Which ChatGPT Ads attribution window fits a seven-day versus ninety-day buying cycle?

A ninety-day buying cycle does not fit neatly inside ChatGPT Ads reporting.

That does not automatically make ChatGPT Ads a bad channel. It does mean the founder has to stop pretending that a 30-day click window is a 90-day revenue view.

The current Ads Manager documentation gives advertisers three click-through reporting windows: 7, 14, or 30 days. View-through reporting is either off or one day. OpenAI also says the setting changes reporting only. It does not change campaign optimization, bidding, or billing, and the system uses last-touch attribution rather than multi-touch attribution.

So the first decision is not “which window sounds normal?” It is “what part of the buying cycle am I actually asking this campaign to measure?”

If the buying cycle is seven days

Use the seven-day click window when the product normally moves from ad interaction to conversion inside a week and the event is close to the business outcome.

That could be a direct ecommerce order, a paid subscription, or a signup that becomes meaningful quickly. The important part is not the product label. It is the time between the click and the event you are willing to call a conversion.

I would start with click-through only for a short-cycle product unless there is a clear reason to include view-through. A one-day view-through conversion can enter the total when someone converts after seeing an eligible impression without a qualifying click. That may be useful for a low-consideration purchase. It also makes the number less clean if the question is “did the click cause the order?”

Use the separate reporting columns. Do not rely on one combined Conversions total when you are trying to understand the path.

If the buying cycle is ninety days

There is no native ninety-day click window in the current documentation.

The maximum is 30 days. That creates a hard boundary. A lead who clicks an ad on day one and becomes a customer on day 74 may still be valuable, but that sale is outside the native click-through reporting window.

The wrong response is to call the campaign a failure because Ads Manager did not claim the sale. The other wrong response is to quietly treat every later CRM conversion as a ChatGPT Ads conversion.

Use a nearer event instead:

qualified demo request;
activated workspace;
completed product setup;
sales-qualified opportunity; or
another event that happens inside the 30-day window and has a documented relationship to revenue.

Then report the later closed-won outcome separately in the CRM. Say what Ads Manager counted, what your owned system counted, and what the window could not see.

That is less satisfying than one blended ROI number. It is also more useful.

The event has to match the campaign

OpenAI’s conversion documentation says a conversion is reportable only when the platform receives an event, the event matches the configured campaign conversion, it falls inside the applicable window, and the platform can connect it to an eligible click using available signals.

That creates several easy failure modes.

A startup can send 'registration_completed' while the campaign is configured for a different event. A redirect can drop the 'oppref' click reference. A browser event and server event can describe the same conversion but use different event IDs, creating a duplicate or an unmatchable record. OpenAI also says modeled measurement may be included when available.

Before deciding that the attribution window is wrong, check the event and the connection:

What exact event is configured?
When does that event fire?
Does the click reference survive the redirect and checkout?
Are Pixel and Conversions API using the same event ID for the same conversion?
Are you looking at ad event time or conversion time?

The founder decision

For a seven-day product cycle, a seven-day click window is a reasonable starting fit if the conversion event is close to value. Add one-day view-through only when you can explain why an impression without a click belongs in the same decision.

For a ninety-day cycle, do not use Ads Manager’s 30-day number as full-funnel revenue. Use an earlier qualified event for campaign learning, connect the click reference and event data correctly, and keep the later CRM outcome in a separate layer.

The fit-based conclusion: short-cycle products can use ChatGPT Ads reporting as a direct response instrument. Long-cycle products should treat it as an early-funnel measurement layer unless their own CRM and cohort analysis can carry the missing time.