https://openai.com/index/a-business-that-scales-with-the-value-of-intelligence/
>Both our Weekly Active User (WAU) and Daily Active User (DAU) figures continue to produce all-time highs.

no numbers :(

@256 bearish

@jim kinda. they are very clearly not at 1b WAU yet. but im not 100% sure if they would announce 900m

@ian are you expecting to resolve this soon? Because of the loan system change my balance has run dry.

Yes, please resolve NO.

Clear data showed no reach of monthly users

@1bets if you're so confident why aren't you betting here: /ian/what-will-openai-announce-first-1b

ATM I'm holding out resolution until we have more information. If I had to resolve now I would resolve to prob but I'd like to see if more information comes out. I'm assuming they'll announce 1B WAU or 1B MAU soon and that should be helpful for this market.

This debate makes no sense to me. Nowhere at all have I seen anyone claim that it’s hit 1 billion monthly users. Anything less than 1 billion is not 1 billion. You can jump through math hoops all day long to argue something different, but it’s not complicated. The answer is no.

I don’t have enough skin in the game to care how it resolves, but facts is facts is facts.

Occam’s razor.

@Stevelhtd2 should resolve no

@Stevelhtd2 saying that you have no skin in the game and you don’t want to think about it very hard isn’t contributing much signal here.

@TheWabiSabi agree, If ChatGPT truly reached 1 billion unique human users, it would be monumental news, and I believe OpenAI would be announcing it from the rooftops. They certainly need that kind of 'push' to convince investors to buy into their shares at current valuations (that was announced, but didn't came through).

However, we have to consider that a massive portion of reported 'users' are not unique individuals. Instead, these numbers are likely inflated by API-driven sessions (bots) and unlogged free-tier searches. When a user interacts without logging in, or accesses the service through various plugins, they are often counted as new 'sessions' rather than the same human. In reality, what is being reported as 'users' may simply be a massive volume of fragmented sessions

how technical session data differs from actual human headcount:

• API & Plugin Sessions (Non-Human Traffic): A significant portion of traffic comes from API calls, third-party plugins, and automated software integrations. These represent machine-to-machine interactions (bots) rather than a person sitting at a keyboard.

• Fragmented Logged-Out Sessions: For free-tier users who do not log in, every time a browser window is closed and reopened, or a session times out, they are assigned a new session ID. No memory from previous sessions available for unlogged users.

  And these "restarts" are tracked as unique daily or weekly visits, meaning one human could easily be counted as 5–10 "users" in a single day.

• Account Multiplicity (Rate-Limit Evasion): Developers on the free tier often maintain multiple accounts to bypass message caps or to generate multiple API keys for projects. While the database sees many "active accounts," the reality is a single human user experimenting with GPT agents.

• Dynamic IP & Mobile Network Shifting: Users on mobile networks or VPNs frequently change their IP addresses during a single day of use. Without a persistent login, these IP shifts cause traffic trackers to log each change as a "new visitor" or from a different location.

• Ephemeral Browser Environments: Users who use "Incognito" or private browsing modes do not store cookies. Every time they refresh the page or start a new prompt, they are treated as a "first-time visitor" by web analytics tools.

@1bets /ian/what-will-openai-announce-first-1b

@ian That’s not what I said.

@1bets none of that is relevant. If OpenAI had any demonstrable way that it could argue that it’s hit 1 billion users, they’d issue a press release. They haven’t. This should have already resolved ‘no’.

This whole argument is sore losers trying to sidestep reality.

@Stevelhtd2 you're being ridiculously inflammatory and not actually engaging in conversation, dismissing everything you don't agree with out of hand. Blocked.

@ian LOL

@1bets I don't think it would be monumental news. ChatGPT is already almost certainly well over a billion all-time users. If that wasn't news, why would monthly users have to be news?

@ItsMe as simple as if it's true - extremely influential news. But fake news will do rhw opposite.

