What a shame!

Although I started writing this article six months ago, it’s now only a few days until the August 12 deadline for ICANN’s application submissions for new gTLDs. But better late than never, as the old saying goes. So here we go!

Fourteen years after the 2012 new gTLD program, ICANN opened the application window for the 2026 Round on April 30, 2026. Although ICANN’s application evaluation fee increased 22.7% from $185K per string in 2012 to $227K in 2026, indications are that this new window will get many more applications than the 1,930 applications filed in the 2012 round.

Technology, also, has evolved tremendously since the 2012 round of gTLD applications. Back then, Artificial Intelligence (AI) was not even on the radar, in contrast to the broad consumer and business product it is today. In three years to 2026 AI achieved an adoption rate of 53% (faster than the adoption rate of the PC or the internet), and AI adoption led to 26% and 50% gains in productivity in software development and marketing output, respectively.

AI can also help at almost every stage of preparing new gTLD applications, from the planning phase to evaluating responses to the application questions. Perhaps the first step in the application process where AI can help is preparing the business plan for a company applying for gTLDs. Although ICANN does not require applicants to submit a business plan, it provides data and information to help an applicant provide better responses to ICANN’s questions.

I settled with Anthropic’s Claude to prepare a business plan and financial model for a hypothetical Cayman Islands company (AIgTLDs) which will apply for two gTLDs. One gTLD is .BHANGRA for Bhangra music community, and the other is .LATTE, as in the coffee drink. As I asked requested, Claude provided the business plan and financial model as Microsoft Word, and Excel files, respectively. The financial model was comprehensive, and included projections of domain name sales, case flow, revenue, and other metrices under various scenarios, along with formulas.

The next step in the application process in which AI can help is the analysis of the Applicant Guidebook (AGB), which provides guidelines for applicants and the rules and regulations of the application. I asked Claude to analyze the AGB from the perspective of the number of questions, application types, and supporting documents (attachments) and financial information required. I got an informative 6-page document that helped me tame the 439-page AGB. My Claude chat revealed that the AGB had 225 application questions organized into 22 Question Sets.

Furthermore, 141 baseline questions apply to all applying entities, while the 84 questions are conditional upon the type of application (Community, Geographic Name, etc.). Claude also revealed that 54 (24%) of the 225 questions allow or require an attachment, with Question Set 7 (for Community gTLD applications) requiring the greatest number of attachments (17). Claude also provided a summary of the financial requirements for the four Profiles (Standard, Registry Operator, Top 25, and Government) of new gTLD applicants.

I then asked Claude to assign the drafting of responses to various to a human team, ChatGPT, and itself. I also asked Claude to indicate which of them will be the primary respondent, and which will be the secondary respondent/reviewer, and the rationale for the assignments. Claude provided me the information in a color-coded table in a Microsoft Word file.

The next step was to draft responses to the application questions, including drafting Registry policies. I thus asked Claude to prepare draft responses for the .BHANGRA and .LATTE applications. Given that my hypothetical. BHANGRA application is a Community Application, I asked Claude to prepare draft responses for 9 Question Sets (QS), starting QS 5 and 7 (Community gTLDs), QS 10-12, QS 14 (Financial Profile), and QS 18-20. On the other hand, the .LATTE is a generic TLD application, and as such, I asked Claude to draft responses to all the above 9 QSes for which draft responses to questions were provided, except for QS 7. The draft responses for .BHANGRA and .LATTE were provided in Word files which I could edit before submitting them.

The success or failure of a gTLD application also depends very strongly on the financial sustainability of the Registry that will operate it. I thus asked Claude to draft a financial profile template of my hypothetical company using data and information from the business plan and financial model it developed earlier. Claude faithfully provided me with an Excel workbook with, as required by ICANN, the Most Likely and Worst Case scenarios for its financial projections, as well as risk assessment framework and domain registration projections.

AI can also help evaluate the responses to the AGB questions before they are submitted on the TAMS. Obviously, it does not make sense to have an AI model evaluate the quality of responses it drafted. For this reason, I asked Gemini to evaluate the .BHANGRA application and the .LATTE application based on responses provided by against the requirements of the AGB give each a Pass/Fail grade. In the same vein, I asked Gemini to evaluate the financial profiles of both applications and provide each a Pass/Fail grade.

Gemini gave the .BHANGRA application and .LATTE application an overall grade of “Pass.” The .BHANGRA application passed because of its technical compliance, its strong definition of the community it will serve, and its commitment to safeguards. On the other hand, the .LATTE application “Pass” grade was based on its strong niche, technical compliance, and founder (myself) experience. However, the Pass grades were conditional upon the completion of other questions which required corporate and financial documentation.

The AI models provide a great start to completing the applications, but they are not a magic bullet. For this reason, the draft responses provided must be edited, and the information provide verified before the responses are submitted to ICANN. Despite this, the AI models can certainly considerably reduce the time, effort and cost of preparing responses to the AGB questions.

All this raises an important question: how can AI help in future rounds of gTLD applications? Given what already can be done with AI, applications almost all gTLD applications in the rounds after this 2026 round will be drafted by AI. Secondly, it is safe to assume that ICANN will leverage AI to evaluate applications in future rounds, meaning a drastic reduction in the cost of, and time taken to conduct application evaluations. This, in turn, should reduce the application and evaluation fees, the length of the application round and the duration between successive rounds, and lead to a Golden Age of inclusiveness in the domain names industry.

So it’s a shame this article is coming a bit too late. Nevertheless, the effort that went into it would have been worth it if only one applicant finds it useful, or if it starts an important discussion on the role of Artificial Intelligence (AI) in ICANN’s new gTLD program, and in the domain names industry at large.