Blog Media

OpenAI ads: strong early returns for eight brands, at small spend

Seven of the eight brands we measured earned more per OpenAI dollar than on their average media dollar, and the eighth broke even. Spend is still tiny, so how far that goes is the open question.

By Senior Principal, Data Science 4 min read

LinkedInXEmail
An executive studies charts drawn on a glass wall

Brands are starting to test ads inside OpenAI's products, and the first question is always the same: is it any good? We put each brand's OpenAI results into Alli Scenario Planner next to every other channel it buys, and compared them on the brand's own numbers.

How we did this

  • Brands: eight, one in each of travel, media and publishing, footwear, home appliances, value apparel, sporting goods, luxury apparel, and designer fashion. Client names are withheld.
  • Scope: 24 platforms across the eight brands' media mixes, 162 channels in all.
  • Performance index: OpenAI's average return per dollar against the brand's blended media, where 100 is parity. It's an average, not a marginal, measure.
  • Limits: small spend, short history, and platforms that measure results differently. That's why we compare each channel with the brand's own average and don't quote raw figures.

The result: ahead of the brand average for seven of eight

OpenAI beat the brand's blended media for seven of the eight brands and matched it for the eighth. Across all eight, its performance index was 167, so each OpenAI dollar earned about 67% more than the average media dollar.

The spread is wide. Value apparel led at 539, then footwear, travel, and home appliances between about 270 and 370. Sporting goods came in at 154 and luxury apparel just ahead of parity at 107. Designer fashion matched its blended media at 100. Media and publishing reads 964, but its blended baseline is mostly awareness spend on out-of-home, print, and linear TV, so it isn't directly comparable with the others.

Figure 1

OpenAI's performance index, by vertical

OpenAI's average return per dollar against each brand's blended media (100 = parity)

OpenAI performance index by vertical: media and publishing 964 (not comparable), value apparel 539, footwear 368, travel 329, home appliances 272, sporting goods 154, luxury apparel 107, designer fashion 100; portfolio 167

Note: Client names withheld; each brand is shown by its vertical. The media and publishing bar is cut off at 600. Platforms measure results differently, so each channel is compared with the brand's own blended media rather than quoted as raw figures.

It also ranked in each brand's top four platforms, and first for three. That's against Google, Meta, YouTube, Amazon, Bing, TikTok, Pinterest, connected TV, and more.

Figure 2

Top four for every brand, first for three

OpenAI's rank among each brand's platforms

All eight brands ranked OpenAI in their top four platforms: first for three brands, second to fourth for five

Why this looks better than what you've read

The public picture has been mixed. Trade reports this year have cited click-through rates near 0.91% against a 6.4% Google Search benchmark, and some advertisers couldn't see their own campaign data (The Keyword).

Both things can be true. Click-through rate measures clicks, not what those clicks are worth. And a small budget buys the best opportunities first.

There's outside evidence that the audience is valuable. In September 2026, visits to US retail sites from AI services earned 83% more revenue per visit than other traffic and converted 90% better, according to Adobe. Twelve months earlier, an AI visit was worth less than half of a typical one. Those are referrals, not ads, but they show what the people on the other side of an AI answer are worth.

83%

Higher revenue per visit from AI-referred traffic than other visits, US retail

Adobe, September 2026

90%

Better conversion for AI-referred visits than other visits

Adobe, September 2026

71%

Year-over-year growth in AI-referred traffic to US retail sites

Adobe, September 2026

Average return flatters a small channel

This is the caveat that matters most. We're comparing OpenAI's average return at very low spend with a brand average that includes mature channels at full scale. Any channel looks its best at the start of its curve.

The fairer test is marginal return: what the next dollar earns in OpenAI against the next dollar anywhere else.

Scenario Planner's response curve shows the shape. Pooled across the eight brands, total response keeps rising as OpenAI spend grows, but the average return per dollar falls fast, to about 103 at a typical mid-size budget, roughly level with blended media.

Figure 3

More spend keeps paying off

Modeled response to OpenAI spend, with its range, indexed to 100 at a typical mid-size budget

Line chart: modeled response rises from about 24 at the lowest spend to 100 at a typical mid-size budget, with a range around it that narrows as spend grows

Note: Pooled across the eight brands. Spend runs across roughly the 5th to 60th percentile of the brands' active channel spend (162 channels). Response is modeled as spend times the average index, so it is an average, not a marginal, measure.

Figure 4

Each dollar pays off less

OpenAI's average return per dollar at each spend level (100 = blended media)

Line chart: the efficiency index falls from about 607 at the lowest spend modeled to about 103 at a typical mid-size budget, near parity with blended media

Note: The same spend range as the chart above. This is the average return per dollar at each level, which falls even as total response keeps growing. The two charts measure different things.

With total budgets held flat, the model found OpenAI could take meaningfully more money than it gets today, and several brands could absorb multiples of their current spend. But the model has only seen the early part of the curve, so its estimate of where returns flatten comes with a wide range.

Where Alli fits

Scenario Planner is where this comparison was run. It blends a media response model built on historical sales with a brand equity model, and for every channel it shows minimum and maximum spend with predicted revenue and ROAS. That's what lets a new channel be judged on the same footing as an old one.

Testing OpenAI ads? See where they rank in your own mix.

Frequently asked questions

Are OpenAI ads effective?

Early results are strong. Across eight brands PMG analyzed in Alli Scenario Planner, OpenAI beat blended media for seven and matched it for the eighth, with a performance index of 167 overall. Spend is still small, so those returns will fall as spend grows.

Why do OpenAI ads show low click-through rates but strong returns?

Click-through rate measures clicks, not what those clicks are worth. Trade reports cited a 0.91% click-through rate against 6.4% on Google Search, but a small budget buys the best opportunities first, and AI-referred shoppers tend to be worth more per visit.

Why is average return misleading for a new channel?

A channel looks its best at the start of its curve. Comparing its average return at low spend with a brand average that includes mature channels flatters it. Marginal return, what the next dollar earns, is the fairer test.

How much should brands spend on OpenAI ads?

The model found OpenAI could take meaningfully more than it gets today, but it has only seen the early part of the curve. Raise spend in steps, keep everything else steady, and re-measure each round.

BlogWhy your MMM's range matters more than its numberA mix model that says TV returns 2.6 is also saying it might return 1.5 or 3.8. Budget decisions made on the single number ignore most of what the model knows.Read BlogThink Bayesian without the equationStart with what you already know, ask how well new evidence fits each explanation, and adjust your confidence. One brand search question shows the whole habit.Read BlogSynthetic research: what to trust, and when to add real peopleAI estimates landed about four points from real survey results on average. That's close enough for a first look, and not close enough for every decision.Read BlogWhy finance doesn't trust marketing's numbersFinance and marketing expect similar payback. What finance wants is a number it can trace, define, and put a range around.Read 

Sources

  1. Adobe, AI Traffic Drives 83% Higher Revenue Per Visit, Vivek Pandya, October 29, 2026. Based on more than 1 trillion visits to US retail sites, 100 million SKUs, and 18 product categories. Adobe sells analytics and marketing software.
  2. The Keyword, Coverage of ChatGPT ad pilot performance, citing Adthena data, March 24, 2026.
  3. The Keyword, Coverage of the ChatGPT ad pilot's revenue and reporting problems, March 27, 2026.

Last updated November 12, 2026. We'll update this article as spend grows.

The Alli by PMG blog

All blog articles