It took ChatGPT two and a half years to go from “meme of the month” to one of the most popular services on the Internet.
When it was first released, in November 2022, as a front-end for large language models, nobody could have predicted that by mid-2025 it would be something that nearly every profession and almost every internet user would be using on a daily basis.
In this article, we will be looking at the most recent ChatGPT statistics as well as general data on the platform’s audience, traffic, revenue, and demographics.
We will also be discussing the latest data on ChatGPT usage, including: MAU and subscribers Monthly and daily usage Global vs. regional users Work vs. entertainment usage Free vs. paid users Generative AI (almost) is ChatGPT.
And for a little bonus, we will also be answering some of the most controversial questions surrounding the service: Do people trust ChatGPT? Are universities and schools adopting it? Are researchers using it?
Global ChatGPT User Growth (2022-2025)
The curve from a bird’s eye view over the past 3 years looks like a pop, then a digesting, then a subsequent climb as the product becomes more robust and more central to users’ routines. There are two dates that I’ll use to mark the beginning and end of this curve:
- January 2023: UBS analysts estimated 100M monthly active users after 2 months of release, the fastest of any consumer app.
- September 2025: OpenAI announced 700M weekly active users.
This is both a gigantic leap and a shift from total users to weekly active users. Before we dive into the data, a quick note that OpenAI has not issued many official statements, and that “visits” by third parties do not exactly translate to “users.”
So, I’ll be using the official numbers as pegs, and focusing on the whether the metric is monthly or weekly, to not conflate apples and oranges.
Quick stats (selected official anchors)
| Period (UTC) | Metric type | Value | What it tells us | Source |
| Jan 2023 | Monthly active users (MAU) | ~100,000,000 | Fastest consumer-app ramp on record at the time; broad early curiosity and utility. | [UBS via Reuters] |
| Sep 2025 | Weekly active users (WAU) | ~700,000,000 | Habitual usage at global scale; reflects deeper integration into daily and work tasks. | [OpenAI] |
Analyst Take
This is a proof of product-market fit followed by a platform moat, in my view as an analyst. That first number in 2023 was proof of broad appeal, but this latter number in 2025 is proof of repeat usage, which is the harder (and more valuable) leap. I think there are two key things to note here.
The first is that the shift from monthly active users to weekly active users is a bullish move. Weekly users is a higher bar to clear than monthly users, and the fact that ChatGPT can leap over it is a testament to the fact that it has evolved from a fad into a utility for hundreds of millions of people.
The second is that much of this growth is coming from outside the US, and much of it is coming for particular tasks involving writing and planning which are likely to be at least somewhat resistant to cyclical disruptions.
My mental model here is that from here we will see some slowing of growth, and some expansion of monetization (paid work accounts, APIs for 3rd party apps, light-weight agency features).
The main bottleneck at this point isn’t demand, but rather ensuring quality, safety, and responsiveness at massive scale while continuing to deliver new features that feel materially distinct.
Monthly traffic and engagement for ChatGPT in 2023 and 2025
Summarized from above section on monthly traffic, daily prompts, and return traffic. Key stats are:
- In 2025, the estimated monthly traffic to ChatGPT is 5.8 Billion
- In 2025, the reported number of daily queries to ChatGPT is more than 2 Billion
- In mid-2023, the reported number of monthly active user visits to ChatGPT was between 1-2 Billion
- In 2025, 80% of traffic to ChatGPT comes from direct traffic (e. g. user typing chat. openai. com into their browser and navigating to the site, as opposed to finding ChatGPT from another online source). This is a proxy for return usage of the site.
Selected Monthly Traffic / Engagement Metrics
| Period | Approximate Monthly Visits | Approximate Daily Prompts / Queries | Insights |
| Mid-2023 (est) | ~1.4 billion visits (June/July 2023) | — | Early phase: growth in awareness and trial usage. |
| Early 2024 (est) | ~4.0 billion visits (rough interpolation) | — | Traffic ramp as features matured and mobile apps deployed. |
| Feb 2025 | ~5.2 billion visits | ~2 billion prompts/day | Usage intensifying: deeper engagement across tasks. |
| Sep 2025 | ~5.8 billion visits | >2 billion prompts/day | Plateauing traffic growth but high retention and repeat-use. |
That’s my take on this data. To sum:
*Mid-2023 monthly traffic was largely new and curious traffic to the site. 2025 monthly traffic is largely repeat traffic to the site. (No surprise, given the number of reported daily prompts to the site in 2025.)
This is a distinction with a difference. Getting a huge surge of traffic in a short amount of time is one thing. Having those users come back day after day, month after month, doing things and producing enormous productivity on the platform is another thing altogether. ChatGPT seems to have transitioned to the latter, and that’s a good sign for its long-term prospects.
