Which custom GPTs get the most users now?

Can we actually tell which custom GPTs have the most users?

We can identify the biggest custom GPTs today, but we cannot rank them precisely by unique users because OpenAI does not publish comparable monthly-active-user numbers for individual GPTs.

The public number we can track is conversations. GPTStore.ai, which indexes public GPT listings, currently shows rounded totals such as 65M+, 49M+ or 18M+. Those figures tell us how much usage a GPT has accumulated, but one person can generate many conversations.

That distinction changes how we should read the leaderboard. A GPT used every day by a relatively sticky audience may accumulate more conversations than another GPT tried once by a much larger number of people.

Ratings help a little, although they have the same problem. Write For Me currently has roughly 147,000 ratings, while Consensus has around 66,000. That tells us both products reached large audiences, but there is no reliable formula for turning ratings into unique users.

So throughout this analysis, “biggest” means the GPTs with the largest observable public usage. When we discuss current momentum, we look separately at how those counters have moved over time.

Which custom GPTs are the biggest right now?

The biggest custom GPTs right now are surprisingly mainstream: local-language ChatGPT alternatives, a website-and-app builder, writing, academic research, design and video creation dominate the top of the public leaderboard.

GPTStore.ai currently puts the Russian-language чат from gptonline.ai at roughly 65M+ conversations. Best Website & App Builder Agent follows at 63M+, Write For Me at 49M+, the Korean-language 챗 at 46M+, Scholar GPT at 42M+ and the Japanese-language チャット at 37M+.

Canva is around 28M+, Chat Português around 22M+, Video AI by invideo around 19M+ and Consensus around 18M+.

Together, those ten GPTs have accumulated at least 389 million displayed conversations. The top five alone account for about 265 million, or 68% of that top-ten total.

The list is revealing because there are few exotic AI-agent ideas near the top. People overwhelmingly used custom GPTs for jobs they already understood: chatting in their own language, writing, researching, designing, building something or making a video.

Custom GPT	Public conversations	Main use case
чат	65M+	Russian-language general chat
Best Website & App Builder Agent	63M+	Website and app creation
Write For Me	49M+	Writing
챗	46M+	Korean-language general chat
Scholar GPT	42M+	Academic research
チャット	37M+	Japanese-language general chat
Canva	28M+	Design
Chat Português	22M+	Portuguese-language general chat
Video AI by invideo	19M+	Video generation
Consensus	18M+	Scientific research

Are the biggest custom GPTs still growing fastest today?

The largest custom GPTs are still adding usage, but their lifetime rankings make current growth look stronger than it really is.

Recent GPTStore.ai snapshots illustrate the difference. The Russian чат moved from roughly 64M+ conversations to 65M+. Write For Me went from about 48M+ to 49M+. Canva moved from 27M+ to 28M+. Consensus rose from 17M+ to 18M+. Video AI by invideo stands out slightly more, moving from roughly 17M+ to 19M+.

Millions of extra conversations are meaningful, but these are modest percentage gains for products that already have tens of millions.

The longer history looks much more dramatic. Public GPT Store snapshots from 2024 placed Write For Me around 5M conversations; today it is around 49M. Canva went from roughly 3M to 28M. Scholar GPT rose from about 2M to 42M. Consensus went from around 5M to 18M.

Scholar GPT therefore grew by roughly 21 times from that early snapshot, Write For Me by almost ten times and Canva by more than nine times. Consensus grew strongly too, although its multiple was closer to 3.6 times.

The current leaderboard is partly a record of who won the first few years of the GPT Store. It is much less useful as a leaderboard of who is winning new users this week.

GPT	Earlier public count	Current public count	Approx. multiple
Scholar GPT	2M+	42M+	~21×
Write For Me	5M+	49M+	~9.8×
Canva	3M+	28M+	~9.3×
Consensus	5M+	18M+	~3.6×

Why are language-specific ChatGPT GPTs so huge?

Language-specific ChatGPT GPTs are one of the biggest custom-GPT businesses hiding in plain sight, with several individual products reaching tens of millions of conversations.

The clearest example is gptonline.ai. Its Russian-language GPT currently shows around 65M+ conversations, the Korean version around 46M+, Chat Português around 22M+ and Chat Español around 14M+. The same creator also operates Turkish, Dutch, Romanian, Croatian, Danish and German GPTs.

