LinkedIn
My pick, up front: if you sell leads, jobs or sales calls, hire a done-for-you search service that runs ChatGPT alongside Google and Bing, then agrees in writing what a lead means to your sales team. Not a form fill. Not a click that stayed for 40 seconds.
ChatGPT ads are too new for anyone to own a long shelf of lead-gen case studies worth trusting. The format launched on February 9, 2026, with self-serve access opening to all US businesses on May 5, 2026 (Ads Manager opening). When a provider says it can show real results, I do not want a logo slide. I want its lead-quality test design before you spend: what counts as qualified, how the team will connect that judgment to the ad session, and what must happen before budget goes up.
That is the ranking. The providers that sound strongest on a sales call do not necessarily have the best answer.
Why a ChatGPT ads case study is the wrong opening question
One B2B pilot spent $1,470 over a month and recorded 39,284 impressions, 259 clicks, a $5.68 average CPC, a 0.66% CTR and zero conversions. I used to tell clients that a zero-conversion test meant the channel failed. I was wrong. On a channel this young, I would first ask whether the setup could recognise a valuable lead when one arrived. The figures alone cannot settle that.
The plumbing matters. Ads appear as Sponsored cards below the answer, separate from the organic answer, and only for logged-in adults on Free and Go tiers. There are no keywords to mine. You write Context Hints describing the ideal customer at ad-group level, and OpenAI decides when a conversation matches. Standard reporting stops at impressions and clicks. If a provider treats that as a complete lead-gen report, you will scale blind.
I rank the six options below on three questions:
- What do they count as a lead? A sales-accepted enquiry, booked job or held sales call tells you more than a form fill. Ads Manager does not show pipeline, so someone has to connect ad sessions to what happens afterward.
- Who does the daily work? Advice is useful. It is not the same as setting up tracking, checking lead quality and managing the account when the test is live.
- What earns the next budget dollar? I want a qualified-lead threshold and a cost threshold, agreed before the first increase. Not a good-looking CTR.
That takes UTMs, server-side capture and CRM-stage feedback matched to deal progression (tracking setup). A pixel and Conversions API are part of the conversion plumbing. Keep one distinction straight: the Conversions objective optimizes toward a supported standard event, such as lead submission; it does not support custom events for that objective. Your CRM’s sales-accepted stage is still the test’s quality check and the gate for more spend. Do not let anyone describe a custom CRM stage as though it were already an available bidding objective.
1. Done-for-you search service: best when lead quality needs one owner
This is where I put groas. Yes, I work with them, so read the ranking with that in mind. Its advantage is treating ChatGPT as part of search management, not a separate science project. Search Ads costs $1,499 a month and covers Google, Bing and ChatGPT ads, with bids, keywords where the channel uses them, and budgets worked around the clock, plus a landing page for every ad group. A landing page matters here because the ad cannot tell you enough about the person who clicked. Qualification has to continue after the click and through sales.
Who is it for? A lead-gen business that wants someone working in the accounts rather than delivering a quarterly workshop. You share access to ad accounts, GA4 and Search Console, agree goals on an onboarding call, and the service goes live in your accounts within 24 hours of that call. It is month to month, with no cut of ad spend. If you do not want that level of account access and hands-on involvement, rule it out. The arrangement will irritate you before the first report.
I would still make the lead definition explicit in the agreement. Do not mistake evidence from established search work for a ChatGPT ads case study: groas’s B2B-adjacent record includes an 85% increase in qualified leads over two months. That is a reason to ask how the same discipline will be applied here, not permission to skip the test. Pick this option if you want one accountable owner for lead quality across search surfaces, and get the CRM-stage check and budget gate written down first.
2. Specialist PPC freelancer: best when one good operator has the hours
For a lean B2B or local-services account spending, say, $10k to $20k a month, a good freelancer can beat a big logo. The advantage is focus: one person who already tracks offline conversions can keep UTMs consistent across channels, connect enquiries to Salesforce or HubSpot, and check which of those enquiries sales accepts. That follows the identical-UTM, lifecycle-stage comparison HubSpot recommends. It is much better than calling every click a prospect.
The rule-out is bandwidth, followed closely by proof of process. A freelancer can know exactly what to do and still lack the hours to check it. Ask to see how they have defined and tracked a sales-accepted lead in an account they managed. Then ask how they would use ChatGPT Context Hints; they are not keywords in a new box. If the only metric they can show is cost per form fill, I would keep looking. Cheap management gets expensive when nobody checks what came through the door.
3. Google Ads agency with a ChatGPT line item: best at production, weaker at the test
This provider often looks safest on a sales call. It has designers, copywriters and an established process, so creative production is its advantage. If you need a stack of polished Sponsored cards, that team can produce them.
For lead generation, the rule-out is a Google workflow copied into a different system. ChatGPT uses Context Hints rather than keywords, and early tests had no search-terms report or device and scheduling controls. A clean-looking campaign structure does not give you the same ways to inspect and tighten traffic. If the agency talks through ad production but cannot explain how it will identify a sales-accepted enquiry, the important part is missing.
Fee incentives deserve a question too. I have watched percentage-of-spend retainers make a small test awkward: a $25-a-day minimum test does not generate much management fee at 15% of spend. That does not prove an agency will push you to scale. It does mean you should ask who approves an increase and what qualified-lead result triggers it. If the answer is spend pacing rather than lead quality, you are paying to supply the test budget.
4. AI-visibility or SEO shop: best for citations, not booked jobs
An AI-visibility shop has a real strength: organic citations. If your goal is to appear in answers without paying for each ad click, that work belongs in the conversation. It does not make the shop my pick for a ChatGPT lead-gen test.
