ChatGPT now sells ad space inside its own answers, and paid media teams are already testing budget on it.
Kondado is launching the ChatGPT Ads (OpenAI Ads) data source: your ad account's data replicated into your own warehouse, next to the channels you already track.
What ChatGPT Ads is
ChatGPT Ads is OpenAI's advertising platform, managed in the ChatGPT ad manager.
Like any brand-new media channel, it launches with no built-in way to compare its performance against the channels you already run: Google Ads, Meta Ads, TikTok Ads, Pinterest Ads.
That is exactly the gap the Kondado data source closes: it pulls data out of the ChatGPT Ads API and puts it side by side with the rest of your paid media stack, inside your own report or data warehouse.
What Kondado delivers
The data source replicates 8 tables from the ChatGPT Ads API:
| Table | Content |
|---|---|
| Ad account | Ad account metadata |
| Campaigns | Campaign setup, budget, bid type, configured conversion events |
| Ad groups | Ad group targeting, product feed, max bid |
| Ads | Ad creative: title, body, URLs, images, review status |
| Ad performance | Performance metrics with configurable grouping and breakdown |
| Ad preview | Visual ad preview, valid for 24 hours |
| Conversion event settings | Configured conversion events and their attribution window |
| Custom audiences | Custom audiences registered on the account |
On the performance table, you can configure:
- Grouping period: hour, day, or month.
- Entity: account, campaign, ad group, or ad.
- Breakdown: product, country, device, or platform.
Which view returns which metrics
This is the part that shapes your report design the most. The API does not return the same metric set across every combination:
| View | Metrics available |
|---|---|
| No breakdown, by day or month (account, campaign, ad group, or ad) | Media metrics + revenue, ROAS, CPA, and attributed sales (the conversions column only comes on daily grouping) |
| Country or device breakdown | Media metrics + conversions, on daily grouping only (no revenue, no ROAS) |
| Product or platform breakdown | Media metrics only |
| Hourly grouping | Media metrics only |
The six media metrics are: impressions, clicks, spend, CTR, CPC, and CPM.
In practice, revenue, ROAS, CPA, and attributed sales only exist in the aggregated view, with no breakdown at all, grouped by day or month.
If you need revenue tied to a catalog product, the API does not hand that to you. It comes from your own model, joining the aggregated view with the sales data you already hold in your warehouse.
Other limits worth knowing before you design a report on top of it:
- There is no age or gender breakdown.
- Revenue and ROAS only exist from the date the ChatGPT Ads API itself started reporting that field; history before that date simply has no value there.
- The API serves a maximum of 5 years of history.
- Conversions can be attributed up to 30 days after the click, which is why the data source's default update window (days) is 30: the last 30 days of each sync are reprocessed to capture delayed conversions.
- The ad preview expires 24 hours after it is generated, so that table needs the pipeline running at least once a day to stay valid.
How to build a report that doesn't contradict the API
- Pick granularity by objective, not by habit: hourly to check spend during the day; day or month for conversion, revenue, and ROAS analysis.
- Accept that a complete report is more than one pipeline. One aggregated, no-breakdown query for revenue, ROAS, CPA, and attributed sales; one country or device breakdown when you need segmented conversions; one product or platform breakdown when the question is about how spend is distributed.
- Reprocess the attribution window. Configure a recent-days update window on every recurring sync to catch late-attributed conversions.
- Don't mix campaigns from different accounts in one call: the API only sees the ad account that owns the key you used.
How to connect
- Sign in to the ChatGPT ad manager with an account that can generate API keys.
- Go to Settings and generate (or copy) your API key.
- In Kondado, create a new ChatGPT Ads data source and paste the key.
One thing to know upfront: each API key belongs to a single ad account. If you manage more than one ChatGPT Ads account, create a separate data source for each one, with that account's own key.
How to move the data into your warehouse or BI tool
Kondado replicates ChatGPT Ads data from an API key you generate in the ChatGPT ad manager.
Your warehouse receives the tables, and your BI tool (Power BI, Looker Studio) reads from that warehouse.
There is no manual export step, and conversions that land late are picked up automatically.
- Generate the API key in the ChatGPT ad manager, under Settings, with an account that can generate API keys.
- Create the ChatGPT Ads data source in Kondado and paste the key. Each key belongs to one ad account, so repeat this once per account you manage.
- Pick the destination: your data warehouse (BigQuery among the supported destinations), which is what Power BI and Looker Studio then connect to.
- Configure the performance table with the grouping period, entity, and breakdown your report needs. The breakdown you pick changes which metrics come back, so check the table above before you commit to a layout.
- Build the report in Looker Studio, Power BI, or your tool of choice, joining ChatGPT Ads against your other channels on campaign or date.
Besides the warehouse, visualization tools can also read the data through a direct connection, via Via Kondado.
Why a scheduled pipeline beats a manual export
A manual export from the ChatGPT ad manager can't capture late-attributed conversions (the API allows attribution for days after the click) or keep the ad preview valid (the preview expires hours after generation).
