you. It is also why the third gate above matters more than any setting in Ads Manager. ### The click window: constant in the table, maybe not in your accounts Nothing in the table varies the click window, but your accounts may. Ads Manager offers a 7-, 14- or 30-day click-through window and a 0- or 1-day view-through window. OpenAI says changing them affects reporting only, not optimization, bidding or billing. Google applies a 30-day click window to Search conversions unless you change it, and allows up to 90 days ([About conversion windows](https://support.google.com/google-ads/answer/3123169?hl=en&ref=cascader.io)). For the platform rows, set ChatGPT's click window to match your Google lead action's window, and set the view window to 0 Day. If Google uses a window Ads Manager doesn't offer, such as 60 or 90 days, the platform rows can't be matched exactly. That costs less than it seems. The verdict comes from the CRM rows, and those rest on the UTM tags of the visit that produced each lead, not on either platform's window. ## Set the budget and the stopping points before launch Two qualified leads is a thin basis for a verdict. To see how thin, measure spend in units of your target cost per qualified lead. That is the client's Google figure: $200 in the example. If ChatGPT were exactly as good as Google for this client, every $200 spent would bring in one qualified lead on average. Some stretches would bring none and some two. Treat qualified leads as arriving independently at that average, and the chance of seeing none at all after spending k times your target is about e^-k: If columns extend beyond the page, scroll sideways. With a keyboard, focus the table and use the arrow keys. | Spend so far, in multiples of your target cost per qualified lead | Qualified leads expected if ChatGPT matches the target | Chance of seeing none anyway | | ----------------------------------------------------------------- | ------------------------------------------------------ | ---------------------------- | | 1× | 1 | 37% | | 2× | 2 | 14% | | 3× | 3 | 5% (about 1 in 20) | | 5× | 5 | under 1% | The first row is the trap. "We spent a whole target CPL and got nothing" sounds like a verdict, but a channel that exactly matches Google produces that result more than a third of the time. By 3× the target, zero qualified leads would happen only about 1 time in 20 if the channel were at your target. That is the early stopping point. Notice what the table doesn't need: a CPC. Cost per click only decides how many clicks, and how many days, the budget buys. That helps on a platform where Osmundson found CPCs that "range from $3 to $22 across advertisers". ### Why the first test is five times the target, and the second 20 times A budget of 3× is enough to act on a zero, but reading an actual count takes more. Set the test budget at five times the target: $1,000 in the example. At 5×, counting without the early stop, a channel exactly at target produces three or more qualified leads about 7 times in 8, and two or fewer about 1 time in 8\. Two or fewer therefore marks a clear shortfall in the test's observed cost per qualified lead: at least 2.5 times Google's. Add the early stop at 3×, and an at-target channel is stopped by one rule or the other about 1 time in 7 in all: about 1 time in 20 by the early stop, and about 1 time in 11 by getting past it and ending with two or fewer at 5×. That is the price of a test this size. A 5× test can't tell a channel 20% better than Google from one 20% worse. At this size, both usually produce between three and seven qualified leads. A second, larger test doesn't settle that either: under the same assumption, a channel 20% costlier than Google's still passes it, early stop included, about 2 times in 3\. What it screens out is a channel much costlier than Google's, and you can fix its rules now. Run it as a fresh test after the first, with the same measure, matured spend only and its own zero-by-3× early stop, and set its budget at 20 times the target: $4,000 in the example, separate from the first test's $1,000\. Its early stop saves budget only if average daily spend stays under (20 − 3) × the target divided by the maturity days: $242.86 a day in the example, at least about 16.5 days for $4,000. - **15 or more qualified leads** is a result worth acting on: a cost per qualified lead of at most about $267 against Google's $200\. Budget ChatGPT as a running channel, still judged the same way. - **14 or fewer** means stop. In this second test, counting its early stop, a channel exactly at Google's cost is stopped about 1 time in 7, the same price as the first test, and a channel at twice Google's cost passes about 1 time in 13\. It is not a guarantee: a channel half again as costly as Google's still passes about 1 time in 3. Back to the first test. The hypothetical spent exactly 5× and produced two qualified leads, so under the brief below it stops. That isn't proof that ChatGPT costs more for this client: an at-target channel gets past the early stop and still ends with two or fewer at 5× about 1 time in 11\. But a test this size can't buy more certainty than that. ### Count only spend that has had time to qualify - **Ads Manager:** OpenAI says attributed conversions can take [24 to 48 hours](https://help.openai.com/en/articles/20001214-launch-your-campaign-and-monitor-performance?ref=cascader.io) to appear. - **The CRM:** qualification takes as long