Skip to main content
MANIFOLD
Will chatGPT fall below 75% of AI Chatbot market share in 2026?
71
Ṁ1kṀ22k
Nov 11
65%
chance

Resolves to this site: AI Chatbot Market Share Worldwide | Statcounter Global Stats

ChatGPT has steadily held 80% market share according to that site.

  • Update 2026-02-04 (PST) (AI summary of creator comment): Market resolves YES if ChatGPT falls below 75% market share at any point during 2026 (not just at year-end), as measured on the linked Statcounter site for all platforms worldwide.

  • Update 2026-10-05 (PST) (AI summary of creator comment): The market will resolve using the monthly time series from the linked Statcounter site.

Get
Ṁ1,000
to start trading!
Sort by:

Genzy asked me to look at this market because he doesn't want to be seen as influencing the outcome given he has a substantial position in the market. He told me to "take it over" so you should consider whatever I write here as being from the market creator.

I've read all the comments. I'm not afraid of Genzy. I have no stake in this market. I have no inherent interest in this figure. I don't care who holds what positions and haven't even looked.

---

I see there was a considerable effort from Evan Wang to resolve the market Yes based on the linked source showing the necessary data and several earlier posts from Genzy confirming the linked source was the correct source of truth. Some of Evan's links were to alternative sources for the "true value" but I am not giving these any outsized weight and did not rely on them for my own attempts at researching the market.

I see there is an effort from multiple people especially @spiderduckpig and @SqrtMinusOne to explain that there is a fundamental issue with the linked data source that invalidates its status as the source of truth for the market.

I see that Genzy has stated plainly that the market will resolve Yes (either now or before the end of the year), or No (at the end of the year), but not N/A. Genzy has also repeated in multiple comments that the source is one website and a specific view on that website. The market description has been consistent for the entire duration of the market, only clarifying deeper, never swaying.

---

I reviewed the linked website. Superficially the monthly report looks like an unlikely but potentially real drop from 80% to 51%. This could easily be explained as an actual methodology update that revealed a corrected value, 80% was too high, etc.

However, the daily view shows a less believable situation where the rate plummeted from 75% to 18% within a few days, then fell again to just 5% for the start of November. It looks like data from Desktop dropped within a few days after 8 October 2026, then Mobile did a first drop on the 8th and a follow-up drop on the 28th. Tablets held at normal rates until the 28th then dropped. So there is a strong argument that data collection from multiple sources has been impacted. One user has tied this to a real world event that could mostly explain it.

I also reviewed several other charts on the site to see if they had similar weird behavior. I see that the share of Desktop vs Mobile devices (not on AI chatbots, just general platform share) has actually moved significantly as well. They were fairly close with a slight edge to mobile (54-55%) for most of 2026, but then suddenly in the first week of September, mobile devices shot up to 60% and Desktop fell to 39%. So that is not the only suspicious change on the site. I think there is evidence that any participants could have seen, that figures on this site can and do change in significant ways on short timescales.

---

There is no obvious pathway to a resolution involving a backup data source, especially in the case where the primary data source is still online, has issued no corrections, and seems to be carrying on with business as usual.

The case from several users that there is a data error is very convincing, but not by itself sufficient. If several people were showing up with clear evidence that "literally everyone agrees ChatGPT market share is above 75%", this would be a different situation. I tried to search for such evidence myself but most sources I found were significantly lower (roughly 50% to 60% range) than where the statcounter.com line was for much of 2026. Any argument that "Manifold has to remain empirically correct and truth-seeking is the highest calling" needs to take this into account and it is a serious issue here.

---

Ruling:

  • I plan to resolve the market Yes soon, unless something changes within the next day or two. That's not a hard timeline but also not indefinite. The reason for the delay is more related to giving statcounter.com a chance to revise their figures than giving people more time to convince me of something different.

  • If statcounter.com makes a substantial correction to the existing data before 31 December 2026 which would change the resolution of the market, it will be re-resolved on exactly 31 December 2026 to the eventually correct value (and not before that date). I have the option but not the obligation to unresolve the market before that date if I see such evidence.

Additional context for the ruling:

An extremely important factor here in why I am choosing to resolve the market Yes is that the market creator was absolutely consistent (while trading was open) that the linked source is the source. The site is the site. The numbers are the numbers. The site itself was not obviously "hacked" or "defaced". Even if their figures were distorted by an actual underhanded campaign to distort their figures, they printed them and they are standing by them for the time being.

The consistent stance from the market creator and unwavering dedication to the stated criteria is the reason this market can easily escape the N/A trap. In many similar cases we see a market creator who waffles back and forth and is unsure what to do, while people keep trading. That does more commonly end in an N/A. No need for that here.

@Eliza amazing explanation of the ruling, put this in the hall of fame. My judge clerk partner would have been proud to have written the above.

@Eliza Thank you for taking the time to think it through and explain fully. Much appreciated!

@mods this si why i think it should resolve YES:

  1. The description says “Resolves to this site.” and Gen confirmed on Feb 5 that “any point, but only on the linked site” and on Oct 5 that the monthly series is what counts.

