TTAB Deems CHATGPT Merely Descriptive of Chatbot Software, but USPTO Accepted Claim of Acquired Distinctiveness

The Board affirmed a Section 2(e)(1) mere descriptiveness refusal of the proposed mark CHATGPT for downloadable and non-downloadable “chatbot software for simulating conversations.” Applicant OpenAI argued that ordinary consumers would not know that GPT is an acronym for "generative pre-trained transformer," but the Board pointed out that "the evidence demonstrates that many users of Applicant’s goods and services actually are sophisticated users," and further that the term GPT appears in mainstream dictionaries and mainstream media. In re OpenAI OpCo, LLC, Serial No. 97733261 (March 4, 2026) [not precedential] (Opinion by Judge Jennifer L. Elgin).

OpenAI submitted two expert reports asserting that the term "GPT" is not generic, but the Board found them largely irrelevant since the only issue before the Board was mere descriptiveness, not genericness.

The Board first found that consumers "would recognize the separate component terms: CHAT, which is a recognizable word to consumers; and GPT, which . . . is an initialism." The Board then turned to a consideration of the individual terms.

[T]he evidence shows that “chatbots,” as used in the identification of Applicant’s goods and services, electronically converse or “chat” with humans in the ordinary sense of that word. There is no multi-stage reasoning needed; rather, CHAT immediately describes Applicant’s downloadable and non-downloadable software “for simulating conversations” or “chatting.”

Open AI contended that the term CHAT is merely suggestive because it has several meanings relevant in the context of Applicant’s goods and services, including “to talk in an informal or familiar manner” and “to take part in an online discussion in a chat room.” The Board found, however, that these definitions are equally descriptive of Open AI's downloadable or non-downloadable “computer programs and … computer software for the artificial production of human speech and text” and “for natural language processing, generation, understanding and analysis.”

Based on the identifications of goods and services, we must presume that in addition to being used by software developers, Applicant’s downloadable and non-downloadable software is available to the general public and casual users who do not take time to craft “effective prompts,” but rather are likely to interact with the software to facilitate everyday communication or “chatting.”

And so, the Board concluded that the term CHAT "is merely descriptive of a purpose or function of Applicant’s software, i.e., to assist or facilitate communication in a conversational manner, as reflected in Applicant’s identifications of goods and services."

OpenAI and the Examining Attorney agreed that GPT stands for "generative pre-trained transformer," and the evidence supported that conclusion.

Because “generative pre-trained transformer” refers to a machine learning algorithm that is pre-trained to produce human-like text in response to a human’s prompts, we find that it merely describes a feature or characteristic of Applicant’s downloadable and non-downloadable software, namely, that it facilitates “the artificial production of human speech and text, … natural language processing, generation, understanding and analysis,” and it simulates conversations.

Next, the Board considered whether relevant consumers will recognize the initialism GPT as the merely descriptive wording it represents, “generative pre-trained transformer.” The Board answered that question in the affirmative: first, many users of OpenAI’s goods and services actually are sophisticated users; second, the term appears in mainstream dictionaries, such as CAMBRIDGE DICTIONARY, Dictionary.com, and URBAN DICTIONARY, and in publications like THE NEW YORK TIMES, THE ECONOMIST, NEWSWEEK, FORBES MAGAZINE, USA TODAY, and BBC NEWS.

Finally, the Board considered the term CHATGPT in its entirety and found that CHATGPT "does not present an incongruity or lose its descriptive meaning, rather it immediately informs the consumer that Applicant’s goods and services feature a conversational chatbot, utilizing a generative, pre-trained transformer algorithm."

Having no doubt as to the descriptiveness of CHATBOT for OpenAI's goods and services, the Board affirmed the refusal.

Read comments and post your comment here.

TTABlogger comment: Why didn't OpenAI claim acquired distinctiveness? My bad. They did so, and the mark will proceed to publication. I should have read the last sentence: "The application will proceed with Applicant’s claim of acquired distinctiveness under Section 2(f)." So the bottom line is that CHATGPT is not inherently distinctive. Perhaps the Board should have said that inherent distinctiveness was the issue.

Text Copyright John L. Welch 2026.