---
title: "By May 2028, ChatGPT Ads Will Be Two Jobs"
description: "Buying a ChatGPT ad won’t mean owning the conversation. Three predictions for Sponsored Agents, plus the proof to demand before paying for management."
url: "https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and"
image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/a996e343-c442-4d6b-a522-6caeee3a3478.png"
published: "2026-11-10T05:58:39.948Z"
modified: "2026-11-10T05:58:40.000Z"
---

November 10, 2026 · 10 min read

# By May 2028, ChatGPT Ads Will Be Two Jobs

[](https://groas.com/author/alexander-perelman)

[Alexander Perelman](https://groas.com/author/alexander-perelman)Personally Tested

Head Of Product @ groas

[LinkedIn](https://www.linkedin.com/in/alexander-433793253/)

![A proud doorman holds open a door marked Sponsored, while inside a customer waits before an empty desk chair and a blank nameplate: nobody staffs the room.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/a996e343-c442-4d6b-a522-6caeee3a3478.png)

In this article

1. [Prediction 1: By May 2028, media buying and conversation ownership will have separate price lines](#prediction-1-by-may-2028-media-buying-and-conversation-ownership-will-have-separate-price-lines)
2. [Prediction 2: Website conversion reports will miss part of the sponsored conversation’s contribution](#prediction-2-website-conversion-reports-will-miss-part-of-the-sponsored-conversations-contribution)
3. [Prediction 3: A named person will have to sign for what the agent says](#prediction-3-a-named-person-will-have-to-sign-for-what-the-agent-says)
4. [What I would buy now—and leave unsigned](#what-i-would-buy-nowand-leave-unsigned)

By May 10, 2028, a ChatGPT advertising proposal that bundles everything into one “management fee” will look outdated. Clients will buy two jobs: **buying the slot and managing the sponsored conversation that follows.** I am writing that date down so you can hold me to it.

The reason is mechanical, not hype. OpenAI’s Sept. 16, 2026 announcement defines [Sponsored Agents as an opt-in second conversation](https://openai.com/index/reimagining-advertising-with-ai/): after seeing a relevant ad, a user can choose to start a clearly labeled chat with a business-sponsored agent, ask follow-ups, and follow a link to the website when ready.

I spent years mining negatives at 1am for ecommerce and home services accounts. I know media-buying work when I see it. Deciding what a business-sponsored agent can promise a prospect is different work.

OpenAI has drawn a boundary around it. The [sponsored conversation is walled off by design](https://openai.com/index/reimagining-advertising-with-ai/), distinct from ChatGPT’s independent answers and separate from the original conversation. The ad buys entry to a labeled side room. What the agent says inside that room is a second product.

Right now, that room barely exists as inventory. Sponsored Agents are [being tested with select advertisers in the United States](https://www.pymnts.com/artificial-intelligence-2/2026/openai-tests-sponsored-ai-agents-in-chatgpt-ads/), not offered as generally available inventory. And [OpenAI has not specified what the advertiser receives from the exchange, which model runs the agent, who writes its instructions, or how the unit is priced](https://ppc.land/openai-lets-advertisers-run-chatgpt-ads-from-hubspot-and-shopify/).

That uncertainty does not make the distinction academic. It makes the scope of work more important. A vendor selling the doorway has not established that it can staff the room.

![Media-buying tools beside a sponsored-conversation transcript](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/ef49b74b-33e0-4cbe-b8be-79d4f7fcc263.png)

## Prediction 1: By May 2028, media buying and conversation ownership will have separate price lines

Buying the slot is account work: bids, budgets, product feeds, UTMs, and creative tests. OpenAI is already pushing that work toward [prompt-based creation and HubSpot/Shopify integrations](https://ppc.land/openai-lets-advertisers-run-chatgpt-ads-from-hubspot-and-shopify/), with natural-language campaign creation, AI-suggested copy and imagery from a landing page, and opt-in text customization.

I used to tell clients that setup was the strategy. I was wrong. Setup is the commodity part.

Running the conversation is responsibility work. Someone has to approve what the agent can say about price, availability, and guarantees. Someone has to decide when it stops selling and escalates to a human. Someone has to choose the exact next-step link it offers.

Those are not extra creative variations. They are decisions about what the business is prepared to stand behind. I call that _conversation ownership_, and I expect buyers to stop accepting it as an unexplained item inside a media-management fee.

