LinkedIn
Do not launch a standalone ChatGPT Ads package. It is a margin trap: you will sell a new line item, then spend the next two quarters reconciling reports and doing optimization work you cannot bill for.
Your client asking to advertise in ChatGPT is not asking for a new buyer. Say someone types “best commercial plumber in Austin” into Google on Monday and asks ChatGPT for a vendor shortlist on Tuesday. That is the same commercial question in a different box. Give it two budgets, two trackers, and two reports, and you have volunteered to explain your own double counting every month.
The placement is new. The buyer intent is not.
ChatGPT Ads look new because the placement is new. Ads appear as labeled sponsored cards below the answer, shown to logged-in adults on the Free and Go plans; Plus, Pro, Business, Enterprise, and under-18 accounts stay ad-free. Advertisers buy through a relevance-weighted second-price auction where a better-matched ad can beat a higher bid. The cited starting points are $60 default CPM for Reach and a $3–$5 max CPC for Clicks. Different interface, familiar fight for intent.
There are no keywords to buy. Matching uses conversation context, landing page, ad copy, and 5–15 advertiser context hints per ad group, plus personalization if the user has left it on. I used to tell clients keywords were control. I was wrong about how long that control would last. Ads Manager has no search-terms report, auction insights, or impression share. The reporting described in the draft is limited to spend, impressions, clicks, and CPC, split by device, country, and platform.
That makes copy and landing pages do more of the work keywords used to do. You validate intent on your own site with UTMs, pixel data, and a CRM join. Same buyer question, same answer page, same budget fight.
One buyer can ask in two places
Take a lead-gen client spending $8,000 a month on Google Search. A buyer asks Google for “managed IT support Denver” and asks ChatGPT who handles IT support for a 40-person firm in Denver. That is not two funnels. It is one buyer expressing commercial intent twice.
The sensible test shifts 10–15% from existing search spend, agrees on CPA and a decision date upfront, and reads campaign-level results before and after. Other AI search placements already sit within search buying: Google’s AI Overviews and AI Mode require no new campaign type for eligible Search, Shopping, and Performance Max campaigns, while Copilot serves through existing Microsoft Advertising. Search inventory expanded. Your client did not get a second buyer.
Two channel reports do not make two conversions
Sell ChatGPT as its own line item and you own the reconciliation. Its reporting gives you spend and clicks, with no query, prompt-theme, or intent report to explain what drove them. Your site and CRM can show what happened after a click, but a buyer who visits through Google on Monday and ChatGPT on Thursday is still one person.
Say your channel exports show 62 Google leads and 41 ChatGPT-tagged leads, while the CRM contains 74 unique people. The client asks which number is real. You spend the meeting explaining attribution instead of budgets, bids, and landing pages. One deduped conversion count belongs at the center of the report. Sell two reports as two products and the reconciliation becomes the product you sold.
Where the standalone package breaks
A small budget still demands a full setup
A standalone ChatGPT campaign has a reported US daily floor of $25 per campaign, or about $750 for a month-long test. At a $5 click, that buys 150 clicks. At the $9.29 B2B CPC seen in one early test, it buys about 80; that test found only 5 of 146 organizations were ICP-fit. Other reported ranges run from $3–$22 CPC and $25–$45 CPM. The price varies. The problem does not: a $25-a-day silo buys you a thin read, not a reliable CPA target.
Put that test inside an $8,000 search program and the ChatGPT clicks sit alongside the Google traffic under one business goal. A weak week calls for a budget shift, not a defense of why a separate package has too little data to judge.
The management work does not shrink with the media budget. You still write the context hints, check the landing page, tag every URL, and explain why Creative Enhancements may rewrite a headline or swap a landing page by default unless you switch it off. Four hours a month on a $600 standalone retainer consume a lot of what that fee was meant to cover. The same work inside a bundled search retainer can reuse tracking and pages you already built for Google and Bing.
Separate reporting becomes recurring labor
The reconciliation does not disappear after launch. With no query or prompt-theme report and AI surfaces that do not break out cleanly, neither platform report can settle which touchpoint deserves credit. So you build a spreadsheet to dedupe the CRM records, then a slide to explain the spreadsheet.
Do that once and it is account work. Do it every month for a separate fee and it is a product with terrible margins. One coordinated search program gets one report with one deduped conversion count. Anything else turns you into an attribution explainer instead of a media buyer.
A specialist is hard to justify for one line item
Nobody hires a full-time ChatGPT Ads specialist for six clients spending $750 each. The channel is too small, the tooling too thin, and much of the work is what Google already needed: offers, landing pages, tracking, and follow-up speed. What the deck calls synergy, I call two people doing the same job.
White-label pricing makes the problem visible. Typical PPC fulfillment runs $300–$600 per SMB account and $800–$1,500-plus for mid-market, or 10–20% of spend. Standard resell markup is 30–50%, before $150–$500 setup fees, dashboard add-ons, and creative fees. Put $450 of fulfillment under a $600 ChatGPT-only retainer and you have $150 left to answer Slack for a month. Put the fulfillment inside a $2,000 search retainer and the economics are different.