as the client's sales process. If Google leads usually qualify within 14 days, both stopping points count only spend on clicks at least 14 days old. Keep the campaign running while recent clicks mature. The stopping points look only at matured spend, and the campaign-total budget is what limits your exposure. When the budget is spent, wait until the last clicks have matured before reading the result. That has a cost. The early stop can save budget only if average daily spend stays under twice the target divided by the maturity days: $28.57 a day in the example, which stretches the $1,000 test to at least 35 days. Spend faster, and the whole budget is gone before matured spend reaches 3×. The early stop then gives you the answer sooner, but it saves nothing. ## Set the test up so the columns mean something ### Attach the lead event before traffic starts OpenAI says to attach the conversion to the campaign before traffic starts. Once a mismatched setup is corrected, events from it won't backfill ([Measure Results](https://help.openai.com/en/articles/20001214-launch-your-campaign-and-monitor-performance?ref=cascader.io)). Ads Manager's Conversions column also counts only what measurement is set up to count. So a screenshot showing zero conversions tells you nothing about leads until you know an event was attached. ### Pick the event closest to a qualified lead OpenAI's standard lead-type events are `lead_created`, `registration_completed` and `appointment_scheduled` ([Conversions API](https://developers.openai.com/ads/conversions-api?ref=cascader.io)). None of the standard events means "qualified lead". Custom events can be measured, but they can't be an optimization goal ([Conversion-optimized Campaigns](https://help.openai.com/en/articles/20001412-conversion-optimized-campaigns?ref=cascader.io)). So by default, an optimized campaign chases the form or the booking, not the qualification. A booked appointment sits closer to qualified than a form fill does. OpenAI warns, though, that events which happen too rarely make performance harder to evaluate. ### If you optimize for conversions, choose click billing A conversion-optimized campaign works toward one standard event, which you choose at creation and can't change afterwards. You pay per click or per thousand impressions, never per conversion. Choose click billing. OpenAI's help page says delivery then optimizes toward conversions that follow a click, and those are the leads your CRM can check. OpenAI's pages disagree on whether impression billing also optimizes on view-throughs; click billing doesn't depend on the answer. ### Cap spend at the test budget OpenAI recommends that advertisers new to the platform start with a daily budget, so they can watch delivery before committing to a campaign-total budget. For a test with a fixed ceiling, though, the campaign-total budget is the right tool. OpenAI calls it the stricter control for total spend, while a daily budget can spend up to twice its amount on a given day ([Creating campaigns](https://help.openai.com/en/articles/20001210-create-campaigns-for-chatgpt-ads?ref=cascader.io)). If you start with a daily budget as OpenAI suggests, track cumulative spend yourself against the running test's ceiling: 5× the target in the first test, 20× in the second. In US dollars, the minimum daily budget is $25. ## What to put in front of the client Something like this, with your own numbers in the brackets: > We'll test ChatGPT ads on \[one eligible offer\], with a campaign-total budget of \[5 × our Google cost per qualified lead\]. Every ChatGPT click carries UTM tags, so its leads land in the CRM under their own source. We'll judge it the way we judge Google: click-through conversions only, a \[30\]-day click window, and leads qualified by the same rule. We'll count only spend on clicks at least \[14\] days old. If that spend reaches three times our target cost per qualified lead with no qualified lead, we stop. That stop saves budget only if we spend under 2 × our target ÷ \[14\] a day; at a faster pace, it just tells us sooner. At the full budget, two or fewer qualified leads means we stop. Three or more earns a fresh second test with its own budget of 20 × our target, using the same measure, matured spend only and its own zero-by-3× early stop, which saves budget only if we spend under 17 × our target ÷ \[14\] a day. In that test, 15 or more qualified leads means we budget ChatGPT as a running channel, and 14 or fewer means we stop. View-through conversions will be reported, not counted. The brief makes the decision before the data can argue for one. Counting matured spend only, every result lands in one of three places: - **Zero qualified leads by 3× the target:** stop early. A channel at the target stops here about 1 time in 20. - **Two or fewer qualified leads at 5×:** stop. The test's observed cost per qualified lead is then at least 2.5 times Google's. A channel at the target gets past the early stop and still ends here about 1 time in 11. - **Three or more qualified leads at 5×:** a larger test under the same rules, not a channel verdict. - Three leads means $333 each against Google's $200, five means $200, and eight means $125\. None of those counts is far outside what a channel exactly at Google's cost produces at this size. - The second test is a fresh one, budgeted at 20× the target. 15 or more qualified leads there means budget the channel; 14 or fewer means stop.