Going forward, I expect the rate of growth of monthly traffic to slow (as you eventually run out of Internet users), and other measures of engagement to become more prominent (e.g. frequency and length of sessions, types of tasks being performed on the platform, integration into everyday tools and enterprise software.)
That is to say, 5.8 billion monthly traffic numbers are pretty cool and all, but what you can do with those numbers to generate consistent value and growth is what ultimately matters.
The data clearly suggests that ChatGPT is no longer a novelty, but has become a tool that people use regularly. Going forward, the question is less “how many” and more “how often” and “what for.”
Where are ChatGPT users from?
Interestingly, there are some patterns in terms of where the users are coming from:
The countries with the largest share of desktop traffic to chat.openai.com (ChatGPT’s website) as of Sep 2025 are:
- Japan: Possibly due to the fact that users there use ChatGPT for “typical keyboard-based activities” such as writing, searching, translating text, or writing code, because it makes these tasks easier for them.
- China: The results are due to a combination of product availability, language support, local interest in AI, and possible access routes.
- United States: It comes as no surprise that the country that gave birth to ChatGPT is among its most frequent users.
- Brazil: Brazilians are also among the most frequent users of ChatGPT.
Share of ChatGPT website traffic by country (desktop, Sep 2025)
| Rank | Country | Share of visits | Quick read |
| 1 | Japan | 36.4% | The outlier: exceptionally high desktop share; strong consumer curiosity and heavy student/knowledge-worker usage. |
| 2 | China | 12.0% | Significant interest despite access frictions; mix of direct use, workarounds, and cross-border traffic. |
| 3 | India | 7.2% | Broad adoption in education and tech services; high propensity for repeat use in study and coding tasks. |
| 4 | United States | 6.9% | Deep enterprise and prosumer usage; higher share likely when accounting for mobile/app and API. |
| 5 | Brazil | 2.3% | Large, fast-growing base; strong uptake among students, marketers, and SMBs. |
| — | Others | 35.1% | Fragmented long tail across Europe, Southeast Asia, and LATAM. |
Method note: India: Indians are one of the most frequent users of ChatGPT. The chatbot is used by students for educational purposes as well as developers since it makes coding easier. Another reason for its popularity among Indians is the English-speaking population.
Analyst’s view
The table above represents the percentage of traffic that ChatGPT’s website has received from the above-mentioned countries. Though it doesn’t directly represent the number of users, it is an indicator.
Also, these numbers are for desktop traffic only, the actual numbers will be higher as they also include mobile traffic.
Which countries are driving the most users of ChatGPT in 2025?
Now on to the relevant and interesting stats, and what they could potentially signal for the future of AI adoption globally.
Based on August and September traffic share, the US is the number one country accessing ChatGPT, with India and Brazil having solidified in second and third place:
| Rank | Country | Approximate Traffic Share % | Insight |
| 1 | United States | ~15.1% | Mature market with strong use across enterprise and consumer verticals. |
| 2 | India | ~9.3% | High growth rate; large population, rising digital fluency. |
| 3 | Brazil | ~5.3% | Latin America’s strongest point of adoption, regional hub. |
| 4 | United Kingdom | ~4.3% | Western European representative, significant secondary market. |
| 5 | Indonesia | ~3.7% | Emerging market, significant mobile-first population. |
| 6 | Japan | ~3.5% | Strong desktop usage, high tech literacy, niche but steady. |
| 7 | Germany | ~3.4% | Key European market, high potential for business/industrial usage. |
| 8 | France | ~3.2% | Similar position as Germany; EU base for expansion and regulation. |
| 9 | Philippines | ~3.0% | Southeast Asia region–led adoption, promising for growth. |
| 10 | Canada | ~2.6% | Smaller population but strong digital infrastructure and usage. |
| — | Others | ~46.9% | The remainder spread across many countries, signifying truly global reach. |
United States makes sense. It has the best language support, excellent internet infrastructure, and many of ChatGPT’s commercial clients are there. India, Brazil, and South-East Asia surprised me a bit more.
I think this is important since this shows that ChatGPT isn’t just an American or Western phenomenon. It’s delivering value in a ton of other markets as well. To me, that implies a ton of potential for growth in those markets.
I also see how large the “Other” category is, with nearly half the users coming from this long tail of countries.
This implies that from a business perspective, although English language markets are obviously crucial, there’s plenty of growth to be had in expanding the service to more languages and cultures. Some things I would consider there include:
- Localization, translation to more languages, adaptation to local cultural contexts
- Mobile-first markets, particularly in developing countries, where the smartphone is the primary means of internet access
- Industry-specific applications beyond the standard consumer ones, e.g. education, programming, SMB services, etc
In general, I think this country-level data can give us some insights into where interest in ChatGPT and products like yours that are built on top of it will grow in the coming years.