The margin math vendors leave off the price page
Here is the comparison before the line items. In this example, a standalone ChatGPT retainer brings in $600–$1,000 per client per month, with fulfillment and unbilled hours eating into it. A bundled search retainer brings in $1,800–$2,500 and uses groas’s listed $299 flat per-client agency fee, subject to a five-account minimum, to subcontract Google, Bing, and ChatGPT together. These are retainer examples, not a claim that every agency charges those prices. Platform ad spend is separate from agency revenue in both models.
| Cost or charge | Standalone ChatGPT package | Bundled search on groas | Basis |
|---|---|---|---|
| Subcontract fulfillment, recurring | $300–$600 per SMB account; $800–$1,500+ mid-market | $299 flat per client, five-account minimum | White-label rates / groas agency rate |
| Platform ad spend, recurring and separate from the retainer | $25/day per US campaign, about $750 for a month-long test | The ChatGPT test uses that floor but draws from the coordinated search allocation | Reported daily floor |
| Client retainer, recurring example | $750–$1,000 | $1,800–$2,500 | Illustrative retainers used in this article; compare with the reported 30–50% white-label markup and groas’s listed business search price |
| Difference before other work and overhead | $250–$500, assuming $500 fulfillment | About $1,500–$2,200 after the $299 fee | Arithmetic on the example retainers above; not net profit |
| Setup and add-ons, one-off or variable | $150–$500 setup, plus any dashboard or creative charges; page and tracking work vary | Existing search pages and tracking may be reusable; any new work still has a cost | Reported white-label fees / work described above |
The listed subcontract fee is not the whole cost. Page and tracking builds may sting once. Dedupe, double reports, and Slack explanations recur. Vendors can put a number on setup and leave the reconciliation labor off the price page because you, not they, will do it.
At five clients, the difference is already uncomfortable. Five standalone retainers at $750 bring in $3,750 against $2,250 in fulfillment at $450 each. That leaves $1,500 before overhead and 20 hours of monthly reconciliation work. Five $2,000 bundled search retainers bring in $10,000 against $1,495 in groas fees. That leaves $8,505 before the agency’s other costs, with the ChatGPT work sharing the existing search setup.
At ten standalone clients, four unbilled hours apiece become 40 hours a month. At 25, they become 100. The equivalent bundled retainer examples bring in $20,000 against $2,990 in groas fees at ten clients, or $50,000 against $7,475 at 25. Those differences are not profit forecasts. They show what is left before the rest of the work lands.
The cheapest mistake is overpaying a $400 setup fee. The expensive one is selling ten standalone retainers that each need four unbilled hours a month to defend.
The standalone exception is a short list
I can think of two clients who might justify separation. One spends enough to fund a real holdout: say $50,000 a month on search, with $5,000 ring-fenced for ChatGPT and an analyst to read incrementality. The other has stopped running Google entirely and only wants ads inside chat. Neither describes the usual small-business account.
I ran those accounts for years. Split budgets produce split learnings, and the smaller half rarely gives you enough to act on. If your client spends under $20,000 a month in total and still runs conventional search, there is no useful standalone ChatGPT program to sell. There is a search program with a ChatGPT placement. The argument about keeping AI-enabled margin inside the agency lands here, not on a shiny new rate card.
Put one outcome in the proposal
Stop selling placements. Sell the outcome the client already buys: one coordinated search allocation, one CPA or ROAS target, and one report with deduped conversions.
I would write the test scope plainly: shift 10–15% of search spend into ChatGPT placements for 30 days, take it from a tracked campaign, and judge blended CPA against the prior 30-day baseline. No separate ChatGPT management fee. No second dashboard. The client asks about ChatGPT because they want coverage, not another invoice. If your proposal needs a second budget table, you have already lost the packaging argument.
Ask the subcontractor who owns the reconciliation
Before you resell fulfillment, ask four questions:
- Will you manage Google, Bing, and ChatGPT toward one CPA target, or charge separately to turn the third placement on?
- Will you deliver a deduped conversion count joined to our CRM, or leave us with platform screenshots to reconcile?
- Will you log bid, budget, and page changes with plain-language reasons, the way execution with reasoning should work?
- Is the fee flat per client, or a percentage of spend that rises as the account grows?
If a partner needs a second management package for the same buyer intent, they are selling you the reconciliation work, not just the media work. groas offers Search Ads across Google, Bing, and ChatGPT for one flat fee, which is why the bundled example uses its $299 per-client agency rate.
Sell the standalone package and you may win the pitch because the slide says ChatGPT. Then you face the renewal with two reports that still do not answer whether the search program worked. The client has one budget and one question; they do not need two programs competing inside your agency. Fold ChatGPT into the search retainer and subcontract it flat. If you are still fielding the weekly version of “are we in ChatGPT?”, that is your proposal. Ignore it, and the second report becomes the reason you lose the account.