How do I measure conversions and ROI from ChatGPT ads alongside my organic AI visibility?

When clients start testing ChatGPT ads, the first question we hear is some version of "so is it working?" The second, a week later, is "and how does that relate to what we get from ChatGPT for free?" Both deserve an answer, and the answer starts with refusing to blend the two into one number.

Here is how we set it up on client programs.

Keep paid and organic in separate buckets first

Paid ChatGPT ads have a spend, a click count and a conversion count, and the authoritative record of spend and clicks sits in the ad platform. We never ask a visibility tool to explain our own campaign costs. Promptwatch, the platform we run client programs on, says explicitly that its Ads Radar does not report on a client's own OpenAI campaigns. That is fine. We want the ad platform for spend and Promptwatch for everything around it.

Organic is the other bucket: how often the brand is named or cited in AI answers for the prompts that matter, and what traffic and conversions those answers send. It has no media cost, which is exactly why it gets undervalued when someone only looks at paid ROI.

The setup, in the order we do it

Agree what a conversion is

Before anything else, we write down one conversion definition per client and use it for both buckets. A demo request counts the same way whether it came from a paid click or an organic citation. Without that, any ROI comparison is arguing about definitions.

Tag the paid links

Every ad destination gets consistent UTM parameters that say paid and ChatGPT, used nowhere else. We do not rely on a default channel grouping to sort paid ChatGPT clicks from organic AI referrals. We test with a real click and look at what lands in the client's analytics before we trust the report. We have not seen a sourced description of how every analytics tool classifies these visits, so we verify per property.

Measure organic AI visits properly

This is where the visitor analytics inside Promptwatch earns its place. It tracks AI-referred traffic and conversions from ChatGPT, Perplexity, Gemini, Claude, Copilot and others through a lightweight script, a Google Tag Manager template or the WordPress plugin. The client does not have to maintain a pile of regex rules in GA4 to keep an "AI" segment alive.

Ads in AI answers are triggered by prompts, and so is organic visibility. We keep one prompt list per client and tag the prompts where the client advertises. Promptwatch lets us organize prompts by topic, tag, type and intent, so the question "where are we both paid and organic?" is a filter, not a spreadsheet exercise.

Watch competitor ads on those prompts

Ads Radar captures sponsored ads in ChatGPT Search, Google AI Overviews and Google AI Mode answers to tracked prompts, including advertiser, ad copy, landing page and the prompt that triggered it. It does not cover Perplexity. On brand plans it starts at Business; it is included on every agency plan. When a client's paid results shift, the first thing we check is whether a new advertiser appeared on the same prompt.

What the ROI number can and cannot say

For paid, ROI is revenue from tagged ad sessions minus spend, divided by spend. Nothing exotic.

For organic AI visibility, there is no media spend, so we use program cost (software plus content hours) in the same formula and tell the client up front that the result is a floor. Some AI answers influence a buyer without sending a click. Some visitors return later through branded search or direct visits and get credited elsewhere.

We also say three things plainly in every report:

Where paid and organic overlap on a prompt, the two results are not independent. Some paid clicks might have arrived organically anyway.
Tracking shows presence and traffic. It does not prove visibility caused a sale, and no tool can guarantee a brand gets recommended.
Early numbers are small. Show counts next to ratios so nobody celebrates a 300 percent change built on three conversions.

The monthly report we actually send

One page, three blocks.

Paid: spend, clicks, conversions and cost per conversion, pulled from the ad platform.

Organic: visibility score, share of voice, the pages being cited and AI-referred conversions from visitor analytics. For clients who want a deeper read, the 360 Insights view joins crawler logs, citation clicks and AI-referred visits per page, which shows which pages carry the organic result. Crawler logs start on the Professional plan.

Overlap: the prompts where the client appears in both ways, plus what competitors are advertising there.

We review it monthly. AI answers move faster than a quarterly deck can follow, and a paid campaign can change character the week a competitor's ad shows up.

When to push back on the question

Sometimes a client wants a single "AI ROI" figure for a board slide. We give them two figures and a sentence explaining why. A blended number hides the one thing leadership needs to decide, which is whether to move budget between paid and organic effort. If organic AI visits are converting well and paid is expensive, the right call may be to invest in the content and citation work that earns organic mentions. If organic is thin and the competitor ads are crowding the prompt, paid may deserve more.

Either way, the decision comes from separated data, and the separation comes from tagging links well and measuring organic visits with a tool built for it. That combination is why we run these reports out of Promptwatch rather than stitching them together by hand each month.