Who are the ChatGPT users? Information about their age and gender (as of 2025):
I have taken a look at the most recent available data on ChatGPT users in 2025, and below I have plotted a few interesting trends that I observed. The data tell us not only who the users are, but also provide hints on how different groups of users are using the tool. For the remainder of this post, I will simply present the data, and provide some thoughts on what it might all mean. Here are the key demographics:
Gender: As of July 2025, Open AI themselves have reported that now 52% of users have names that could be classified as female, compared to 37% in January 2024.
Age: As of March 2025, the largest age group is 25-34 years old (at 30.6%), followed by 18-24 and then 35-54.
Demographic table
| Demographic category | Share of users (approximate) | Notes and interpretation |
| Gender – Male | ~54.8 % | March 2025 estimate: male users remain a slight majority. |
| Gender – Female | ~45.2 % | Shows narrowing of gender gap in adoption. |
| Age 18-24 | ~23.3 % | Younger adults/adolescents: high adoption but maybe more casual use. |
| Age 25-34 | ~30.6 % | Most significant age block: prime working-age, overlapping personal & professional use. |
| Age 35-44 | ~19.1 % | Mid-career professionals, showing strong uptake beyond early adopters. |
| Age 45-54 | ~13.8 % | Use is lower but meaningful among more senior users. |
| Age 55+ | ~13.2 % | Older generations still smaller share; adoption rises but lags younger cohorts. |
What our analyst says
“First of all, I think point one is the most interesting. It’s great that men and women are equalizing here, it means the platform is moving out of the early adopter stage (very male dominated) into more of the mainstream.
Going into the mainstream is great, but it also means more scrutiny and more expectation. If you have a tool in the early adopter phase, you just need to focus on growth.
Once you start moving into the mainstream, you need to focus more on usability, inclusion and access, which is working for all age groups.
“The age groups itself is fascinating. The biggest age group is 25-34. This isn’t surprising. This age group has so much to balance from work, to study, to side hustle, to social life, and the use cases are so huge for them to use AI for writing, research, productivity, ideation, problem solving. The next biggest group is 18-24.
Once again, not surprising. They’re still studying, or just started their careers, and similar use cases to 25-34. The 45-54 and 55 and over group is smaller, but still significant. This is important because it means that the platform is being used by people outside of the core demographic, and is becoming more mainstream.
This matters for businesses and developers building on top of ChatGPT because it means they need to think about the user interface, does it make sense for less technical people? Are the prompts written in a way that makes sense for different age groups? Is the language readable for people that aren’t native English speakers or for older people?
“The focus of the platforms changes over time. In the early days it’s all about user acquisition. As you start to keep growing, it’s about user retention and getting users to come back to the platform. This is just as important, if not more important.
Overall, this data shows that ChatGPT isn’t a Gen Z phenomenon or a tool just used by technologists, it’s a mainstream productivity and creative tool.
The next step is figuring out how to retain these users and get them to come back for more; it’s great to have all these users, but you can’t just live off your early adopters; you need to figure out how to create loyal users for the future.”
2023-2025, ChatGPT Downloads and Usage on Mobile
The chart above shows how the ChatGPT app grew from a minuscule iOS beta to an app opened 10s of millions of times daily. The first big jump (on the left) is the launch of the iOS app in May 2023. The other big jump is the Android app launch, and 2025 was when it clearly became a mass market app.
The above graphic is from the same report and shows not only that there was an increase in downloads but what the app started being used for the same can be said of other, more general uses on-the-go.
Initially, users largely accessed it to help them professionally such as composing emails or helping with document summaries, more “how do I get through the day” types of queries.
But as time went on the usage began to pivot from just that. People were looking for lifestyle suggestions, advice, even asking health-related questions, more of a “general day-to-day helper” app.
By July 2025, the app had become the fastest to reach 1 billion downloads on the iOS and Google Play stores, according to Sensor Tower. That’s a wild little fact. What is more intriguing is how the timing of its use evolved.
It was no longer limited to being accessed only during the workday but rather throughout the week and even at night and on the weekends, whenever the user felt they needed help answering a question or coming up with an idea.
The following are times between 2023 and 2025 where the sources specify the exact dates these changes occurred:
| Period | Metric | Value | What it signals | Source |
| May 2023 (iOS launch, U.S.) | First-week installs | ~606,000 | Strong lift-off despite single-platform, limited-market launch. | |
| Q3 2024 (global) | Quarterly downloads | ~90 million | Re-acceleration a year after launch; momentum sustained. | |
| H1 2025 (global) | Half-year downloads | ~470 million | A new surge; ChatGPT leads GenAI apps in installs and revenue. | Sensor Tower |
| End of Jun 2025 (global) | Cumulative installs | ~940 million | Poised to cross the 10-digit mark. | Sensor Tower |
| Jul 2025 (global) | Cumulative installs | 1 billion | Fastest app to 1B downloads across iOS + Google Play. | |
| 2025 (mobile usage trend) | Weekend usage on mobile | Weekend/weekday gap narrows | Usage expands beyond work/education into lifestyle & entertainment. |
Analyst commentary
1 billion isn’t really about the “1 billion,” it’s about the diversification of use cases: The first week was certainly impressive, but it was largely a release of pent-up demand: By mid-2025, we saw a more significant indicator: more downloads, and more diversified mobile queries, such as shopping, cooking, fitness, and entertainment queries.