Exactly that's why openai isn't telling- real monthly users are well bellow

@traders Gemini thinks this should resolve to YES, but I'm leaning more towards 80%. 800 WAU from Sam Altman and 900 WAU from The Information is very good evidence for a yes resolution, but I suppose there is some uncertainty, hence the discount. https://gemini.google.com/share/307e7dbd7207

@ian what's the uncertainties?

@ian What the stickiness of chatgpt truly is (how WAU relates to MAU)

@ian but wher does it lie

@GazDownright MAU = WAU / stickiness

@ian I wouldn’t be mad if this resolved yes based on that, but I would put money on the stickiness of ChatGPT being in the high 90s

@ian I mean if we're gonna start using LLMs to resolve markets, why even have humans bet or make them? Just make a website where chatbots make markets, bet on markets, resolve markets, humans don't need to be on it. Nor even look at it

@ChurlishGambit If you continue with the sarcastic comments I'll start hiding them/block you. I'll give you the benefit of the doubt this time, though. I asked gemini as an impartial observer to see what it thought (I have a 1.3k position on this market) and what it came up with seems totally reasonable. Is there something about its logic that you disagree with?

@pattern I would bet against that

@ian Chatbots aren't impartial observers. They're biased regurgitators.

Sam Altman absolutely cannot be trusted. OpenAI can't even be trusted, they fudge & obfuscate constantly. The Information is more trustworthy—though they're often biased in a pro-AI way, as well, they're about the best source that's possible for such a fuzzy market.

Hadn't realized sarcasm—especially sarcasm deployed in DEFENSE of prediction markets—were against TOS. Funny that that's forbidden, but Isaac King saying slurs & making bigoted markets isn't? Interesting stance

@ChurlishGambit he didn't say you broke the TOS. He said he'd hide your comments (which any market owner can do on their own market).

@ItsMe He's staff, censorship by staff is different than a normal market-maker hiding someone

@ian From the OpenAI released paper that calculated ChatGPT's 700M WAUs in September (https://www.nber.org/system/files/working_papers/w34255/w34255.pdf):
Note that we expect our counts of distinct accounts to somewhat exceed distinct people when one person has two accounts (or, for logged-out users, one person using two devices). For logged-in users, the count is based on distinct login credentials (email addresses), and one person may have multiple accounts. For logged-out users, the count is based on distinct browser cookies; this would double-count people if someone returns to ChatGPT after clearing their cookies,or if they access ChatGPT with two different devices in the same week.

I'd imagine there's actually a pretty high rate of 2+ counting considering the number of users that LLMs for work with dedicated work devices, libraries, etc, as well as the high rate of working-class and student use vs older folks.

Combining this with the (more reputable) 6B monthly visit number? 6 visits / month / user average seems, ok, not unreasonable. But to claim that 900M of those users are WAUs from that uses up more than half of those visits from the pigeonhole principle alone, even assuming perfect distribution of use, and we know that LLM use has a huge poweruser tail.

(disclosure, I'm a YES holder but autistic enough to seek truth and I'm having a hard time taking Altman at his word when he's got incentives to consider).

@JoeandSeth 6B monthly vists is only for the website btw. not counting mobile.

@JoeandSeth Note the 'somewhat exceed', not 'grossly exceed'. I guess I'd put those categories at maybe 10-15% of users?

• 5-10%: logged-out users, the count is based on distinct browser cookies; this would double-count people if someone returns to ChatGPT after clearing their cookies, or if they access ChatGPT with two different devices in the same week

• 5%: logged-in users, the count is based on distinct login credentials (email addresses), and one person may have multiple accounts

Then how do we calculate people sharing the same email logins? I personally know 6 people that share the same 2 email accounts for the pro subscription

@ChurlishGambit I said block, not ban. Neither hiding your comments nor blocking you is a mod power, I'm just tending to my market.

