On November 6th 2023 (Source), OpenAI, the research firm behind the globally recognized ChatGPT, released a public market place for consumers to create and access tailored versions of their LLM (large language model). According to OpenAI, anyone with ChatGPT plus subscription is able to build their own custom version of ChatGPT without the need to code (Source).
At Originality.AI, we have already gone into great depth covering both OpenAI as a company (Source) and their latest iteration of the ChatGPT model, GPT-4 (Source). However, for the first time, there now exists a semi-public (restricted to ChatGPT plus members) market for creators to access and alter the ChatGPT model by equipping it with specific instructions or knowledge aimed to improve performance on specialized tasks. Not only does the open sourcing of the ChatGPT model spell for momentous advancements in the technology used by AI applications, furthering the understanding and utility of them. It also boosts the widespread adoption of large language models in the daily lives of those not previously interested in programming or coding.
From having a pocket horticulturist to math tutoring, the many different applications of custom GPTs have been pointed out by OpenAI and showcased by independent creators since the public debut of the GPT store. What has not been made equally clear is the quantifiable impact of the GPT store itself. How many custom GPTs are currently available for use? Where are GPTs being used most commonly? How are these custom GPTs being used? Read on below to better understand the metrics and trends of the custom GPT market.
What Are GPTs?
Generative Pre-trained Transformers, commonly known as GPT, are a family of neural network models that uses the transformer architecture and is a key advancement in artificial intelligence (AI) powering generative AI applications such as ChatGPT. (Source)
GPTs are custom versions of ChatGPT created by OpenAI users. All you have to do is tell the GPT builder, in plain English, what you want to create, and the builder will take it from there. (Source)
GPT models give applications the ability to create human-like text and content (images, music, and more), and answer questions in a conversational manner. Organizations across industries are using GPT models and generative AI for Q&A bots, text summarization, content generation, and search. (Source)
How Can I Access the GPT Marketplace?
GPTs can be built and shared only among ChatGPT Plus members, who can share their GPTs by sharing a link to their GPT or by publishing their GPT in the OpenAI GPTs store (Source)
Enterprises with enterprise plans can also access custom GPTs. They may now empower users inside their companies to design internal-only GPTs without code and securely publish them to their workspaces. The admin console lets the enterprise choose how GPTs are shared and whether external GPTs may be used inside the business only. (Source)
** Disclaimer: The data below and any statistics listed hereon that are not explicitly followed by a source reference are part of an in-house analysis conducted by the Originality.ai team. Data visualizations were also produced in-house using data made available on ahrefs. Please be advised that data from ahrefs is updated frequently to represent best estimates rather than concrete figures. **
How Many GPTs are there?
As of mid-February, there are:
80,763 total unique GPT models
71,901 active unique GPT models
8,862 inactive unique GPT models
Below we can see a trendline of the total active unique GPT model count over time since the debut of the GPT store in early November 2023:
As we can see in the graph above, the GPT Store has seen a healthy growth in the amount of unique GPT models made available throughout its entire 3-month tenure. On average, the GPT store has added nearly 800 pages to its marketplace every day since its public debut.
Below we can see another graph that details the daily changes of model count including both additions to the GPT store and removals of models from the marketplace:
From the graph above we can see that for most of the GPT store’s tenure, the daily addition of models was in the range of 500-1000.
Remarkably, between December 29-30, 2023, the amount of additions and losses of GPT models to the GPT store evened out, resulting in no change to total count of models available in the marketplace.
During the period analyzed, only one day out of the 3 months had a negative change in model count, with a decrease of nearly 2000 pages happening on January 20, 2024.
As of February 3, 2024, there has been an upsurge in model additions to the GPT store, with an average of nearly 1500 models added to the marketplace daily.
Where are GPTs most popular?
As mentioned earlier, the total model count of the GPT store sits at around 80.8K (both active and inactive). We can track those models’ origin to see which countries they were developed in, leading to the following list of most common sites for custom GPT creation (by percentage, rounded to nearest hundredth) :
United States- 26.22%
India- 9.83%
Canada- 8.86%
United Kingdom- 7.41%
Australia- 7.11%
Germany- 5.97%
Russia- 5.49%
Spain- 5.37%
Netherlands- 5.12%
Brazil- 5.06%
France- 4.82%
Indonesia- 4.4%
Italy- 4.34%
It seems that a large proportion of custom GPTs originate in some sense from the United States. Although this may paint a picture of the supply side of GPT creation, we should also look at the demand side. Through this lens we can better understand which markets are currently most attractive to GPT creators. Looking at the graph below, we can see the main targets of GPT creation by deriving the top geographic keyword as it relates to total traffic (as percentage, rounded to nearest hundredth):
Japan- 31.02%
Korea- 25.13%
United States- 10.68%
Colombia- 8.80%
Brazil- 8.67%
India- 4.66%
France- 4.34%
Mexico- 2.99%
Philippines- 2.21%
Ukraine- 1.51%
From looking at the data above, we gather:
Between Japan, Korea, and India, more than half (60.80%) of custom GPT content is skewed toward appeasing the Asian market.
Although not significantly present in the GPT creation metrics, the Latin American market (Colombia, Brazil, and Mexico) represents about a fifth (20.46%) of custom GPT content targeting.
Lastly, the United States seems to be more of a producer than target market, accounting for 26.22% of GPTs but only 10.68% of targeted GPT content.
What are Custom GPTs being used for?
As noted above, the applications for which custom GPTs are used for can be nearly limitless. It seems that the limiting factor of GPTs capabilities will at one point be someone’s creativity and ingenuity. That being said all GPTs share the same backbone structure and will require the creator to determine the following settings:
Knowledge- If a creator requires a custom GPT to hold particular or specific reference information to guide desired response, creators can upload pdfs or other files to the custom GPT to provide it with the additional needed context as deemed necessary.