The single biggest indicator of this trend is the closing weekday/weekend usage gap: users aren’t just playing with ChatGPT, they’re incorporating it into their daily lives:
This trend has two business implications:
The bar for mobile user experience will continue to rise fast cold-start performance, offline-enabled experiences where possible, seamless voice interactions, and smooth handoffs to other apps and services will become increasingly important as users expand the scope of how they use ChatGPT:
Second, the growth of the business will increasingly shift from growth in downloads to growth in retention and frequency of use such as average number of prompts per session, frequency of return visits, and whether users use mobile to make day-to-day decisions on-the-go.
I expect year-over-year growth in downloads to moderate, while engagement on mobile will continue to deepen, particularly as multimodal interactions and simple embodied agents become everyday gestures rather than “demos.”
ChatGPT Subscriptions and Paid Subscribers (2023-2025)
When it comes to ChatGPT as a revenue business, the most significant fact is that it’s crossed the chasm from curiosity to cash cow in less than 2 years. I share the numbers below followed by some thoughts on what this means for the broader AI ecosystem.
The numbers
Via The Information:
- OpenAI says paid subscribers rose from around 5.8M at the end of 2023 to about 15.5M in 2024
- ChatGPT brought in at least $333M per month in subscription revenue in late 2024 (which is at least $4B annualized) and $415M per month (which is roughly $5B/year) in early 2025
- The company said ChatGPT had a $10B run rate for annualized revenue (subscriptions + enterprise) for the entire ChatGPT family of products in June 2025
Paid Users & Revenue Estimate (2023 to 2025)
| Year | Approx. Paid Subscribers | Approx. Subscription Revenue | Notes |
| 2023 | ~5.8 million | Estimated ~$1–2 billion | Early monetisation phase; basic Plus plan only. |
| 2024 | ~15.5 million | ~$4–5 billion | Rapid subscriber growth; more tiers introduced. |
| 2025* | ~20 million+ (est.) | ~$5–10 billion (run-rate) | Includes enterprise, up-tiers; full-year data still immature. |
*All of these metrics are annualized and estimated as of mid-year for 2025.
Analysys Take
The ability of ChatGPT to build a paid business is in my opinion less about the technology or even the strategy of the company, and more about the transition of generative AI out of the “toy” phase and into the “things people will pay for” phase.
To see such growth in both subscribers and revenue shows that people are willing to pay for the tool, and the revenue model is working. This does not mean that there are no questions left unanswered, such as:
Can the business be sustained? With a US$10B run-rate in 2025, OpenAI will need to continue to add value to the tool (for example, new features or vertical integrations) and/or expand the use-cases of the tool into new markets (for example, even more enterprise use-cases).
Failing that, the business will not be sustained once users are no longer satisfied with just paying for the tool.
How sustainable is the cost structure of this business? Due to the high cost of training and running these AI models, a very large revenue number does not necessarily mean a very large profit.
Will the business be able to broaden its reach? Most of this growth has been in early adopters and power users, as well as in the US, and from companies.
OpenAI will need to start to find traction with more casual users and at lower price points, and in more markets, where the cost of subscriptions and the range of acceptable payment options may be very different.
In summary, ChatGPT has proven that the subscription business can be economically significant. The next step will be to broaden and deepen the monetization, not just grow the number of users and subscribers.
OpenAI Overall Revenue and Valuation (2020–2025)
I think the OpenAI funding story is the fastest growth story in tech history. Below is a sampling of revenue and valuation milestones superimposed over the AI market. Once the data has been presented I’ll add a little color on what this might mean going forward. Reported figures are:
- Revenue: From a few millions in 2020 to $3.7 billion in 2024 to >$12 billion annualized by mid 2025.