**The split follows the work, not the vendor’s org chart.** One company might do both jobs. One strategist might oversee both. Neither arrangement makes the responsibilities interchangeable, and a record of running campaigns does not prove a process for approving answers.

The evidence today is the mismatch between increasingly accessible campaign creation and an undeclared conversation-management toolkit. No agency can point to publicly described, platform-provided conversation-quality metrics because OpenAI has not described them. That leaves buyers unable to assume that a familiar Ads Manager workflow covers the second job.

What would prove me wrong? If Sponsored Agents ship with full transcripts, claim controls, and priced units inside Ads Manager, making conversation ownership part of the same manageable service, the case for separate pricing weakens. If that becomes the normal buying arrangement by May 2028, I called the split wrong.

Until then, I would ask for two scopes, two named responsibilities, and two prices. “We manage your ads” is not an answer to “Who approves what the agent says?”

## Prediction 2: Website conversion reports will miss part of the sponsored conversation’s contribution

The split creates a measurement problem before the invoice catches up. [Standard ChatGPT reporting today covers impressions, clicks, spend, CTR, average CPC/CPM, and conversions through Pixel and Conversions API](https://www.contexthints.com/guide/chatgpt-ads-roi.html), without chat content or user-level detail.

That can describe campaign activity. It cannot tell me whether an agent answered a pricing question correctly, created an expectation the website could not meet, or helped a prospect reach a useful next step.

I recognize the broader pattern from early Performance Max work: when the platform owns the assist and the website owns the count, a last-click sheet leaves part of the story out. Add a sponsored conversation in the middle and I have four separate stages to track:

1. **Ad engagement:** the prospect responds to the placement.
2. **Sponsored conversation:** the prospect asks questions and receives answers.
3. **Website handoff:** the agent sends the prospect to the next step.
4. **Qualified outcome:** that activity becomes qualified pipeline or revenue.

Collapsing those stages into “click, then conversion” does not make the middle disappear. It makes it harder to judge.

![Four-stage measurement diagram from ad engagement to qualified outcome](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/571facc3-c427-4489-b809-4b2b91ab6992.png)

Agencies already see the reporting gaps. [Accuracast saw form fills arrive while ChatGPT still showed zero conversions two weeks later](https://digiday.com/marketing/openais-measurement-gaps-are-keeping-chatgpt-ads-budgets-at-test-level/), with roughly 100 platform clicks against 20 in its own tracking. Web Guide Partner reported conversions arriving 24 to 36 hours late, without demographic or prompt data.

OpenAI has responded with more plumbing: [expanded Pixel and Conversions API matching on phone, name, region, postal code, and Android GAID, plus card-level impressions and clicks with product fields through the Insights API](https://searchengineland.com/openai-expands-chatgpt-ads-globally-and-builds-out-its-ad-stack-487125).

Useful. But none of that logs what the agent said.

**Better click matching is not conversation reporting.** I want tracking tied to qualified pipeline rather than raw leads, but I would not mistake a CRM row for evidence that the sponsored conversation worked as intended. The outcome and the exchange need separate scrutiny.

This prediction fails if OpenAI ships conversation-level reporting that ties claims to pipeline. If buyers can follow the exchange through to a qualified outcome, the missing middle becomes visible.

Until then, I would refuse to pay for Sponsored Agent management against a CPA or ROAS claim that the vendor cannot substantiate. Pay for the handoff it can verify. Keep measuring the qualified outcome, but do not pretend that a website report explains the conversation that preceded it.

## Prediction 3: A named person will have to sign for what the agent says

A media buyer can optimize the placement without deciding every answer a prospect might receive. A sponsored conversation removes that comfortable distance.

Say I sell garage doors in Dallas. The agent quotes a price, promises same-day installation, and links to a booking page showing a three-day lead time. Who fixes the mismatch? Who answers the angry call?

What the deck calls “full-funnel ownership,” I call one person whose name is on the approved answers.

That work means claims lists, escalation rules, and a defined next step. It is not bid tweaking with a new label. Before I fund it, I want the proposal to answer:

- Which claims about price, availability, and guarantees are approved?
- When does the agent stop selling and escalate to a human?
- Which page does it send the prospect to next?
- Who reviews transcripts, where access permits?
- Who owns the correction when an answer promises something the business cannot deliver?

The concern is not abstract. [Some advertisers already refuse to sign tracking terms or upload customer lists for custom audiences](https://digiday.com/marketing/openais-measurement-gaps-are-keeping-chatgpt-ads-budgets-at-test-level/). Liability makes lawyers nervous before media does. An agent speaking in the business’s name gives them another reason to ask questions.