A separate Japanese GPT from gptjp.net has around 37M+ conversations.

Just the Russian, Korean, Japanese and Portuguese products contribute about 170 million conversations. Among the current top ten GPTs, that represents roughly 44% of all displayed conversations.

That scale changes the way we should think about custom GPT adoption. Some of the biggest winners did not invent a new AI capability. They gave users an obvious local-language entry point with culturally familiar positioning and a name people could immediately understand.

The GPT Store ended up functioning partly like search. A simple query such as “ChatGPT Korean” or “ChatGPT Portuguese” could lead users into a specialized wrapper that then became their habitual way to use the underlying model.

Is writing still one of the biggest custom GPT use cases?

Writing is still one of the largest custom-GPT use cases today, and Write For Me remains in a league of its own at roughly 49M+ conversations.

The gap inside the category is huge. GPTStore.ai currently lists Unbound Limitless Story Writer around 3M+, Writing Assistant around 2M+ and Cover Letter around 1M+. Write For Me is therefore more than 16 times larger than the next writing GPT in that directory category.

Its growth has also lasted. Earlier public snapshots placed Write For Me near 5M conversations, so the product has accumulated almost ten times that amount since then.

A second writing market grew alongside it: rewriting AI-generated text. AI Humanizer from mmchdigital.solutions currently has around 12M+ conversations. Competing humanizer GPTs range from hundreds of thousands to about a million.

That is a much more specific behavior than “help me write.” Users first began generating text with AI, then a separate audience appeared that wanted the output rewritten to sound more natural or avoid AI-detection systems.

Writing has stayed large because it creates repeat usage. Someone may generate a logo once; emails, reports, articles, applications and rewrites come back every week.

Why are research GPTs getting so many users?

Research GPTs have become the clearest specialist success story in the GPT Store, with Scholar GPT and Consensus alone reaching roughly 60 million combined conversations.

Scholar GPT currently has around 42M+. Consensus has about 18M+. Behind them, SciSpace and Scholar AI each have roughly 6M+.

These products solve a problem ordinary ChatGPT originally handled poorly: finding identifiable papers, searching academic literature and giving answers connected to sources the user could inspect.

The underlying databases are also much larger than a typical GPT knowledge upload. Scholar GPT advertises access to more than 200 million resources across sources such as Google Scholar, PubMed, bioRxiv and arXiv. Consensus launched its GPT around a scientific database of more than 200 million papers. SciSpace currently advertises access to about 287 million papers.

That gives research GPTs a stronger reason to exist than a clever system prompt. The user is getting access to a search layer and corpus that sit outside the base conversation.

And the audience is broader than “academics.” Students, doctors, analysts, journalists, researchers and knowledge workers can all need essentially the same task: find the literature, tell me what it says and show me where it came from.

Are image and video GPTs still huge now?

Creative GPTs are still huge today, although the strongest products have moved well beyond basic AI image generation.

Best Website & App Builder Agent currently sits near the very top of the entire GPT Store at around 63M+ conversations. Canva has roughly 28M+. Video AI by invideo has around 19M+. AI Music Maker GPT is around 14M+, while Logo Creator has accumulated roughly 11M+.

The important change is what the user is asking the GPT to deliver. “Make me an image” has expanded into “make me a site,” “design the presentation,” “produce the video” or “turn these lyrics into music.”

Video AI by invideo is a good example. GPTStore.ai currently lists around 274,000 ratings in addition to its 19M+ conversations, and the GPT can launch invideo's own generation workflow from a brief or script. It pushes the user closer to a finished asset rather than stopping at advice.

Canva follows the same broader pattern. Its custom GPT sits inside a product people already use to make presentations, logos and social media content.

Creation became such a large GPT category for a simple reason: users can judge the result immediately, and the GPT can turn a vague request into something tangible.

Do external tools actually help custom GPTs get more users?

External tools are not required for a custom GPT to become huge, but they give the product a much stronger reason to keep being used as ChatGPT itself improves.

Write For Me reached roughly 49M conversations without needing a unique scientific database or video engine. The language-specific chat GPTs reached even larger totals through positioning, discoverability and repeat conversation.

Consensus, Canva and invideo work differently. Consensus connects users with scientific literature. Canva connects prompts to design workflows. Video AI by invideo can start a video-generation process through its own service.