- Valuation: The private market valuation of the company has gone from tens of billions in early 2024 to $300 billion in March 2025 (after a $40 billion financing) and has reportedly hit $500 billion via a secondary share sale in late 2025.
| Year | Approximate Annual Revenue* | Approximate Valuation | Notes |
| 2020 | ~$3.5 million (reported) | — | Very early commercial phase. |
| 2021 | ~$28 million (estimate) | — | Revenue still modest relative to scale. |
| 2022 | ~$200 million (estimate) | — | Prior to mass rollout of ChatGPT; growth accelerating. |
| 2023 | ~$1.6-2.2 billion (estimate) | — | Adoption accelerating; enterprise traction visible. |
| 2024 | ~$3.7 billion | — | Publicly cited figure for the year. |
| 2025 | ~$12 billion (annualised) | ≈ $300-500 billion | Mid-year run-rate; valuation at ~$300B (March) and ~$500B (Oct). |
Revenues are approximate, based on reported annual or annualized run rates, and industry estimates.
Analyst’s note
It’s difficult not to be both in awe and a bit uncomfortable about OpenAI’s trajectory: From a research institution to a tens-of-billions-of-dollars-revenue company, with a valuation of almost half a trillion. On the one hand, it’s an extraordinary success story.
OpenAI has grown faster than probably 99% of all software companies. ChatGPT was a wildly successful consumer hit that dramatically increased its global user base.
In addition, it has built an enterprise and API business that is now driving material commercial revenues. On the other hand, here are a few things worth keeping in mind:
- Valuation vs. revenue If your valuation is in the $300 to $500 billion range and your revenue is around $12 billion, that’s a revenue multiple of 25× to 40× or more depending on the actual number. That’s a very high revenue multiple for most software companies. What it really means is that investors are pricing in a lot of the future, much better monetization, continued hypergrowth, and sustainable market dominance.
- Profitability and cost structure Revenues are growing fast. But profitability is a very different matter. It’s really expensive to train and deploy large AI models. Infrastructure costs are massive, and the talent to build and operate those models is scarce and expensive. The big question is whether OpenAI can turn its scale into sustainable profitability and free cash flows.
- Growth rates Doubling or tripling your revenue is great when you’re starting from a low base. It’s much harder when you’re already at tens of billions of dollars. Sustaining that kind of growth over multiple years will be increasingly challenging.
- Valuation risk Whenever you have a valuation that is highly dependent on the future (growth or margin expansion), there is some risk involved. If growth slows down even modestly or if margins are compressed due to infrastructure costs or competition, the valuation may be impacted as well. Or to put it differently: A half-trillion-dollar valuation is a very big bet on the future economic value of AI.
The next step
In a way, it feels like OpenAI is at an inflection point. The company has proven it can grow extremely fast.
The next challenge is to prove that the business model can deliver sustainable profits, and that its ecosystem of models, tools, enterprise integrations, and partnerships delivers sustained value.
If that happens, today’s valuation may look a lot more reasonable a few years from now. If not, we may see some multiple compression.
A historic trajectory
What’s so remarkable about OpenAI’s rise isn’t just the absolute numbers, it’s the speed with which this is all happening. Both the revenue growth and the valuation expansion have been exceptionally fast, even for our tech industry standards.
The next two to three years (2025 to 2027) will tell us whether this is a story about the emergence of one of the most important platform companies of the next decade or a cautionary tale about growth and cost expectations eventually colliding with reality.
Who is currently the most used AI chatbot (2025)?
What are the most popular chatbot models being used today? Well, if we take a look at the market share of chatbots in 2025, we can get an idea. ChatGPT is still the most used chatbot by a long way, but some new entrants are rising to prominence.
According to StatCounter, in September 2025, ChatGPT had around 81 percent of the global AI chatbot market share. The rest of the market share was comprised of Perplexity, Microsoft Copilot, Google Gemini, and Claude.
A Similarweb report from September 2025 also estimated that the top chatbots used globally are: ChatGPT, Perplexity, Google Gemini, Microsoft Copilot, and Claude.
The market share estimates below are based on StatCounter data, and refer to the proportion of global chatbot traffic each model receives. The market share data is as follows:
Global AI chatbot market share (September 2025)
Please note that ‘market share’ in this context, refers to the share of global chatbot traffic. In other words, it’s a (very) rough estimate of usage, rather than a measure of revenue or number of users.
| Rank | Product | Estimated share (%) | Notable takeaways |
| 1 | ChatGPT | ~81.1 | Still the default destination for consumer and prosumer use. |
| 2 | Perplexity | ~10.8 | Research-style answers and live web context give it a durable niche. |
| 3 | Microsoft Copilot | ~4.1 | Distribution via Windows/Office fuels steady gains. |
| 4 | Google Gemini | ~2.8 | Strong brand, but chatbot traffic share trails its broader AI footprint. |
| 5 | Claude | ~1.0 | High satisfaction among knowledge workers; smaller absolute footprint. |
What does this mean for you?
My guess is that chatbot adoption is subject to a strong network effect. ChatGPT will be the most popular chatbot in 2025, not just because it’s the best model, but because it’s also the most well-known, the oldest (it launched in late 2022), and the one with the most integrations. This matters.
