Brief IA
›

ChatGPT Advertising: Returns Still Hard to Compare

AI Tools·Tom Levy·

ChatGPT Advertising: Returns Still Hard to Compare

ChatGPT Advertising: Returns Still Hard to Compare
Key Takeaways
1Out of 102 studied sites, researchers measure $0.89 in revenue per ChatGPT ad session in the United States compared to $2.45 in paid search.
2The returns reported per dollar spent combine these measures with click costs from other accounts.
3Bank of America's survey focuses on the budget intentions of 117 advertisers and agencies, not on actual transfers made.
Why it matters : these data describe observed performances but do not demonstrate the effect of a budget transfer for a given advertiser.

The first cases from advertisers show mixed results against Google, often failing to track leads all the way to sales. A comparison of 102 sites reveals lower revenue per visit with ChatGPT, but the returns reported per dollar spent are based on advertising costs from different samples.

Agencies Are Not Yet Measuring All Sales

At StubGroup, which markets advertising management on ChatGPT, a 30-day test for a B2B compliance company yielded a cost per lead of $92.21 with ChatGPT, compared to $157.57 with Google. The agency reports clicks at $0.76 versus $10.46, but conversion rates of 0.8% compared to 6.64%. It notes that this ChatGPT cost per click is low compared to some of its other clients. Google campaigns, optimized over the years, appear to provide leads that convert to customers at a higher rate, with a seemingly lower cost per customer. The long sales cycle does not yet allow for knowing the final number of customers derived from leads on each platform, according to StubGroup.

Workshop Digital invested $2,319 of its own funds on ChatGPT, without obtaining a qualified lead, compared to two with Google. The agency reports costs per click of $3.14 on ChatGPT and $6.76 on Google Search, along with click-through rates of 1.1% and 5.2%. It indicates that the campaigns are not directly comparable, without providing the corresponding dates or budget for Google.

Brafton, which sells ChatGPT advertising services, claims to have spent $1,524.11 for 263 clicks and a form submission with an ad running from June 17 to August 28. Its comparisons with Google and LinkedIn do not match benchmarks, and its paid media director, Zach Shah, believes that ChatGPT is priced like search but behaves like paid social media. The three agencies focus on results after the click, where the outcomes are the fewest. In September, Brooke Osmundson reported the lack of context to interpret these results.

Among the client testimonials highlighted by OpenAI is Jotform, which claims a cost per sign-up about 30% lower than that of paid search. This result is based on internal data from July, without details on costs or the method of attributing sign-ups. A sign-up does not equate to a sale.

Reported Returns Depend on Prices from Other Accounts

The Frankfurt School reports $0.52 in revenue for every advertising dollar spent on ChatGPT, compared to $3.40 on Google Search, yielding a ratio of 6.6 according to the institution. Unlike revenue per visit, these returns are not directly measured: Kaiser and Schulze divide revenue per session by cost per click, equating each session to a billed click. The amounts reported by the institution are not included in the body of the study, and its communication does not detail the calculation. Therefore, the result depends on the actual prices paid on each channel.

The calculation retains $0.72, the median from 16 primarily U.S. accounts that connected their Google Ads spending, without matching this cost with the revenue of each site. For ChatGPT, the $1.72 comes from a single account presented in June by Opascope, a paid media agency. It mentions about $60,000 spent in 15 days on campaigns developed after promising starts, and not an average of all its tests; it reports a ROAS of 1.49x. In the researchers' sample, only one site achieves conversions and implicit revenue per visit comparable to Opascope, with just nine ChatGPT advertising sessions.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

From these references, Kaiser and Schulze describe a ChatGPT click as about 2.4 times more expensive and a cost-effectiveness ratio less than one-sixth that of Google Search. Their table places the ChatGPT price at 6.57 times the level needed to match search returns. However, they consider $1.72 a conservative estimate, at the lower end of the prices cited by practitioners and below the initial maximum bid of $3 to $5 suggested by OpenAI. This bid remains a ceiling chosen by the advertiser, not necessarily the amount billed.

A resampling of advertisers results in a ChatGPT price of about $0.26 to match search returns, with a range of $0.05 to $0.57. This procedure keeps the selected click prices constant for both platforms.

The 102 Sites Show Lower Revenue per Visit with ChatGPT

On the U.S. traffic of the 102 sites studied, Kaiser and Schulze measure $0.89 in revenue per ChatGPT advertising session, compared to $2.45 in paid search, or about one-third. The transaction rates per session are 0.37% and 1.17%, respectively. The entire dataset covers 60 weeks up to August 22, with up to 28 weeks available for U.S. ChatGPT sessions.

Their model, which accounts for the site, device used, and week, finds lower conversions and revenue per session with ChatGPT. According to the researchers, this finding holds even when removing individual advertisers: among 87 advertisers using both channels in the U.S., 84 report lower revenue per session on ChatGPT. About 20% of the sites belong to services, software, education, or finance, where reported conversions may include bookings, leads, licenses, or sign-ups. The appendix describes a transaction as a completed order, without specifying how these categories relate.

Advertisers who chose to buy these ads themselves lead the researchers to claim a descriptive comparison, not a causal demonstration of ChatGPT's inferior effectiveness. Their results do not allow for determining whether reallocating a budget between matched campaigns would benefit or harm a particular advertiser. OpenAI responds that the data pertains to an earlier phase of development and that extrapolation is based on a small sample. The company indicates it has continued its launches, including Target CPA and Target CPC.

The data comes from Grips Intelligence, where Kaiser is a former employee and scientific advisor. The researchers state that he received a salary as well as virtual shares from this company. They clarify that Grips Intelligence did not participate in conducting the research.

Revenues Do Not Quantify Any Transfer from Google

According to a survey conducted by Bank of America among 117 advertisers and agencies, the results of which were reported on October 14, Google Search is identified as the most frequently cited potential source of funding for future ChatGPT campaigns. However, it gathers intentions without measuring the amounts already shifted. Neither these responses nor the revenues published by the platforms allow for quantifying a transfer from Google to OpenAI.

OpenAI introduced its ads in February and then launched targeted cost auctions last week. The company reported as of August 31 an annualized revenue run rate of $1 billion; on October 21, it indicated it had surpassed this level and counted over 100,000 advertisers since launch. Alphabet simultaneously announced $63.35 billion for its search and other category in the third quarter, up 12%. This growth does not preclude that some advertisers may have reallocated part of their spending elsewhere.

During the earnings presentation, Philipp Schindler, Alphabet's chief business officer, cited the lower monetization of the AI mode among the factors slowing search growth. He did not accompany this explanation with any figures or references.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.