A recurring data source solves both: it reprocesses the attribution window on every sync and refreshes the preview daily.
What you can build with it
- Consolidated media report: once ChatGPT Ads data lands in your warehouse, join it with your existing Google Ads, Meta Ads, TikTok Ads, and Pinterest Ads pipelines and see spend, clicks, and conversions across your whole paid media stack in one place.
- Revenue and ROAS by account or campaign: use the aggregated, no-breakdown view grouped by day or month, the only one that carries revenue, ROAS, CPA, and attributed sales.
- Spend, clicks, and CTR by catalog product: use the product breakdown to see where budget is being consumed inside the feed.
- Conversions by country or device: those two breakdowns are the only ones that preserve the conversions metric, and only on daily grouping, useful for deciding where to concentrate budget.
- Hour-by-hour tracking: use hourly grouping to check spend and volume during the day, useful during campaign launches or bid tests.
What Kondado gives you beyond this data source
The ChatGPT Ads data source is one piece of a larger platform. Kondado connects +80 data sources and replicates them into reports, spreadsheets, databases, and AI assistants, with no code.
Everything is managed in the Kondado app: you create the pipeline, set how often it refreshes, and follow every run from one place.
So you don't start from scratch, the platform ships report templates: ready-made reports you pick, connect to your own sources, and customize to your metrics.
Your data also answers in ChatGPT and Claude.
With the Kondado MCP, you ask in natural language how many conversions ChatGPT Ads brought this month, or how each channel is performing, and the answer comes back in the chat.
The MCP reads the base Kondado replicates, and the query is read-only: it changes nothing in the connected systems.
The biggest win is the join.
With ChatGPT Ads side by side with Google Ads, Meta Ads, TikTok Ads, and Pinterest Ads, you compare spend, clicks, and conversions across your whole paid media stack, by campaign and by date.
A single report or warehouse, read through the visualization tool you already use: Power BI, Looker Studio, Tableau, Metabase, and others.
Frequently asked questions
What is the ChatGPT Ads API?
It's the data interface for ChatGPT's advertising platform (OpenAI), accessed with an API key generated in the ChatGPT ad manager, under Settings, returning both setup data (campaigns, ad groups, ads) and performance data.
Does the API return ROAS by product?
No. Revenue, ROAS, CPA, and attributed sales are only returned in the aggregated view, with no breakdown at all. With a product breakdown, the API returns only impressions, clicks, spend, CTR, CPC, and CPM.
Can I see conversions by product?
No. Conversions survive only the country and device breakdowns, and only on daily grouping. A product or platform breakdown returns the six media metrics only.
Is there an age or gender breakdown?
No. Available breakdowns are product, country, device, and platform.
Does one API key cover multiple ad accounts?
No. Each API key belongs to a single ad account; different accounts need separate keys and data sources.
Can I send ChatGPT Ads data to BigQuery, Looker Studio, or Power BI?
Yes. Kondado replicates the data source to your warehouse (BigQuery among the supported destinations), and Looker Studio, Power BI, or another BI tool reads from it.
How often should the pipeline run?
How often the pipeline runs is whichever interval you pick among the options your plan allows. To keep the ad preview valid and reprocess the conversion attribution window, we recommend the most frequent option available, and at least once a day.
Get started
The ChatGPT Ads data source is available on Kondado. Connect your account and bring your paid media data into the report or warehouse you already use.
See your whole paid media operation in one place: start with the 14-day free trial.
How to connect ChatGPT Ads to Kondado and report on it
Step by step to generate your ChatGPT Ads API key, create the data source in Kondado, and read the data in your warehouse or BI tool.
Generate the API key in ChatGPT Ads
Sign in to the ChatGPT ad manager (ads.openai.com) with an account that can generate API keys. Open Settings and generate (or copy) the key for the ad account you want to connect.
Create the ChatGPT Ads data source in Kondado
In the Kondado app, create a new ChatGPT Ads data source and paste the key. Each API key belongs to a single ad account, so create a separate data source for each account you manage, with that account's own key.
Pick the destination
Choose where Kondado replicates the tables: your data warehouse (BigQuery among the supported destinations) or a BI tool connected to it, such as Power BI or Looker Studio. Visualization tools can also read the data through a direct connection, via Via Kondado.
Configure the performance table
Pick the grouping period (hour, day, or month), the entity (account, campaign, ad group, or ad), and the breakdown (product, country, device, or platform). The breakdown changes which metrics come back: revenue, ROAS, CPA, and attributed sales exist only with no breakdown, and the conversions column only comes on daily grouping.
Set how often the pipeline runs
How often the pipeline runs is whichever interval you pick among the options your plan allows. We recommend the most frequent one available, and at least once a day, to keep the ad preview valid (it expires after 24 hours) and to reprocess the conversion attribution window.
Build the report
Build the report in Looker Studio, Power BI, or your tool of choice, joining ChatGPT Ads against your other paid media channels on campaign or date.
