top of page

“Red Soles” and Search Engines: Louboutin Trademark Lawsuit Proves ChatGPT is Unreliable Evidence of Secondary Meaning

By: Gianna Policastro

"Louboutin altadama140" by Arroser is licensed under CC BY-SA 3.0.

Across the globe, high fashion fans recognize “red-soled”[1] shoes as “more than a [mere] color. It’s an attitude.”[2] Christian Louboutin “turned his trademark [red] shade into a[n] internationally recognized symbol of luxury[,]  … passion, power, sensuality, love, vitality, and a certain stylish insouciance à la française.”[3] Christian Louboutin created his eponymous brand in 1993, and the French designer has filed trademark applications for its “RED SOLE” mark in numerous countries, including “France, Mexico, Singapore, New Zealand, Australia, the [United States], Moldova, the Philippines, Cambodia, Laos, Brunei, Indonesia, Morocco, Bahrain, Chile, Israel, Switzerland, Vietnam, Malaysia, and India.”[4] Given Louboutin’s global presence, it is no surprise that the brand “has filed numerous [trademark infringement actions] in several jurisdictions,”[5] often litigating its trade dress[6] rights over the “RED SOLE” mark.[7]


Recently, on August 22, 2023, the Delhi High Court decided Christian Louboutin SAS v. M/S the Shoe Boutique – Shutiq, where Louboutin alleged a small, New Delhi-based shoe brand infringed its RED SOLE and SPIKED SHOE STYLE marks.[8] Louboutin successfully enjoined Shutiq from infringing these marks.[9] However, Shutiq is noteworthy because the Delhi High court set global precedent, cautioning trademark owners worldwide against relying on AI-generated evidence as proof of secondary meaning.


In Shutiq, as one piece of evidence supporting Louboutin’s “extensive advertising and long and continuous use of” its RED SOLE mark, Louboutin presented a response generated by ChatGPT,[10] a generative AI tool that provides answers to prompts in “dialogue format.”[11] The ChatGPT response stated: “Louboutin is known for their iconic red-soled shoes, including spiked styles for men and women.”[12] Louboutin intended for this AI-generated response to further support its claim that the RED SOLE mark has secondary meaning;[13] specifically, that consumers identify the RED SOLE as a source-identifier of the Louboutin brand.


However, the Shutiq court announced a bright-line rule that AI-generated evidence cannot establish a mark’s secondary meaning.[14] As a result, Shutiq makes clear that, while person-made evidence is subject to ethical disparities, biases, and outright human error, these potential shortcomings are preferential to AI’s algorithmic flaws.[15] Moreover, the Shutiq court encouraged other judicial systems to adopt similar guidelines in their own trademark enforcement practices.[16]

In today’s digital age, Shutiq’s decision begs the question: why can AI not be a reliable source of acquired distinctiveness, and why might United States’ judiciaries be inclined to adopt a similar rule?

According to Shutiq, there are several reasons why AI provides unreliable evidence of secondary meaning, and thus cannot replace human-made evidence. For example, responses created by “Large Language Model based chatbots such as ChatGPT … [depend] on a host of factors including the nature and structure of query put by the user [and] the training data.”[17] These factors increase the likelihood that generative AI tools could produce “incorrect responses, fictional case [law, or] imaginative data.”[18] Further, at present, the “accuracy and reliability of AI generated data is still in the grey area.”[19]

Nevertheless, Shutiq leaves open the possibility for generative AI to play a useful role in other aspects of trademark practice. For instance, the Shutiq court noted that lawyers may use platforms like ChatGPT “for preliminary research.”[20] Contrastingly, the Delhi High Court emphasized that “human intelligence,” as well as the “human element in a judicial process” are indispensable elements of common law systems.[21]

Although generative AI may facilitate efficient legal drafting, restricting lawyers’ use of these tools is permissible in trademark practice. Doing so would protect mark owners that are smaller businesses throughout the trademark clearance, prosecution, and portfolio management procedures. In addition, prohibiting trademark lawyers from relying on AI sites will prevent their small-business clients from inadvertent infringement liability that may arise from AI’s algorithms.[22] Without this rule, those small businesses could face detrimental impacts of costly litigation, especially where the parties possess unequal bargaining power. Finally, domestic trademark jurisprudence will clarify the merits of fair use defenses where ChatGPT may arguably use a valid trademark for descriptive or educational purposes.