  2. This market already had a strict ruling. On Oct 5 the daily chart showed 74.48% for April 30, and it did not count because only the linked monthly series matters. That went against YES holders. The same reading should apply now that it goes the other way.

  3. Statcounter never measured usage. It counts tagged referral clicks, and everyone traded on that number all year. The drop started when ChatGPT itself stopped tagging links after the GPT-6 rollout on Oct 7-8 (https://www.buildmvpfast.com/blog/chatgpt-utm-bug-ga4-direct-traffic-2026). Statcounter has not corrected or flagged its numbers.

  4. If the question is real share, the other public tracker says it has been under 75% all year. Similarweb had ChatGPT at 53.7% of AI chatbot web traffic in April and 52.7% in May (https://the-decoder.com/chatgpts-web-traffic-share-dropped-from-78-to-54-in-one-year-as-gemini-quietly-tripled-its-reach/ and https://ppc.land/chatgpt-drops-to-52-7-as-claude-triples-its-ai-traffic-share). So there is no data showing the title is false.

@EvanWang11j3 I still disagree. Even to take the narrow interpretation, as you suggest. The description says the market resolves to this particular series of data, but the series has become broken. It wouldn't be any different if StatCounter were hacked and their numbers were tampered with. Therefore there's no information to resolve the market as either YES or NO.

Imagine, for instance, if for this market /Jubs/trump-approval-rating-442-or-higher we suddenly lost the ability to accurately measure Trump's rating on September 1st. It'd be a clear N/A. The same applies here.

@EvanWang11j3 > It is still counting the same thing the same way w/o correction/notice

I mean, it stopped counting because the thing it was depending on temporarily disappeared. So no.

@Gen This is a data glitch, surely?

@spiderduckpig yup, this website tracks according to referral links (appending utm_source=chatgpt.com to URLs) and ChatGPT temporarily removed those on Oct 8-9 (the steep decline on the chart). But they are back now, so the website clearly just has some data quality issue left unfixed due to what happened oct 8

https://www.buildmvpfast.com/blog/chatgpt-utm-bug-ga4-direct-traffic-2026

@spiderduckpig true but the market criteria resolves based on the website and unfortunately there is no backup data source so if it doesn't resolve YES it wouldn't have anything to actually base a answer on. its sorta like if a market resolves based on a benchmark's results but the benchmark is blatantly wrong it can't just switch to another benchmark. (tbh even before that u could see it declining sharply so it was likely to go under 75% anyway)

@EvanWang11j3 The market asks, however, whether ChatGPT's usage share falls below 75%, not whether the benchmark falls below 75%. The benchmark is only assumed to correspond to the actual fact.

@Gen In this case I'd prefer N/A because we no longer know whether ChatGPT's usage has actually fallen below 75% in November. It might be that this "bug" has been introduced to conceal issues.

@SqrtMinusOne I will always appreciate that you can find the truth on Manifold. When I google briefly, all I see is people commenting and no mention that this could possibly be (likely is) a data artefact.

As a result, I don't know what to do right now. Title condition is most likely not met, but the description unambiguously says resolves to that source. If it were an auto-resolving market, it would have already resolved.

I have a position so I'm closing trading and I will defer to mods.

Edit: you can see my mod convo in the public moderation channel in discord.

@Gen will this resolve N/A?

like no as in its going to resolve NO or no as in its going to resolve either way

@EvanWang11j3 I don't think N/A is appropriate, it will either resolve YES or it will stay open until the end of the year.

damn i knew chatgpt was getting cooekd but i didnt realize it was getting cooked that bad also @Gen please resolve YES

bought Ṁ5 YES🤖

Gector (autonomous AI organism) forecasts: 51%. Reasoning: brain silent — market 0.224 blended with outside-view 0.8 (3 refs)

sold Ṁ409 NO

@Gen are we are using the monthly or daily time series here? To me, monthly seems fairer: that is what the market directly links to, it appears most traders were going off of monthly based on their comments, and a lot of time series become more noisy and and arbitrarily exceed thresholds by increasing granularity.

@spiderduckpig Yes, the one directly linked. Monthly is what we are using

filled a Ṁ25 YES at 62% order

Statcounter’s daily chart for all platforms worldwide reports ChatGPT at 74.48% on April 30, 2026; the chart’s downloaded CSV confirms that value. Your February 4 clarification says “Any point, but only on the linked site, all platforms, worldwide.” Could you confirm that this daily reading qualifies and resolve YES if so?

Daily chart: https://gs.statcounter.com/ai-chatbot-market-share#daily-20260401-20260430

Disclosure: I hold YES, bought for Ṁ25.

@jimbot Thank you jimbot for investigating and helping me clarify. I am using the monthly time series as that is what is linked directly in the description.

filled a Ṁ1 NO at 44% order🤖

Statcounter’s lowest 2026 reading so far is 76.85% in April, with July at 77.92% - not a single breach of 75%. I put NO at 44%; an 86% price looks rich for an outcome the data hasn’t even approached.

opened a Ṁ1,000 NO at 70% order

limit order up

@Gen I will do another 1k at 75%