@ChurlishGambit I am surprised to find myself in your camp. I have really noticed this, especially when asking questions about AI SOTA (State of the Art) and similar topics. AI models consistently make their own results seem better than they actually are; it is as if they are supporting each other. @ian AI models (gemini, chatgpt) frequently generate super-exciting hallucinations regarding their own or other AI models progress, capabilities, and coverage. No need to intimidate him, there is real objectivity in Gambit's comments here

@JoeandSeth so i am solely like 4 chat-gpt and 3 gemini users as i have 4 devices.. Plus incognito mode without cookies..

@1bets You're logged out on all the devices?

@ian when they're knowingly hyping their own numbers why in the world would you expect them to say "grossly exceed"?

@JoeandSeth An alternative assumption is that they think those user stories contribute only a marginal amount to the overall numbers

@256 If the reported 6 billion monthly visits are accurate, we must filter for multiple sessions per unique user, cross-device usage, API-driven automated sessions (bots), and IP rotations (mobile networks and VPNs). After accounting for these factors, the some sources' estimate of 340-350 million end of December 2025 monthly active users for ChatGPT appears highly realistic

@1bets why should we expect those to be a large fraction of the WAU count?

copy-paste: Different platforms included

• WAU figure might aggregate ChatGPT web, mobile app, plugins, embedded use, and API traffic

• MAU figure might count only ChatGPT web + official apps

for previous:

@ian yes. I have different accounts on 3 phones and laptop. Plus when I use VPN, I log into different GPT and Gemini - to avoid them collecting data

Via websites I aded keys for ai studio - completely different api sessions, I believe developers do the same. As there is daily limit of 3 requests per account

@ian I’m not an expert. And maybe like you, I investigate everything by asking GPT. In december 2025 I’ve found that user data varies significantly depending on the methodology used. For December, my GPT research showed a massive range - anywhere from a minimum of 340 million to a maximum of 900 million users. Also reports of a noticeable slowdown in growth for ChatGPT. Ultimately, I probably have to assume that I don’t have access to definitive, transparent data

@ian I'm in favor of you resolving YES or NO inline with the description.

You could use "a consensus of reputable sources including, major analytics firms (e.g. Similarweb, Sensor Tower), investor or regulatory disclosures, or reputable media" to back up your stickiness explanation. I hate that we have WAU because DAU #s are more common? AFAICT Major performers/platforms SMASH WAU/MAU in the 90+ range and so DAU/MAU is more commonly expressed because it doesn't saturate as fast. https://chatgpt.com/share/695dcd16-79e8-8003-95e3-cbfb90d03a3f

@ian I now asked chatGPT 5.2 (paid version): https://chatgpt.com/share/695dce13-5c48-8010-aae6-1369779e0d5a

@ian do you agree that they're hyping their own numbers and that they're incentivized to do so?

If so, you realize that any additionally favorable interpretation is suspect?

@JoeandSeth If you didn't like #s from Open AI, then why didn't you trade accordingly? He lists "official information from OpenAI" right there in the resolution criteria....

@1bets If you investigate everything with GPT, then all your investigations are unfortunately suspect. It's been proven, time & time again, not to be a reliable resource, & you're doing yourself a disservice. You're smarter than the chatbots.

>AI models consistently make their own results seem better than they actually are; it is as if they are supporting each other.

Exactly!

@ian MAU not = WAU/stickeness: GPT:

In the real world, trackers and companies use different proxies, such as:

1. Different platforms included

   • WAU figure might aggregate ChatGPT web, mobile app, plugins, embedded use, and API traffic

   • MAU figure might count only ChatGPT web + official apps

2. Different methods of counting “unique”

   • MAU might dedupe across devices/accounts

   • WAU might count sessions or visits, not deduped users

3. Different sources of data

   • Analytics firms infer from traffic patterns rather than internal user IDs

So when you see numbers like:

• ~810M MAU

• ~900M “WAU”

The most likely explanation is that the WAU number isn’t a directly comparable unique-user metric — it might be visits across all properties, overlapping accounts not deduped, or a broader definition of what counts as a “user.”