In Shutiq, as in most trademark infringement disputes, a likelihood of bias likely refers to a more powerful litigant’s competitive advantage, both in the marketplace and in the courtroom.[23] In this case, Louboutin exemplifies the power imbalance when global powerhouse brands assert trademark rights against smaller, less expansive brands that lack the financial resources to effectuate a perfect defense. But, as Louboutin experienced itself in the United States and India, young brands purposefully lack the ability to establish secondary meaning. The Shutiq court’s reasoning prevents bad faith trademark registration by prohibiting brands from relying on generative AI tools as evidence of a mark’s sufficient use in commerce and its reputation amongst the public.[24] Given the United States’ commitment to maintaining use-based trademark rights, Shutiq would advance domestic trademark policy and protect mark owners’ goodwill.[25]

As jurisdictions like the United States grapple with AI’s place in judicial processes, courts may seek guidance from Shutiq and consider adopting similar rules. At the very least, restricting AI’s role in United States trademark infringement litigation would uphold trademark policies of fostering commercial competition, protecting bona fide use of a mark in commerce, and rewarding brands that maintain high-quality goods and services.


Gianna Policastro is a Staff Editor at CICLR.

[1] Christian Louboutin, [].

[2] Id.

[3] Id.

[4] Irene Calboli, Hands Off “My” Colors, Patterns, and Shapes! How Non-Traditional Trademarks Promote Standardization and May Negatively Impact Creativity and Innovation, in The Protection of Non-Traditional Trademarks: Critical Perspectives 287-308 (Oxford Academic, 2018).

[5] Id.

[6] Trade dress is defined as trademark “protection of ornamental and visual representation of goods, either product packaging or [product] design.” Barbara Kolsun & Douglas Hand, The Business and Law of Fashion and Retail 277 (2020). Such ornamental features “can be a valid trademark” upon a showing of secondary meaning. Id.

[7] In the United States, Mr. Louboutin has maintained USPTO trademark registration of the RED SOLE trade dress since January 29, 2008, for goods including leatherwear, footwear, headwear, gloves, belts, and scarves. See The color(s) red is/are claimed as a feature of the mark. The color red appears in the design representing a stylized red sole, Registration No. 3,376,197. In India, Louboutin received trademark protection over the RED SOLE mark in December 2017 following the Delhi High Court’s determination that the mark was sufficiently “well-known.” See Christian Louboutin SAS v. Mr. Pawan Kumar, CS(COMM) 714/2016 (India).

[8] Christian Louboutin SAS v. M/S the Shoe Boutique – Shutiq, CS(COMM) 583/2023 3-5 (India).

[9] Id. at 17-18.

[10] Indian Court Says No to ChatGPT Use in Louboutin Trademark Lawsuit, The Fashion L. (Aug. 28, 2023), [].

[11] Introducing ChatGPT, ChatGPT (Nov. 30, 2022), [].

[12] Shutiq, CS(COMM) 583/2023 at 13.

[13] Id.

[14] Id.

[15] Id. at 17-18.

[16] IP Protocol Puts India on the Map, The Lawyer, [].

[17] Id. at 16.

[18] Id.

[19] Id.

[20] Shutiq, CS(COMM) 583/2023 at 17.

[21] Rohit Kulkarni, Arguments by ChatGPT Are Invalid in Court; Cases Can’t Be Judged Based on ChatGPT – Delhi High Court, (Sept. 2, 2023), [].

[22] Some ChatGPT users question whether AI tools impermissibly apply others’ intellectual property to respond to users’ prompts. For example, generative AI “is capable of writing ‘in the style of’ many human authors if instructed to do so.” Bernard Marr, Is Generative AI Stealing from Artists?, Forbes (August 8, 2023, 02:36 AM), [].

[23] Shutiq, CS(COMM) 583/2023 at 16-17.

[24] Id.

[25] Id.

1 Comment


bottom of page