According to Similarweb data published on October 18, 2026, the most visited X profile in September 2026 by web traffic share was @thsottiaux — Thibault Sottiaux, who oversees ChatGPT, Codex, and the API at OpenAI. He surpassed football journalist Fabrizio Romano and Elon Musk, rising from fourth place to first over three months: this engineer's profile has become the primary hub where the community checks news about ChatGPT and Codex limit resets.

What Happened

Similarweb published a ranking of the most visited X account pages for September 2026 on October 18, 2026. The @thsottiaux account took first place in web traffic share, followed by football journalist Fabrizio Romano in second and Elon Musk in third. Sottiaux's trajectory has been steady: he ranked fourth in July, second in August, and first in September. According to X statistics, he published 83 limit reset announcements over the year, 60 of which concerned full resets, prompting users to create unofficial websites to track such announcements.

Context

The Similarweb ranking is based on the share of social network web traffic directed to an account's page in a browser, excluding the X app. This measures the frequency of profile visits, not post reach or follower count: Sottiaux has approximately 893,000 followers, while Musk has 240 million, making the gap in page visits even more notable. Sottiaux is responsible for ChatGPT, Codex, and the API at OpenAI and has become the de facto news channel for usage restrictions: quotas and limits on paid plans cause regular friction, and no official source for availability status data is mentioned in the material, so unofficial reset trackers emerged in response to this deficit.

Why This Matters for the Industry

For the AI industry, the ranking is an indirect operational signal: in September 2026, attention to OpenAI products concentrated on quotas and limit resets rather than new models, meaning access to computational resources became the bottleneck determining user experience and company reputation. Information about limits is becoming a standalone product space: demand for monitoring is already confirmed by user behavior, so builders can assemble an MVP — tracking reset announcements with alerts in Slack or Telegram and integration with calendars or CI — but the project must be designed for degradation from the start, because the data source is unofficial and any change in publication format will break it. Simultaneously, provider limits should be treated as a first-class architectural parameter: implementing own queues and throttling, retries with exponential backoff, aggressive caching of results, fallbacks to other models or providers, and setting honest expectations for users.

Why This Matters for Users

If you pay for ChatGPT or Codex, this case explains why limits feel so acute: reset messages are published dozens of times a year, so they should be perceived as a regular, not exceptional, event, and users should be aware of unofficial trackers in advance. Capacity planning, deadlines, and promises to clients should not be built without accounting for provider quota volatility, and tracker data cannot be blindly trusted until OpenAI has an official way to check remaining limits or availability status. A separate lesson concerns metrics: profile page traffic diverges from follower count, making it unreliable to judge an account's significance solely by followers.

What Is Unknown / Limitations

Causality is not established: the correlation between profile traffic growth and reset announcements does not prove that operational restrictions, rather than model releases, are the main driver of attention to OpenAI; comparable data on traffic around official model announcements would be needed for such a conclusion. The metric itself is a proxy for news-checking frequency: it counts the share of web traffic to the profile page excluding the X app, and drawing conclusions about reach or account influence based on it is methodologically incorrect. Sources provide ranking data for only three months (July–September 2026), there are no absolute traffic values, and the statistic of 83 resets obtained from X has not undergone independent verification; the reliability of unofficial tracker data is also unconfirmed.

Sources

Author

Look at AI, editorial team