**A named owner matters more than a claim of “full-funnel” service.** A vendor can put both jobs on the same slide. That does not establish who takes the call when the agent gets the offer wrong.

What would prove this prediction wrong? OpenAI could absorb that risk. If Sponsored Agents ship with strict claim controls, advertiser-approved scripts, and a clear policy that OpenAI takes responsibility for hallucinations, conversation ownership stays lighter and media buyers may retain it within their existing role.

I doubt that outcome. The platform already separates sponsored conversations from its independent answers. I would not build a buying brief around the assumption that it will also take responsibility for every business promise made inside them.

My practical test is simple: a proposal that prices the media but leaves the conversation owner unnamed is half a proposal. Ask who writes the allowed claims and who gets the call when the agent promises something the business cannot ship.

## What I would buy now—and leave unsigned

Nobody can sell hands-on management of a Sponsored Agent today without live access to that test. General campaign experience is not a substitute. Neither is an AI strategy deck.

Sponsored Agents remain a limited test, with no public pricing, instruction controls, or conversation reporting. That means I would separate consulting about a possible service from execution inside an account. Both can be described honestly. Only one deserves an execution fee.

Before paying a vendor that claims live capability, I would ask for four things:

- **Live account proof.** The Ads Manager account name, date of the last change, and a redacted screenshot. Test access beats a partner badge.
- **Exact format names.** Did the placement run as a bottom ad unit, carousel card, or Sponsored Agent test? “ChatGPT ads” is not specific enough.
- **Auditable tracking.** Pixel and Conversions API firing, intact UTMs, and a CRM row tied to a click. I want evidence beyond a platform report, especially with delayed conversions.
- **A named conversation owner.** If the scope includes an agent, who approves claims and reviews transcripts when that access is available? No name, no signature.

A vendor without Sponsored Agent access can still do useful campaign work. It just cannot sell the second job as though access already exists. I would pay for the test it can run today or walk. I would not prepay for the agent it hopes to run next quarter.

That is why I tell first-time buyers [their first hire should cover campaign execution rather than an agent build](https://groas.com/post/don-t-hire-an-agent-builder-to-run-your-first-chatgpt-ads). The immediate work is less glamorous and more checkable: account setup, feed and product data, prompt coverage that earns the next step, and tracking tied to qualified pipeline rather than form fills.

If I were hiring this month, I would write a two-page brief.

### Page one: Buy the slot

This page covers account build, product-feed cleanup, creative tests from landing-page assets, UTM discipline, and Pixel plus Conversions API tied to qualified pipeline.

**Price the work flat, not as a percentage of spend.** Prompt-based setup keeps getting cheaper; the audit trail matters more. I want to see what changed and who owns the outcome, not pay a larger management bill simply because the media budget grew.

This is where groas fits today without pretending: live campaign execution, tracking integrity, and landing-page work, with a named senior strategist accountable for the outcome and changes logged with reasoning. The scope is work that can be verified, not future agent inventory that cannot be accessed.

That does not make Sponsored Agent management a standard deliverable. The test remains limited to select U.S. advertisers, and public details about pricing, instruction controls, and conversation reporting remain absent. I would rather buy a clearly bounded service than an ambitious label.

### Page two: Own the conversation, conditionally

This page starts only when Sponsored Agent access reaches the account. It names the owner, approved claims, escalation rules, next-step links, and transcript-review responsibilities, subject to the controls actually available.

**I would leave page two unsigned and unfunded until the vendor proves access.** If the promised controls are unavailable, the scope needs to say so. A future capability belongs in a conditional plan, not this month’s management invoice.

Before signing either page, I would also understand [how ChatGPT keeps ads separate from answers](https://groas.com/post/ads-in-llms-explained-why-the-ad-never-touches-the-answer). That boundary is the reason these responsibilities need to be spelled out in the first place.

Check me on May 10, 2028. I expect separate price lines, four-stage measurement, and a named owner for what the agent says. If one integrated service handles both jobs, reporting connects conversations to pipeline, and OpenAI absorbs the conversation risk, I called it wrong.

If I am right, buyers who separate the jobs now will pay less for the learning curve later. Buy the test that can be audited today. Fund the conversation only when someone can demonstrate access and the controls the scope requires. The rest is consulting with a media invoice attached.

