Do ChatGPT ads work? What Q3 2026 data from DTC brands shows

Do ChatGPT ads work? For most DTC brands advertising in ChatGPT, the answer is yes, and by more than their dashboards show. WorkMagic’s first-party data from Q3 2026 shows that last-click attribution misses a large share of the orders ChatGPT ads drive, and a geo lift test with Dose found the ads lifting Amazon sales too. This post walks through the numbers and how to measure ChatGPT ads with incrementality instead of clicks.

Key takeaways

• Incrementality-adjusted attribution (iDDA) credits ChatGPT ads with 83% more orders and 74% more revenue than last click (per-brand median, Q3 2026).

• iDDA also credits 60% more new-customer orders than last click, so last click understates ChatGPT ads’ contribution to acquisition.

• In a geo lift test, Dose found 2.3x more incremental DTC orders than click attribution reported, with 67% from new customers.

• The same test measured a 34.5% halo effect on Dose’s Amazon orders, sales that click attribution never sees.

• To evaluate ChatGPT ads fairly, read them with incrementality: iDDA for day-to-day decisions and a lift test to confirm the true effect.

Do ChatGPT ads work? The short answer

Yes. Across WorkMagic customers running ChatGPT Ads in Q3 2026, the typical brand got 1.8x more orders and 1.7x more revenue from the channel than last-click attribution credited. A controlled geo lift test with Dose confirmed the gap: ChatGPT ads drove 2.3x more incremental DTC orders than click attribution reported, plus a lift on Amazon.

The catch is measurement. ChatGPT ads often do their work before the click, inside a conversation where a shopper is researching and comparing. Many of those shoppers buy later through another path, so a click-based report hands the credit to whatever touchpoint came last. If you judge the channel on last click alone, you will undervalue it.

ChatGPT ads performance: last click vs. incrementality

We compared two views of the same ChatGPT Ads activity across WorkMagic customers from July 1 to September 30, 2026: last-click attribution and iDDA, WorkMagic’s incrementality-adjusted data-driven attribution. The figures below are per-brand medians for brands with at least five last-click orders from ChatGPT Ads.

Three things stand out:

• The gap is large. Last click credits the typical brand with little more than half the orders iDDA attributes to ChatGPT ads. A channel that looks marginal on last click can be one of the stronger ones on an incremental basis.

• Revenue follows orders. The revenue gap (+74%) is close to the order gap (+83%), so the missing credit is not limited to small baskets.

• New customers are undercounted too. iDDA credits 60% more new-customer orders. For brands that buy ChatGPT ads for acquisition, last click understates the very thing they are paying for.

Last click has always struggled with channels that work earlier in the journey (see the challenges of marketing attribution). Platform documentation such as Google Ads’ guide to attribution models makes the same point: the model you choose changes which channels get the credit.

What a ChatGPT ads lift test found: the Dose case study

Attribution models estimate. A lift test measures. To check whether ChatGPT ads were truly incremental, Dose ran a geo lift test with WorkMagic. You can read the full Dose ChatGPT Ads case study; here is the summary.

The question

How much of ChatGPT Ads’ performance was truly incremental, under a strict customer acquisition cost (CAC) goal?

The test

A pause-to-measure geo holdout across matched U.S. DMAs: ChatGPT ads were paused in a set of test regions and kept running in comparable control regions. Results were measured on both DTC (Shopify) and Amazon orders to capture any cross-channel halo.

The results

“Incrementality helped us see the demand ChatGPT was actually creating,” said Ari Jessel, Head of Paid at Dose.

The Amazon result is the part click attribution can never show. A shopper who meets a brand in ChatGPT and then buys on Amazon leaves no click path back to the ad. Only a test that measures every sales channel captures that halo effect.

How to measure whether ChatGPT ads work for your brand

The most reliable way to know if ChatGPT ads work for you is incrementality: compare what happened with the ads to what would have happened without them. In practice that means five steps.

• Stop judging on last click alone. Use it as one signal, not the verdict on whether to keep spending.

• Use incrementality-adjusted attribution day to day. iDDA gives ChatGPT ads credit for the orders they cause, including ones that close through another touchpoint.

• Run a geo lift test to confirm. A pause-to-measure test like Dose’s, or a launch-to-measure test when you are starting the channel, gives a causal read. Our incrementality testing guide covers the setup.

• Measure every sales channel. Include Amazon, retail and other marketplaces, or you will miss halo effects like Dose’s 34.5%.

• Track new-customer share. Judge ChatGPT ads on incremental new customers and CAC, not only on blended ROAS. Our guide to calculating incremental lift shows the math.

Once you have a lift test result, use it to calibrate attribution so day-to-day numbers stay close to the measured truth. That is the core idea behind channel incrementality: each channel, ChatGPT ads included, gets credit for what it actually adds.

ChatGPT ads vs. organic ChatGPT traffic

Paid placements are only part of how ChatGPT sends shoppers to DTC sites. Organic ChatGPT traffic, the visits that come from links in ChatGPT’s answers, is growing and converting well. In the same Q3 2026 data, organic ChatGPT visitors converted at 5.3%, 68% of those orders came from new customers, and visitor volume grew 45% from Q2 to Q3.

Together the two datasets point the same way: shoppers who reach a brand through ChatGPT arrive ready to buy, and many are new to the brand. That makes accurate measurement of the paid side more important, not less.

What this means for your ChatGPT ads budget

If your ChatGPT ads look average on last click, do not cut them yet. Based on the Q3 2026 medians, the channel is likely driving more orders, revenue and new customers than your report shows, and possibly sales on other channels such as Amazon. Read the channel with incrementality-adjusted attribution, then confirm with a lift test before you scale up or pull back. For a broader view of marketing attribution and where it falls short, see our full guide, or the Wikipedia overview of attribution models.

Conclusion

Do ChatGPT ads work? The first-party data says yes, and that last click hides much of the effect. The way to know for your brand is incrementality: read the channel with iDDA and confirm it with a lift test that covers every sales channel. See how Dose did it in the Dose ChatGPT Ads case study, or learn how WorkMagic measures ChatGPT Ads.

Source: Aggregated, anonymized data across WorkMagic customers, Q3 2026 (July 1 to September 30). Per-brand median, iDDA vs. last click, for brands running ChatGPT Ads with 5+ last-click orders. Organic ChatGPT data covers April to September 2026. Dose results from WorkMagic’s case study, “Dose | ChatGPT Ads,” October 6, 2026.