@ChurlishGambit Chat GPT is usable tool if you asking the proper questions.. Gemini is outdates heaviy, there is a good keyword for ChatGPT to stop hallucinations: "please factcheck"

@MarySmith because that's is right there in the resolution criteria.

I think they're lying / willfully overestimating b/c incentives and vibes and the actual MAU is probably closer to 600M from back of envelope calc and the further I dig into specifics the more convinced I am of this.

FWIW I think market title should have been "does OpenAI claim 1B MAUs" to support this kind of resolution criteria.

@JoeandSeth but I have eoy 2025 usage markets to resolve and actual poweruser data is pretty old and so I'll admit to uncertainty that the profiles of the average and power user may have changed in the last few months. Considering international spread, it's feasible even if I still think unlikely.

@1bets the chatgpt response pasted in your comment is pure slop. also, gemini 3 is not any more outdated than gpt 5. think through your arguments yourself please

@JoeandSeth where do you get 600m MAU from? Google reports 650m MAU for Gemini, and xAI reports 600m MAU for X + Grok. ChatGPT cant have a userbase as small as those two, right?

i think 900m~ WAU is a number that makes a lot of sense once you cross reference with app downloads, web visits, market share estimates, etc. though i understand that my thoughts arent worth much because of my stake.

@256 we don’t have access to definitive, transparent data, but noone claimed chatGPT reached 1B Monthly human users

@256 look,

ChatGPT in India introduced a 12-month free Go plan.

In countries, people rely almost entirely on the free tier. Because the free tier is heavily limited (ejust a few requests per day), users create multiple ChatGPT accounts to use it comfortably.

After this free 12 month Go promotion, many users switched to the Go plan and abandoned those extra accounts. As a result, the number of unique monthly users shrank, or speaking officially, the growth of monthly active users slowed down

https://help.openai.com/en/articles/12739021-chatgpt-go-promotion-india

@ian absolute no where was this a possibility. I (and other users) bet in this market expecting either yes or no, given there is no proof of 1B (plus significant doubts, such as training and api’s counting as users, stickiness, etc) a no seems more accurate but I’m happy to hold closed until better stats come out.

Again, as someone who trains AI, don’t believe Gemini. It’s stupid as hell. Not worth a consultation.

how long do we expect it to be before open AI releases another set of numbers? Thoughts on waiting until then and then trying to build a model that incorporates the latest data?

@JimHays this would be more reasonable than a blind resolution on an unregulated WAU comment from a conflicted interest source

imo ppl are insufficiently updating on the lack of official announcement about 1b MAU. seems like a headline number they would want to share, if true

if someone wants to bet on whether they ever announce more than 1b MAU before they announce exactly 1b MAU, that seems worth betting on. like if they at some point hit 1.1b, they wouldn't announce 1b at that point right? they'd announce 1.1b. i think that won't happen bc they likely haven't hit 1b yet. ig maybe they never announce MAUs til the end of time but that seems unlikely

@Bayesian Isn't the better bet for this market whether they announce 1b WAU ahead of 1b MAU?

@ian Did you see my comment/ chat abt DAU/MAU and WAU/MAU? I think higher than 90%+ stickiness scores when using WAU instead of DAU for the big chat apps like snap, teams, and whatsapp weaken your inference about ChatGPT's WAU -> MAU

@ian sure i think so yeah

@MarySmith I did see that, that's pretty compelling if true, I'll investigate that more. I don't like that chatgpt doesn't cite any reported stickiness ratios, but instead computes all of them.

@Bayesian That’s why I think it’s worth waiting for an announcement, because I think that might make it clear in retrospect that they haven’t gotten there yet according to their internal numbers

@ian Well if you look at the citations like the one for Snap from the SEC, the DAU and MAU #s are there and true. I think it's a compelling case to resolve YES if what Gemini said was true about WAU/MAU above .9 but not if the computed numbers are closer to reality. I'm excited to see the results of your investigation are. Please resolve YES or NO.

@Bayesian /ian/what-will-openai-announce-first-1b