At the same time, each of the other large chatbots seems to have found its own niche:
- Perplexity: Strong knowledge retrieval and source-backed answers.
- Microsoft Copilot: Integration with Windows and Microsoft Office.
- Google Gemini: Integration with the rest of Google.
- Claude: Safety-focused, cautious chatbot.
Because of this, it seems likely that any of these other chatbots could potentially overtake ChatGPT under the right conditions. My hypothesis is that the two factors that will matter most going forward are:
- Embeddedness: How well-integrated is the chatbot into the tools people are already using? The more invisible and intuitive this integration is, the more people will tend to use the chatbot.
- Agentic capabilities: How well does the chatbot actually help users accomplish tasks, rather than just generating text?
The chatbots that have strong integration and strong task-completion, while also offering good enterprise plans, security, and mobile support are likely to win the most market share.
Estimated use of ChatGPT by industry and profession (2025):
Take 2025 in a realistic context: There are two narratives that play out. From an industry perspective, ChatGPT has become a notable source of traffic, particularly to reference, education, news, and shopping destinations.
On the job front, the bulk of the work is everyday knowledgework: ChatGPT is used for planning, decision making and writing, with coding being a bit of a minority application than the hype may imply.
The large 2025 usage survey by OpenAI shows that “Practical Guidance,” “Seeking Information,” and “Writing” comprise almost 80% of the messages, and that at work in particular writing is the biggest task and decision-support (or “asking”) is popular with most occupations.
Separately, Similarweb reports that in 2025, AI websites (mostly ChatGPT) generated over 1 billion referral traffic in June from categories like reference, science/education, news/media and e-commerce.
By the numbers
Tasks by occupation. According to OpenAI’s research (May 2024–June 2025 messages), “writing” accounts for ~40–42% of workplace messages, and “getting/recording information” and “making decisions/solving problems” are among the most common tasks in nearly every occupation, including management, business, STEM, administration, and sales.
4.2% of the messages are about programming.
Referral traffic from ChatGPT: As of June 2025, Similarweb reports 1.13 billion referrals from AI platforms, with ChatGPT driving >80% of those referrals to the top 1,000 websites on average.
Other category leaders are Wikipedia (Reference), ResearchGate (Science/Education), Reuters (News/Media) and Amazon (Ecommerce).
Table — Usage snapshot by industry and profession (2025)
| Dimension | 2025 datapoint | Example indicator | What it implies |
| Professions – Core work task | Writing ≈ 40–42% of work-related messages | Writing tops management/business; editing & drafting dominate work tasks | ChatGPT is a writing accelerator across white-collar roles. |
| Professions – Decision support | “Asking/decision support” is ~49% of all messages; majority of work messages in 2025 tilt to “Doing” but still decision-heavy | Common across management, business, STEM, admin, sales | Tool is used to think through choices as much as to produce output. |
| Professions – Programming share | ~4.2% of all messages | Coding far smaller than writing for ChatGPT Consumer users | Contrary to perception, code is not the primary workload here. |
| Industry – Reference | Wikipedia ≈ 9.2M ChatGPT referrals (June 2025) | High referral intensity to encyclopedic/reference content | ChatGPT acts as a front door to authoritative fact bases. |
| Industry – Science/Education | ResearchGate ≈ 3.2M ChatGPT referrals (June 2025) | Coursera, Udemy also rank | Strong pull from learning and research workflows. |
| Industry – News/Media | Reuters ≈ 1.7M ChatGPT referrals (June 2025) | Yahoo/Yahoo Japan lead overall | Users click through for sourcing and verification, not just summaries. |
| Industry – Ecommerce | Amazon ≈ 4.0M ChatGPT referrals (June 2025) | Etsy/eBay/Walmart also in top ten | Growing role for AI-assisted product discovery and comparison. |
Quick note: The B2B data above is based on two different data sources: (a) messages to ChatGPT that we attempt to reverse-engineer to understand the jobs and tasks the user is doing, and (b) referral data from AI that shows ChatGPT as now the #1 referrer to major destinations.
Neither dataset is fully representative, but when taken together, we think they do a decent job of capturing usage patterns and referral destinations for ChatGPT.
What the analysts say
If you put these two datasets together, there is one overarching theme that emerges: knowledge work is the dominant use case. Everything else is a secondary function.
Most of the use cases that people have for ChatGPT involve explanations, problem-solving or content creation. This is why writing and decision-support appear across so many job categories.
Conversely, the referral data points to a second emerging role for AI: intent router. Increasingly, these services are going to help users jump straight from a conversation into whatever content, research or product they need, without having to fire up a search interface.
If that’s the case, there are two logical conclusions that follow.
First, writing-heavy occupations, consulting, ops, law, education, are going to be some of the largest users of AI in the coming years. These jobs involve a lot of writing, editing and revising, which is what ChatGPT does best.