AI PDF Drive offers another useful example at around 6M+ conversations. Its proposition includes persistent document storage and analysis across multiple chats, something closer to a workflow than a one-off prompt.

This becomes more important as generic instructions get easier for the default ChatGPT experience to absorb. A product that only tells the model how to answer can lose its edge quickly. Access to a useful dataset, account, file system or external action gives users a clearer reason to choose the specialized workflow again.

Do big companies dominate the most popular custom GPTs?

Big software companies have several major custom GPTs, but independent builders still captured some of the largest audiences in the entire store.

Canva, Consensus and invideo are recognizable software businesses. Yet the current top group also includes puzzle.today's Write For Me, awesomegpts.ai's Scholar GPT, gptjp.net's Japanese chat GPT and gptonline.ai's language portfolio.

The concentration inside individual creator portfolios is striking. Puzzle.today has more than 30 listed GPTs, according to GPTStore.ai, but Write For Me has around 49M conversations while its next products are around 100,000. One hit accounts for nearly the entire visible scale of the portfolio.

Gptonline.ai shows the opposite pattern. Its strategy has produced several large GPTs across different languages: around 65M in Russian, 46M in Korean, 22M in Portuguese and 14M in Spanish.

There were at least two viable ways to win the GPT Store: find one enormous horizontal job, as puzzle.today did with writing, or repeat the same successful proposition across several large language markets.

Brand size alone clearly did not determine the winners.

Do the most-used custom GPTs also get the best ratings?

The most-used custom GPTs do not consistently have the highest ratings, which tells us distribution and use-case frequency have mattered at least as much as user satisfaction.

Best Website & App Builder Agent sits around 63M+ conversations with a rating near 3.7. Canva has roughly 28M+ conversations with a rating around 3.5. Video AI by invideo is around 19M+ with approximately 3.9.

Meanwhile, smaller specialist GPTs frequently score 4.4, 4.5 or higher.

Some leaders manage both scale and strong ratings. Write For Me is around 4.3, Scholar GPT around 4.3 and Consensus around 4.4. The Russian and Korean chat GPTs are both near 4.2.

Still, the pattern is messy enough to kill the idea that the Store behaves like a clean quality ranking. A 3.5-rated product can accumulate tens of millions of conversations when the task is common and the product is highly visible.

A 4.6-rated specialist can remain tiny simply because far fewer people need it.

Did GPT Store visibility help create the biggest winners?

GPT Store visibility almost certainly helped the early leaders compound their advantage, although the public data cannot tell us how much of their growth came directly from featuring.

A University of Amsterdam research project that tracked the GPT Store homepage found Write For Me, Scholar GPT, Canva and Consensus among the products appearing particularly often. Write For Me appeared roughly 158 times in the researchers' sample, Scholar GPT around 156 times, Canva around 152 times and Consensus about 125 times.

Those four products now have roughly 49M, 42M, 28M and 18M conversations respectively.

The relationship can run both ways. OpenAI had good reasons to feature GPTs that were already popular, and featured placement then exposed those products to even more users.

Once a GPT built an early audience, that advantage could become difficult to dislodge. More usage created more visibility, more visibility brought more conversations, and the public counter itself acted as social proof.

That helps explain why several names that were already prominent in 2024 are still near the top today even though thousands of alternatives have appeared since.

How concentrated is custom GPT usage?

Custom GPT usage is brutally concentrated: millions of GPTs have been created, while only a tiny group has accumulated tens of millions of conversations.

When OpenAI launched the GPT Store in 2024, the company said users had already created more than three million custom GPTs. Compare that with today's public leaderboard, where crossing even 10M conversations still puts a GPT in relatively rare territory.

The gap becomes obvious lower down the rankings. AI PDF Drive is a substantial product at around 6M+. SciSpace and Scholar AI are also around 6M+. Grimoire, Tutor Me and WebPilot sit closer to the low millions in recent directory data. Plenty of useful professional GPTs remain in the tens or hundreds of thousands.

A GPT with 100,000 conversations sounds successful until we compare it with Write For Me at 49M. The difference is roughly 490 times. Against the 65M Russian-language leader, it is about 650 times.

That kind of distribution looks much closer to YouTube, mobile apps or search results than to a healthy catalog where traffic spreads evenly across products.

