ChatGPT Ads, running now.

Advertising inside ChatGPT is live and we are running it for clients. The campaign is the easy half. The hard half is the product feed, which is why we built and published a free converter for it.

What this actually is

ChatGPT Ads places your products inside the assistant people are already using to decide what to buy. For ecommerce it works from a product feed, in the same way Google Shopping does, and the mechanics will feel familiar to anyone who has run Merchant Center.

The familiarity is also the trap. Of the 41 attributes in OpenAI's product feed specification, 30 map straight across from a Google Shopping feed. The other 11 have no Google equivalent at all, and five of those are required, which is why a Merchant Center export uploaded unchanged gets rejected. Three of them are eligibility flags carried per product, so you decide row by row what may appear in search, what may be bought inside ChatGPT, and what may run in ads. Google has nothing like it.

None of that is difficult once you know it. It is simply undocumented in the places people look, and it is where the time goes.

What's included

Feed build and validation

We convert your Shopify, Merchant Center or custom catalogue into the format Ads Manager accepts, and validate every row before it goes near an upload.

Shopping campaign setup

Campaign structure, budgets, bidding and geo targeting, configured against the eligibility flags already set in the feed.

Conversion tracking

The OpenAI pixel installed behind proper consent, with the events that matter wired to real outcomes rather than page views.

Feed maintenance

Manually uploaded items expire after 14 days. We move you to a hosted URL or SFTP so your catalogue does not quietly vanish mid-campaign.

Per market feeds

target_countries uses only its first entry, so one feed covers one country. We build and maintain one per market rather than discovering the gap later.

Product eligibility strategy

Deciding which products should be ads eligible rather than flagging the whole catalogue, so budget concentrates where the margin is.

Cross channel measurement

Read alongside your Google and Meta numbers, so this channel is judged on incremental contribution rather than in isolation.

Honest reporting

A new channel produces thin data before it produces good data. We say which is which instead of presenting noise as a trend.

How we run it

Audit the catalogue

We look at your existing feed first. Most problems are already in it: HTML left in descriptions, prices with no currency code, availability values in Google's wording rather than OpenAI's.

Build and validate the feed

Converted, validated row by row, and every rejection understood before upload rather than after. You see the count that will actually go live.

Set eligibility deliberately

Search, checkout and ads eligibility set per product against margin, not switched on wholesale because it was the default.

Launch and instrument

Campaign live with conversion tracking already in place, so the first week produces data you can act on instead of a gap you fill in later.

Maintain and read it honestly

Feeds refreshed, expiry handled, and performance reported against your other channels. Early numbers get called early numbers.

Who this is for

Ecommerce brands with a real catalogue and a working Google Shopping or Meta programme already. This channel is new, so the sensible budget is a test budget carved out of a working account, not a reallocation away from something that pays. If your existing channels have unresolved problems, fix those first. Being early somewhere unproven is a poor trade against fixing something proven.

Frequently asked questions

Can I actually run ChatGPT ads now?
Yes. ChatGPT Ads is live and we run it for clients. For ecommerce it works from a product feed, much like Google Shopping, and the campaign setup will feel familiar. The feed is the part that catches people out.
Can I reuse my Google Shopping feed?
Not unchanged. 30 of the 41 attributes map straight across, but 11 have no Google equivalent and five of those are required, so an unmodified Merchant Center export is rejected. Availability values change wording, prices need an ISO 4217 currency code, and descriptions must be plain text rather than HTML.
What are the eligibility flags?
Three per product attributes deciding whether an item may appear in ChatGPT search, be bought inside ChatGPT, and run in ads. Google has no equivalent: there, eligibility comes from feed configuration and campaign setup. Here it is a column, so you can make your whole catalogue searchable while only your best margin lines are ads eligible.
Can one feed cover several countries?
No. target_countries accepts a list but only the first entry is used, so one feed covers one market. Germany and Austria means two feeds and two uploads. We build and maintain them per market.
How much budget should we start with?
A test budget, carved out of a working account rather than taken from something that already pays. The channel is new enough that the first weeks buy information rather than volume, and anyone quoting you a confident return has invented it.
How do you measure it?
The OpenAI pixel, installed behind proper consent, with conversion events wired to real outcomes. We read it alongside your Google and Meta numbers so this channel is judged on what it adds, not on its own figures in isolation.
Do you also handle the feed if we run the ads ourselves?
Yes. We also publish the converter free for feeds up to 500 products, so you can do it yourself if you prefer. Larger catalogues, multi market setups and ongoing maintenance are where we usually get asked to take it over.

Want this handled properly?

Book a no-obligation intro call. We'll come with real observations, not a sales script.