## Frequently Asked Questions

### Will ChatGPT ad management and sponsored conversation management cost the same thing?

The article predicts that by May 10, 2028, clients will buy two separate jobs: buying the ad slot and managing the sponsored conversation that follows. Buying the slot covers bids, budgets, feeds, and creative tests, while conversation ownership covers approving what the agent can say about price, availability, and guarantees. The article expects buyers to stop accepting conversation work as an unexplained item inside a media-management fee.

### What is a Sponsored Agent in ChatGPT ads?

A Sponsored Agent is an opt-in second conversation defined in OpenAI's Sept. 16, 2026 announcement. After seeing a relevant ad, a user can choose to start a clearly labeled chat with a business-sponsored agent, ask follow-ups, and follow a link to the website when ready. The sponsored conversation is walled off by design, distinct from ChatGPT's independent answers and from the original conversation.

### Can ChatGPT ads reporting show what the Sponsored Agent actually said to a user?

No. Standard ChatGPT reporting covers impressions, clicks, spend, CTR, average CPC/CPM, and conversions through Pixel and Conversions API, but it does not include chat content or user-level detail. That reporting cannot show whether an agent answered a pricing question correctly or created an expectation the website could not meet.

### How should you measure results when a ChatGPT Sponsored Agent sits between the ad and the website?

The article describes four separate stages to track: ad engagement, the sponsored conversation, the website handoff, and the qualified outcome such as qualified pipeline or revenue. Collapsing those stages into click-then-conversion makes the middle harder to judge. Better click matching is not conversation reporting, and the outcome and the exchange need separate scrutiny.

### Who should be responsible for what a ChatGPT Sponsored Agent promises customers?

The article argues a named conversation owner must sign for what the agent says. That work means claims lists, escalation rules, and a defined next step, plus transcript review where access permits. A proposal that prices the media but leaves the conversation owner unnamed is called half a proposal.

### What proof should I ask for before paying a vendor that claims to manage ChatGPT Sponsored Agents?

The article says to ask for four things: live account proof such as the Ads Manager account name, date of the last change, and a redacted screenshot; exact format names like bottom ad unit, carousel card, or Sponsored Agent test; auditable tracking with Pixel and Conversions API firing, intact UTMs, and a CRM row tied to a click; and a named conversation owner. General campaign experience or an AI strategy deck is not a substitute for live access.

### Should I pay for Sponsored Agent management today if the feature is still in testing?

The article advises paying only for a clearly bounded service that can be verified today, such as account setup, feed and product data, prompt coverage, and tracking tied to qualified pipeline. It recommends leaving any conversation-ownership scope unsigned and unfunded until the vendor proves Sponsored Agent access and the required controls, and pricing the current work flat rather than as a percentage of spend.

## Run Your Marketing On groas

Plug ChatGPT, Claude or any AI agent into groas and let it do the work of a whole marketing team: $19 a month flat with every feature included, and autopilot whenever you’re ready.

[Start For $19](https://groas.com/pricing)