Specialists can still become large. Research proved that. Yet the niche needs either a huge recurring audience or some capability people cannot easily get from the default assistant.

What do today's most popular custom GPTs tell us about what people actually want?

Today's most popular custom GPTs show that users mostly want AI to complete familiar jobs faster, with very little interest in learning a complicated new interface or concept first.

Among the current top ten, local-language general chat represents around 170M displayed conversations. Creation products account for roughly 110M. Academic research contributes around 60M through Scholar GPT and Consensus. Write For Me contributes another 49M.

Those categories overlap conceptually, so we should not call this market share. The comparison is still useful because of what barely appears near the top.

There is no swarm of ultra-specific sales agents, obscure professional advisers or complicated autonomous-agent products dominating public usage. The biggest products answer requests people can explain in five words: write this, research this, make this, build this, talk to me in my language.

That simplicity is probably one of the most durable findings from the entire GPT Store experiment.

Use case among the current top 10	Approx. displayed conversations
Local-language general chat	170M+
Creation, design and video	110M+
Academic research	60M+
General writing	49M+

Should someone copy the most popular custom GPTs now?

Copying today's biggest custom GPTs would be a weak strategy because their huge counters reflect years of accumulated distribution as well as genuine demand.

A new generic writing GPT would enter a category where Write For Me already has around 49M conversations and the default ChatGPT can handle most ordinary writing requests. A new language wrapper would compete with established GPTs that already have tens of millions of conversations. A basic image GPT would face built-in image generation plus mature creative products.

The better clue sits inside the products that remain differentiated. Consensus brings scientific search into the conversation. invideo connects the prompt with video production. Canva connects the chat to design. AI PDF Drive adds document storage and analysis workflows.

Those examples point to a more defensible route now: take a job people already perform frequently, then add data, tools or execution that the default model does not automatically provide.

The biggest historical GPTs tell us where demand exists. They do not tell us where competition is still easy.

Are custom GPTs still where ChatGPT is heading?

Custom GPTs are now a legacy format: OpenAI is preparing to retire them and move these workflows toward Plugins, so the current leaderboard captures the peak of an ecosystem that is already changing shape.

OpenAI's Help Center now explicitly says the company plans to retire custom GPTs. Existing GPTs remain usable for the moment, while affected Enterprise workspaces are being given a migration path toward Plugins. OpenAI says other plans are expected to follow the transition, with plan-specific details communicated separately.

The technical direction is revealing. Under the planned migration, a GPT's instructions can become a skill inside a Plugin, while connected apps can move into the same workflow. Custom actions require separate rebuilding rather than automatically transferring.

OpenAI has also stopped allowing new GPT creation on personal Free, Go, Plus and Pro accounts. Existing GPTs can still be used, and business or education workspaces can have different creation rules depending on their settings.

That gives today's usage data a different role. The leaderboard is still extremely useful for understanding what people chose to do with specialized AI assistants. It is becoming less useful as a map of what the next marketplace will literally look like.

As seen above, the strongest use cases already point in the same direction as OpenAI's new architecture: recurring instructions become more valuable when they sit next to useful data, apps and actions.

So which custom GPTs get the most users now?

The custom GPTs with the most observable usage now are the Russian-language чат at roughly 65M+ conversations, Best Website & App Builder Agent at 63M+, Write For Me at 49M+, the Korean 챗 at 46M+, Scholar GPT at 42M+, the Japanese チャット at 37M+, Canva at 28M+, Chat Português at 22M+, Video AI by invideo at 19M+ and Consensus at 18M+.

We cannot honestly call those numbers unique users. OpenAI does not publish the data needed to do that. They are accumulated conversation counters, and they heavily reward products that launched early, appeared frequently in the Store and generated repeat usage.

But the underlying answer is unusually clear.

Four behaviors produced most of the visible mass-market winners: chatting in a preferred language, creating things, writing and researching. The more specialized a GPT becomes, the harder it is to reach tens of millions of conversations, with academic research standing out as the major exception.

The biggest surprise is how ordinary these winning jobs are. Users did not flock mainly to elaborate autonomous agents. They kept opening tools that made an existing task easier to start and easier to finish.

That is probably the part of the GPT Store leaderboard worth carrying forward. Custom GPTs themselves are on the way out, but the demand exposed by hundreds of millions of conversations is still here.