Capabilities- The default and minimum capability your custom GPT possesses is the ability to surf the web. If desired, the creator can also grant its custom GPT the ability to generate AI (artificial intelligence) images with DALL-E or interpret programming code using GPT-4.
Actions- Will your chatbot be able to read and access information from outside sources like an original pdf or word document? Should your chatbot be allowed to follow links provided or provide links itself to third party websites?
The following table below depicts the most popular type of GPT creations submitted by regular users. Top Keywords in GPT descriptions (excluding words like “Chat” or “GPT”) were used to understand what applications are most attractive to custom GPT consumers:
Looking at the graph above, we can highlight the following insights:
Of the top ten custom GPT descriptions, 40% are based on education applications (homeworkify, essay writer, grammar checker, ai math solver).
Only one of the top ten descriptions are directed towards the entertainment sector (cowordle).
Only three of the top ten descriptions are not text based, with two being image oriented (midjourney and uncrop) and one focusing on voice applications (ai voice generator).
What are the most popular GPTs?
The following chart displays the most popular custom GPTs on the GPT store, as dictated by the average volume of traffic as of mid February 2024:
A quick analysis of the visualization above yields the following insights:
The top most visited custom GPT “チャット GPT 日本語”, written in Katakana for the Japanese market, roughly translates to “(internet-based) Chat GPT in Japanese”.
The top two visited custom GPTs account for 110.8% (2,183,547) of the total traffic drawn in by the 38 other GPTs (1,971,243) in the top 40.
30% (12 out of 40) of the top 40 custom GPTs are focused on translation services or making ChatGPT’s capabilities available in other languages.
Is there any Market Value in Custom GPTs?
Much like the previous models and products offered by OpenAI, the custom GPT store presents a strong market potential. Although the direct revenue directly stemming from the custom GPT store has not been publicly reported we can use the following stats to have a better idea:
ChatGPT has become a relied-upon service for many people, as evidenced by the fact that just one year after its introduction, the service has been adopted by around 180.5 million users worldwide. (Source)
The market for LLMs is estimated to reach 40.8 billion USD by 2029, up from 10.5 billion USD in 2022. (Source)
The only way for regular consumers to access custom GPTs is through a ChatGPT Plus subscription, which currently costs $20USD per month. (Source)
The graph above denotes the organic traffic to custom GPTS and the estimated value of that traffic since the public debut of the GPT store since its launch in November 2023. Looking at this graph above, we can ascertain the following:
The custom GPT store draws in around 4-5 million visitors throughout the duration of its tenure.
The average value of the organic traffic drawn by the custom GPT store is around $1MUSD.
Traffic peaked at over 7M (7,358,026) on December 5th, 2023
Hypothetically, using peak traffic, if all site visits came from unique GPT plus subscribers, the revenue derived by the GPT store would be over 100 million USD (147,160,520). Although this is most likely not the case, it stands to illustrate how much OpenAI can stand to earn monthly through its custom GPT store.
What is the Potential Market Impact Stemming from Custom GPTs?
After looking at the potential market value that the custom GPT store poses internally for OpenAI’s profit and loss statements, it is equally, if not more, beneficial to look at the external market impact of custom GPTs. As mentioned earlier, custom GPTs can be created and used for a wide variety of tasks and applications, while this potential is enticing, it may also spell for consequential market effects when it comes to job and labor markets. OpenAI themselves recently conducted a study to estimate and explore some of these potential effects, concluding that:
Findings reveal that around 80% of the U.S. workforce could have at least 10% of their work tasks affected by the introduction of LLMs. (Source)
Similarly, approximately 19% of workers may see at least 50% of their tasks impacted. (Source)
Analysis suggests that, with access to an LLM, about 15% of all worker tasks in the US could be completed significantly faster at the same level of quality. (Source)
When incorporating software and tooling built on top of LLMs, this share increases to between 47 and 56% of all tasks. (Source)
The following table runs different models used to estimate the percentage of daily tasks for a specific profession that are exposed to being disrupted by the further implementation of LLM models like custom GPTs. The table lists the professions that were found to be at the greatest levels of exposure according to each individual model, in addition to the table it was found that:
Findings indicate that the importance of science and critical thinking skills are strongly negatively associated with exposure, suggesting that occupations requiring these skills are less likely to be impacted by current LLM. (Source)
Programming and writing skills show a strong positive association with exposure, implying that occupations involving these skills are more susceptible to being influenced by LLMs. (Source)
This table below examines the potential market impact and effects on US domestic labor through a lens of job specific training. Jobs are categorized into zones which denote the level of training required to competently fulfill expected job tasks, using this measure to determine potential risk of exposure to LLM disruption. Aside from the chart it was found that:
Upon examining barriers to entry, we observe that the jobs with the least exposure require the most training, potentially offering a lower payoff (in terms of median income) once competency is achieved. (Source)
Conversely, jobs with no on-the-job training required or only internship/residency required appear to yield higher income but are more exposed to LLMs. (Source)
Founder / CEO of Originality.AI I have been involved in the SEO and Content Marketing world for over a decade. My career started with a portfolio of content sites, recently I sold 2 content marketing agencies and I am the Co-Founder of MotionInvest.com, the leading place to buy and sell content websites. Through these experiences I understand what web publishers need when it comes to verifying content is original. I am not For or Against AI content, I think it has a place in everyones content strategy. However, I believe you as the publisher should be the one making the decision on when to use AI content. Our Originality checking tool has been built with serious web publishers in mind!