Here is the rule-out: ads and organic citations are separate systems. Paying for ads does not improve your chance of an organic citation. Ads appear as Sponsored cards below the answer, not inside it. A visibility report can tell you about mentions and share of voice, but those are not booked jobs. Standard ad reporting stops at clicks unless the team connects traffic to later lead stages.
I would skip this option if booked jobs this quarter are the goal and the shop cannot show a path from ad session to CRM qualification. Ask to see that path, not another visibility dashboard. You can value organic presence without pretending it answers the paid lead-quality question.
5. Consultant or course: best for learning, not for running the account
A good operator can compress the learning curve. For, say, $1,500, you can learn how Context Hints work and why the Conversions objective needs a clean supported event before it is useful. That is the advantage: faster understanding. It is not daily management.
The rule-out is what happens after the call or the last Loom. You still have to write the hints, build landing pages, set UTMs, check what sales accepted, and decide whether to touch bids and budgets. I bought courses like this when I ran accounts myself. The pattern was consistent: the advice could be right and the account could still drift. Drift is a doing problem, not a knowing problem.
Pick this only if you already have a media buyer with hours available each week. Ask who will make the next account change after the engagement ends. If the answer is you, include your time in the price. A course does not become a done-for-you service because the slides include a checklist.
6. In-house hire: best access to sales, slowest way to start a small test
Proximity to sales is the in-house advantage. An employee can ask why 22 form fills produced three working deals and 19 tire-kickers, then carry that answer back to the campaign. A 90-day account of improving Google Ads lead quality without raising budget shows the shape of that work: import the CRM signal, stop rewarding cheap fills, and watch qualification. It is a useful model for the feedback loop, not a ChatGPT ads result.
The rule-out for a first test is the commitment. One hire costs more than most test budgets, onboarding takes months, and ChatGPT volume alone may not provide many conversions to learn from. Early CPMs sat around $60 before falling toward $25 to $45. Even a small test spends real money while the hire learns a channel without a search-terms report.
Pick in-house when search is already a core pipeline source and ChatGPT is an expansion bet, not when you need someone for a first $25-a-day experiment. Ask how many qualified conversions the person will have to work with each month. If the answer is under 30, I would keep the role fractional.
The first-call questions that expose a weak lead-gen plan
I ask the same three questions on an intro call. They are short on purpose; a provider should not need a deck to answer them.
- What exact event will the campaign optimize toward? A supported standard event such as lead submission is an answer. So is an explanation of how the team will check those submissions against sales acceptance. Saying only “clicks” or “form fills” leaves the quality question unanswered.
- Show me the wiring diagram. I want consistent UTMs on ChatGPT and Google, server-side capture, and a CRM stage matched to the ad session. Without that connection, a channel comparison by revenue becomes a comparison by clicks.
- What earns the next budget dollar? Give me a threshold for sales-accepted leads and their cost. Impressions, CTR and launch-era CPMs are not substitutes. Those CPMs fell from around $60 toward $25 to $45 as inventory opened; they do not tell me whether your sales team wants the leads.
Here is the ranking in one screen. The middle columns describe the provider type, not a promise that every firm or freelancer works the same way.
| Rank and provider | What it is most likely to count | Who does the work? | What should gate more budget? |
|---|---|---|---|
| 1. Done-for-you search service | Qualified or booked leads, agreed up front | Service team in your accounts | Agreed cost and count of qualified leads |
| 2. Specialist PPC freelancer | CRM stages, if the operator already uses them | One operator, if hours allow | Sales-accepted leads at an agreed cost |
| 3. Google Ads agency + ChatGPT | Often form fills unless you specify otherwise | Agency team, usually Google-first | Qualified leads, not spend pacing |
| 4. AI-visibility shop | Mentions and clicks unless tracking goes further | Visibility team; ask who manages bidding | CRM-qualified leads, not visibility growth |
| 5. Consultant or course | A definition you must put into practice | You or your media buyer | Your own quality threshold |
| 6. In-house hire | Closest to the sales team’s definition | Your employee | Enough qualified leads to learn from |
One popular option is absent on purpose: the named-agency leaderboard that ranks firms such as Getcito, Single Grain, NoGood, WebFX and Focus Digital by scale or GEO breadth (that style of roundup). It answers who looks substantial, not who will refuse to scale on junk leads. For this decision, biggest is the wrong axis.
The pick for local services, B2B and lead sellers
Local services: pick the done-for-you service or a freelancer who has wired calls before. You live on booked jobs, not attractive click reports. Invoca found that 49% of answered calls from ChatGPT were qualified, about 10 points above the all-channel average. That potential matters only if someone checks which calls became real opportunities and uses the answer to decide whether to spend more.
B2B with a sales team: same pick, stricter gate. Use the Conversions objective with a supported standard lead event, then check the resulting enquiries against the sales-accepted CRM stage before raising the $25-a-day test budget. Do not confuse a recorded submission with a qualified opportunity.
Lead sellers: do not hire an SEO add-on or buy a course to solve an execution problem. Your buyer will audit lead quality, so your manager has to define and check it from day one. A few shops claim live ChatGPT campaigns are generating enquiries through hands-on message and landing-page testing. Ask for the test design behind the claim, not the screenshot.
If you skipped to the end: hire the provider that agrees what qualified pipeline means, shows how it will measure it, and refuses to scale until the test earns it. For most lead-gen businesses reading this, my pick is the done-for-you search service running ChatGPT with Google and Bing under one owner.