[Or have groas run it for you](https://groas.com/done-for-you)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","description":"groas is the marketing infrastructure for your own AI: it connects your AI to all your marketing and gives it proprietary marketing data collected 24/7. Plug Claude, ChatGPT or any AI agent into groas and run all your marketing from it, for a tiny fraction of the usual cost, with the groas app there whenever you want to see more. It is the layer ChatGPT, Claude and any AI assistant run a business’s online marketing on. Through groas an assistant can do anything you could do natively in Google Ads, Bing Ads, Meta Ads and ChatGPT ads, on your website and in AI search, and, where the owner switches it on, can run connected campaigns by itself, so AI agents do the work a whole marketing team used to. groas also gives those agents large-scale proprietary marketing data it collects around the clock, on any schedule the owner chooses: every ad competitors run on Google and Meta, their landing pages and full funnels with the offers and prices in them, Reddit customer research, AI search answers from ChatGPT, Google AI Overviews, Gemini, Claude and Perplexity, and SEO data from Semrush and Ahrefs, backlinks included ($19 a month, flat, with everything included, for unlimited teammates on one shared workspace and unlimited ad accounts and websites, delivered as an MCP server). Separately, its done-for-you service has senior people and proprietary quantitative systems run paid ads, SEO and AI search for businesses, and white label for agencies, from $5,000 a month.","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and#webpage","url":"https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and","name":"By May 2028, ChatGPT Ads Will Be Two Jobs","description":"Buying a ChatGPT ad won’t mean owning the conversation. Three predictions for Sponsored Agents, plus the proof to demand before paying for management.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/a996e343-c442-4d6b-a522-6caeee3a3478.png"},"breadcrumb":{"@id":"https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and#breadcrumb"},"dateModified":"2026-11-10T05:58:40.000Z","mainEntity":{"@id":"https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"By May 2028, ChatGPT Ads Will Be Two Jobs","item":"https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and#article","url":"https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and","mainEntityOfPage":{"@id":"https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"By May 2028, ChatGPT Ads Will Be Two Jobs","description":"Buying a ChatGPT ad won’t mean owning the conversation. Three predictions for Sponsored Agents, plus the proof to demand before paying for management.","datePublished":"2026-11-10T05:58:39.948Z","dateModified":"2026-11-10T05:58:40.000Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/a996e343-c442-4d6b-a522-6caeee3a3478.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"agent, conversation, sponsored, chatgpt, ads, two, work, chatgpt ads","wordCount":2016,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"My background includes Goldman Sachs and Stanford Computer Science.","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/chatgpt-ads-will-split-into-two-jobs-buying-the-slot-and#faq","mainEntity":[{"@type":"Question","name":"Will ChatGPT ad management and sponsored conversation management cost the same thing?","acceptedAnswer":{"@type":"Answer","text":"The article predicts that by May 10, 2028, clients will buy two separate jobs: buying the ad slot and managing the sponsored conversation that follows. Buying the slot covers bids, budgets, feeds, and creative tests, while conversation ownership covers approving what the agent can say about price, availability, and guarantees. The article expects buyers to stop accepting conversation work as an unexplained item inside a media-management fee."}},{"@type":"Question","name":"What is a Sponsored Agent in ChatGPT ads?","acceptedAnswer":{"@type":"Answer","text":"A Sponsored Agent is an opt-in second conversation defined in OpenAI's Sept. 16, 2026 announcement. After seeing a relevant ad, a user can choose to start a clearly labeled chat with a business-sponsored agent, ask follow-ups, and follow a link to the website when ready. The sponsored conversation is walled off by design, distinct from ChatGPT's independent answers and from the original conversation."}},{"@type":"Question","name":"Can ChatGPT ads reporting show what the Sponsored Agent actually said to a user?","acceptedAnswer":{"@type":"Answer","text":"No. Standard ChatGPT reporting covers impressions, clicks, spend, CTR, average CPC/CPM, and conversions through Pixel and Conversions API, but it does not include chat content or user-level detail. That reporting cannot show whether an agent answered a pricing question correctly or created an expectation the website could not meet."}},{"@type":"Question","name":"How should you measure results when a ChatGPT Sponsored Agent sits between the ad and the website?","acceptedAnswer":{"@type":"Answer","text":"The article describes four separate stages to track: ad engagement, the sponsored conversation, the website handoff, and the qualified outcome such as qualified pipeline or revenue. Collapsing those stages into click-then-conversion makes the middle harder to judge. Better click matching is not conversation reporting, and the outcome and the exchange need separate scrutiny."}},{"@type":"Question","name":"Who should be responsible for what a ChatGPT Sponsored Agent promises customers?","acceptedAnswer":{"@type":"Answer","text":"The article argues a named conversation owner must sign for what the agent says. That work means claims lists, escalation rules, and a defined next step, plus transcript review where access permits. A proposal that prices the media but leaves the conversation owner unnamed is called half a proposal."}},{"@type":"Question","name":"What proof should I ask for before paying a vendor that claims to manage ChatGPT Sponsored Agents?","acceptedAnswer":{"@type":"Answer","text":"The article says to ask for four things: live account proof such as the Ads Manager account name, date of the last change, and a redacted screenshot; exact format names like bottom ad unit, carousel card, or Sponsored Agent test; auditable tracking with Pixel and Conversions API firing, intact UTMs, and a CRM row tied to a click; and a named conversation owner. General campaign experience or an AI strategy deck is not a substitute for live access."}},{"@type":"Question","name":"Should I pay for Sponsored Agent management today if the feature is still in testing?","acceptedAnswer":{"@type":"Answer","text":"The article advises paying only for a clearly bounded service that can be verified today, such as account setup, feed and product data, prompt coverage, and tracking tied to qualified pipeline. It recommends leaving any conversation-ownership scope unsigned and unfunded until the vendor proves Sponsored Agent access and the required controls, and pricing the current work flat rather than as a percentage of spend."}}]}]}
```