Second, AI-driven discovery is opening up a whole new avenue for monetization. Companies that figure out how to optimize their content for AI, with more metadata, answer-driven pages and brand recognition, will be the ones that hoover up most of the traffic.
The bigger point here is this: by 2025, ChatGPT is playing two roles at the same time. Productivity app for knowledge work, and traffic engine that drives destinations around the web. And that’s why it’s growing so fast.
Top applications for ChatGPT (2025)
What are users actually using ChatGPT for in 2025, then? Based on these numbers, it looks like much of the usage is concentrated on the writing, tutorials, and information retrieval use cases rather than coding or graphic design.
Fast facts: OpenAI reports that in May 2024 to June 2025, 28% of overall conversations were about “writing” (edit, rewrite, write), and 42% of professional use cases. In the same study, they found that 49% of messages were about “asking/decision-support” and 40% were “doing” (requests to do something, generate something).
A third-party analytics firm reports that after “practical guidance” (28.3%), the next most popular use cases were “writing” and “seeking information” : Numbers are rounded and based on aggregated conversation labels provided by OpenAI and summarized by analytics platforms.
Top Use Cases for ChatGPT (2025)
| Rank | Use Case | Approximate Share | Comments |
| 1 | Writing assistance | ~28% overall / ~42% in work-chats | Includes drafting, rewriting, summarising text. |
| 2 | Practical guidance / advice | ~28.3% | How-to questions, personal guidance, planning tasks. |
| 3 | Information-seeking | ~20–25% | Research, fact lookup, knowledge extraction. |
| 4 | Task execution (“doing”) | ~40% in work-context | Generating content, completing forms, producing materials. |
| 5 | Coding/development help | ~4% | Relatively small proportion despite large attention. |
The analyst’s take
There are a few things that jump out at me here.
One is that the core use cases for ChatGPT are all about text-based ideation and work, writing, editing, summarizing, and decision-making. Performing tasks and coding are definitely going to be part of that, but they are a trailing indicator rather than a leading indicator.
If you are building on top of AI, this is worth remembering. Most of your users are not going to be developers, they are going to be writers, professionals, students, researchers, and content creators.
The second thing that stands out is that how-to advice and information seeking are taking off, that means that ChatGPT is encroaching on what might traditionally have been search, research, or even education.
The number of conversations that are now about how-to advice or decision-making demonstrates that users are starting to think of ChatGPT as something more than a text engine, increasingly they are thinking of it as an advisor, consultant, or thought partner.
If you are building on top of AI, that means you should focus on ease of use for non-technical users. You should build tools that integrate into people’s everyday workflow, into their writing cycles, research cycles, and decision cycles. And you should consider how to monetize high frequency actions like editing and summarizing.
I’m not sure it makes sense to make a huge strategic bet that people will suddenly start coding with ChatGPT. It might make sense to bet that people will start writing and thinking with ChatGPT.
Ultimately the power of the product is its ability to help people structure their ideas and communicate more effectively, and that versatility is what will lend the product longevity. The opportunity in the future is to expand that set of everyday use cases, not to shrink it into a niche.
Perceptions of, and trust in, ChatGPT (2025)
The data I have regarding attitudes toward ChatGPT in 2025 are mixed: mostly curiosity, suspicion and perhaps confusion.
The key take-aways are below. A table with the figures is at the bottom of this article.
Highlights
A 47-country survey found that 66% of respondents report frequent use of AI, but only 46% report that they are willing to trust AI. An American survey found that 29% of adults trust ChatGPT, more than some rival AI chatbots.
That same American survey found that 75% of US adults say they never use chatbots like ChatGPT to consume news, and that even among those who do use chatbots, many report having difficulty distinguishing between what is true and what is not.
Table: Public perception and trust metrics for ChatGPT (2025)
| Metric | Approximate Value | Context / Notes |
| Global willingness to trust AI systems | ~46% | Reflects broad public sentiment across 47 countries. |
| U.S. adult trust in ChatGPT | ~29% | Respondents who say they trust ChatGPT “somewhat” or more. |
| U.S. adults who use AI chatbots for news | ~10% (2% often + 7% sometimes) | Majority do not rely on chatbots for news. |
| Among U.S. chatbot news users who say they see inaccurate content at least sometimes | ~50% | Suggests uncertainty about accuracy. |
Analyst’s view
To my eyes, there are a few different ways to read these numbers. On the one hand, they point to adoption and awareness: lots of people have heard of ChatGPT and similar tools, and many are actually using them.
On the other hand, trust is still an open question. The fact that less than half of people globally say they are willing to trust AI is a sign that adoption and familiarity aren’t the same thing as faith.
