OpenAI Begins Testing Sponsored Agents Inside ChatGPT
The company rolled out AI-powered ad tools and an Ads Manager that builds campaigns using natural language. Shopify and HubSpot are the first partners.

OpenAI announced on Wednesday that it has begun testing advertiser-sponsored agents inside ChatGPT and rolled out a set of AI tools for media buyers. It's the company's most concrete step yet toward a market it had been avoiding discussing publicly: advertising inside the assistant. For anyone working with growth and paid media in Brazil, the takeaway is that a new channel is being born, and it's worth understanding the mechanics before it turns into a crowded auction.
According to Reuters, via Economic Times, the move has two fronts: the Sponsored Agents feature for the end user and the ChatGPT Ads Manager for the advertiser. The two sketch out an ad format quite different from the banner or the sponsored link the market is used to.
What Sponsored Agents Are
The sponsored agents feature will let the user start a conversation with an agent backed by a brand and, after interacting with the ad inside ChatGPT, be directed to the company's website. It's not a static banner: it's a conversation with the brand's AI agent embedded in the assistant experience.
OpenAI is careful to mark the boundary. According to the company, the agents will be clearly labeled and will remain separate from ChatGPT's independent responses and from the user's original chat. In other words, the model's "organic" answer doesn't mix with paid content, at least that's what OpenAI promises in the current design.
For now, the test is running with select advertisers in the United States. No date has been announced for Brazil, but OpenAI's pattern in previous launches has usually been to expand markets in waves, which gives teams here some window to study the format.
The ChatGPT Ads Manager: Campaigns by Prompt
The piece that directly interests those who run media is the ChatGPT Ads Manager, a plugin that lets advertisers use natural language prompts to create, update, and analyze campaigns. In practice, the manager describes what they want in text and the tool builds the rest.
What OpenAI says it does:
- Turns a website or a campaign brief into a ready-made campaign;
- Suggests copy and images based on the landing page and the advertiser's goal;
- Runs performance review and returns recommendations;
- Offers an (optional) AI-driven copy customization that adapts headlines and descriptions to the context of the conversation and translates the text into the user's preferred language.
That last point is the most relevant for the Brazilian market. An ad that translates and adjusts itself automatically to the language and context of the conversation changes the logic of creative production: instead of running manual copy variations by segment, the advertiser delivers a base and the system personalizes it at the moment of delivery. For campaigns in Portuguese, that could mean less localization rework, but also less fine-grained control over the message that actually shows up.
Shopify and HubSpot Come In as First Partners
OpenAI has already locked in integrations that signal where it wants to go. HubSpot is the first CRM partner and Shopify, the first e-commerce partner. The Shopify app will launch internationally in markets where ChatGPT Ads is already available, starting September 23.
The choice isn't random. Linking the ad directly to a Shopify store's catalog and to HubSpot's CRM shortens the path between the conversation in ChatGPT and the sale or the lead recorded. For Brazilian merchants already running on Shopify, it's the kind of integration that tends to arrive with little technical effort once ChatGPT Ads opens up here, and it's worth watching which markets the expansion starts in.
Why This Matters for Media Work in Brazil
The bigger context is the fight over AI advertising money. Conversational assistants have been eating into traffic that used to go to traditional search, and the marketing industry is already discussing GEO (optimization for AI-generated answers) as a successor to or complement of SEO. With native ads inside ChatGPT, OpenAI starts competing head-on with Google and Meta for channel budget, now inside the very environment where part of the audience already asks purchase-related questions.
What changes in practice for growth professionals here:
- One more channel to master. If ChatGPT Ads follows the partners' script and arrives via Shopify, whoever already sells on that platform could be among the first to test it, with a learning edge over the competition.
- The ad unit is a conversation, not a click. Measuring the performance of a sponsored agent requires rethinking metrics: CTR and CPC aren't enough anymore, interaction quality and purchase intent in the middle of the dialogue come into play.
- Less control over the final creative. With copy customized and translated by AI in real time, what the user reads can vary. Brand teams in Brazil will need clear guidelines to avoid off-tone messaging or translations that distort the pitch.
- The line between organic and paid becomes a watch point. OpenAI promises labeling and isolation, but the boundary between a neutral answer and sponsored content inside an assistant is sensitive, and that's exactly where advertisers and consumers will test their trust in the platform.
What's Still Open
The source doesn't detail the billing model, the auction format, official reporting metrics, or the expansion timeline beyond the US and Shopify's markets. There's also no public data on how effective these ad units are, so any budget decision is still a bet. For the Brazilian market, the useful move right now is to track when ChatGPT Ads opens up here and to map whether the business already runs on Shopify or HubSpot, the two entry doors OpenAI chose to start with.
Translated from the Brazilian Portuguese original · Read the original
Perplexity swaps DynamoDB for in-house database and cuts latency by 5x
The company behind the AI-powered search engine migrated its serving layer to CobbleDB, an internal database written in Rust, and cut batch read latency by up to 5x while saving at least 20% on storage.