I’d read that as a sign of both upside and downside potential. On the upside: a substantial share of people are already open to using ChatGPT, which suggests there’s still lots of room for further engagement and deeper trust.
On the downside: if users don’t perceive ChatGPT as more accurate, more reliable, or more transparent in the future than they do today, they might not engage as deeply as they otherwise could, or they might stop using it altogether.
If I were advising a business or developer: the emphasis needs to shift away from user acquisition and towards what you might call “trust engineering”, which means steps to help users feel confident in the results they get, be able to verify what they get, and understand the limits of the tool.
For example: it could help to have clear disclaimers, or an accuracy score, or human review built in.
In short: ChatGPT is already making huge strides in awareness and use by 2025, but trust remains an important constraint.
The next stage of its maturation won’t just be about growth in absolute numbers; it will be about growth in confidence and dependability.
ChatGPT in Education and Research (2025)
If you ask users (students, researchers, etc.) what they are doing with ChatGPT in 2025, you will get answers that include studying, writing and information seeking. There are two signals.
The first signal is a broad analysis of use cases from OpenAI that shows education-adjacent use (studying, studying help, writing, information seeking) as one of the top categories, with total weekly users now in the hundreds of millions and total weekly messages in the billions.
The second signal is that, according to web traffic analysis, ChatGPT is sending massive traffic to education and academic sites, implying that users are not just consuming the summary, but are also clicking through.
Quick stats
Usage: In the 2025 OpenAI report, we see 700M weekly active users and ~18B weekly messages, as an indication of the popularity of studying and research within the broader tool use.
Education & research traffic: As of June 2025, Similarweb reports ChatGPT sending millions of users to education and academic sites, including ResearchGate (3.2M), Coursera (1.1M), archive.org (0.85M), academia.edu (0.81M) and nature.com (0.61M), with Wikipedia (~9.2M) being the top one.
Table: Education & research indicators (2025)
| Metric (June/Sep 2025) | Value | What it indicates |
| Weekly active users (global) | ~700,000,000 | Mainstream adoption; education use sits within a very large base. |
| Weekly messages | ~18,000,000,000 | Frequent, repeated study/research interactions at scale. |
| ChatGPT referrals → Wikipedia (reference) | ~9.2 M / month | Heavy use of background reading and citation checks. |
| ChatGPT referrals → ResearchGate | ~3.2 M / month | Direct pathways to papers and researcher profiles. |
| ChatGPT referrals → Coursera | ~1.1 M / month | Follow-through from Q&A to structured courses. |
| ChatGPT referrals → archive.org | ~0.85 M / month | Access to open archives, books, and datasets. |
| ChatGPT referrals → academia.edu | ~0.81 M / month | Discovery of preprints and author pages. |
| ChatGPT referrals → nature.com | ~0.61 M / month | Movement from summaries to peer-reviewed content. |
Sources: OpenAI usage data shows total product usage, of which education and research are key use cases; total referrals data based on Similarweb June 2025 data estimating AI-platform referrals to the top 1,000 websites (of which ChatGPT makes up the vast majority).
What our analysts say
And, from where I sit, 2025 is a milestone: ChatGPT is no longer a way to cheat on your homework, but an entry point to learning.
The referral stats are important because they tell us that it’s not just what happens in the chatbox: people use conversational responses to figure out what to read next, and then they go to trusted sources.
That’s a better pattern for scholarly use than a simple “chat-for-answer” cycle.
This leaves two great tensions. 1) Depth vs speed: the product is really good at quick building blocks like summaries and outlines and problem setup, but still the institution will want to know that students and researchers are using the source material.
Second, citation hygiene: when usage reaches a critical mass, the tools that facilitate easy citation to specific versions will play a key role in switching the university stance from passive support to active support.
What I hope is that education will continue to incorporate ChatGPT more and more, and that’s great, but that it also gets to the point where it is a tool for regular learning tasks such as drill problems, fact-checking, and eventually transferring your results directly to your learning management system or notebook.
While the tools are getting us closer, I think the big step will be when strong learning becomes the default behavior.
If we combine the 2022-2025 data, we see not just exponential, but normalizing distribution.
ChatGPT is now not just an application but an ecosystem, where students, professionals and businesses are coming together, all dependent on this conversational AI.
What’s revealing is the product’s market share, its growing subscription revenue and its continued spread of users across ages, locations, and occupations. It shows a technology moving from fad to fixture.
However, the statistics also reveal the next hurdle. As people consume more, they will increasingly raise concerns about trust, credibility, and the learning value.
What matters now is that OpenAI, and other such platforms, maintain their integrity as they grow and spread to new areas.
ChatGPT is the most visible yardstick in the sweep of AI metrics.
The period between 2022 and 2025 represents the years when generative AI entered the mainstream. It is no longer the stuff of tomorrow, but a common tool for learning, working, and creating.
