Blog — No. 08

ChatGPT in Siri underperformed. Integration still isn’t adoption.

OpenAI told a court ChatGPT in Siri underperformed. Platform AI integration isn’t adoption — opt-in friction and who owns the action after the tap.

In brief
  • MacRumors (23 September 2026) reports OpenAI court filings describing Apple’s ChatGPT-in-Siri / Apple Intelligence integration as “dramatically underperforming” and “persistently underperforming.”
  • Per that press summary: ChatGPT was added to Siri in December 2024; users faced a multistep opt-in; by January 2025 OpenAI called the start slow and cut its forecast for incremental logged-in weekly active users from the partnership — no invented percentages here.
  • Platform partnership headlines measure distribution and deal optics. They do not prove adoption, permission clarity, or return visits.
  • Shipper filter unchanged: integration ≠ who owns the action after the tap. Same line as Muse and JetBrains Air.
  • I have not used ChatGPT through Apple Intelligence / Siri as a product under review. This is not a product test.

OpenAI’s filing says the Apple integration underperformed. That is a partnership scoreboard, not a shipping scoreboard. I’ve shipped FTNS.ai, KnowCap, Conquest, and 3D Trading Chart, and I founded Cryudine Group, an AI agency for U.S. SMBs — enough times through install and return to know the hard part is still who owns the action after the tap, and whether anyone comes back. I have not used ChatGPT through Apple Intelligence or Siri as a product under review. This is not a product review.

What the filing and press actually say (not a product trial)

On September 23, 2026, MacRumors reported on OpenAI court documents in its legal fight with SpaceXAI (xAI when the suit was filed). Per that article, OpenAI described Apple’s ChatGPT integration as “dramatically underperforming” and later as “persistently underperforming.”

Attributed timeline from that press summary — not my measurement:

  • Apple added ChatGPT to Siri in December 2024.
  • Users had to complete a multistep opt-in.
  • By January 2025, OpenAI called the integration “off to a slow start” and cut its forecast for incremental logged-in weekly active users expected from the partnership.
  • The filing argues foreclosure was de minimis / minimal usage in OpenAI’s framing; specific foreclosure shares in the article are redacted — I am not inventing them.

I am attributing that map. I am not claiming I measured Apple Intelligence usage or OpenAI WAUs.

The shipper filter — integration isn’t adoption

When I ship, the filter is blunt: after the tap, who owns the action? Did the user understand what would happen next? Did they have to fight settings to turn the thing on? Do they come back after the partnership screenshot cools?

A platform AI integration can win a press cycle and still fail that filter. Off-by-default and multistep opt-in are not footnotes. They are the product. If the path to value is buried in Settings, you did not ship adoption. You shipped a checkbox with a keynote slide.

Partnership PR is distribution news. Adoption is whether the action completed cleanly and whether anyone returned.

Permission and friction beat partnership headlines

Agent and assistant surfaces live or die on permission clarity and friction. Connectors, defaults, and what happens without another tap matter more than co-branding.

Chat wrappers can hide behind “just a suggestion.” A Siri-path assistant that reaches into AI features across a phone cannot. If the user cannot reconstruct what was offered, what was opt-in, and what was done, you did not prove adoption. You proved a logo placement.

That is the same ownership problem I wrote when Muse topped a free chart and when JetBrains unified agent surfaces: distribution and orchestration are not the hard part. See Muse hit #1. That isn’t the hard part of shipping an app. and JetBrains Air unifies agents. It still doesn’t own the merge. Optional desk continuity on permission vs button: Jev grades decisions. It still doesn’t get the button.

Continuity — same ownership line as Muse and JetBrains Air

Charts are not a new filter. Unified agents are not a new filter. Platform AI partnerships are not a new filter.

Muse measured distribution; shipping still meant who owns the action after the tap. JetBrains Air unified orchestration; that still is not merge authority. OpenAI says ChatGPT in Siri underperformed. Integration headlines still are not adoption.

Different launches. Same line. I have not used ChatGPT through Apple Intelligence / Siri as a product under review.

Key takeaways

  • MacRumors (23 September 2026) reports OpenAI court filings calling ChatGPT-in-Siri / Apple Intelligence “dramatically” and “persistently” underperforming — attributed press, not my metrics.
  • Press timeline: Dec 2024 add; multistep opt-in; Jan 2025 slow start and cut forecast for incremental logged-in WAUs — no invented percentages.
  • Platform integration measures deal optics and distribution path. Adoption is permission clarity and return visits.
  • Shipper filter unchanged: integration ≠ who owns the action after the tap. Same Muse / JetBrains Air line.
  • I have not used ChatGPT through Apple Intelligence / Siri as a product under review. Not a product test.

FAQ

Did ChatGPT in Siri really underperform?

Per MacRumors’ September 23, 2026 summary of OpenAI court filings, OpenAI described the Apple integration as “dramatically underperforming” and “persistently underperforming.” That is OpenAI’s filing language via press — not a measurement I made.

Have you used ChatGPT through Siri or Apple Intelligence?

No. I have not used ChatGPT through Apple Intelligence / Siri as a product under review. This is not a product review.

What have you shipped?

FTNS.ai, KnowCap, Conquest, and 3D Trading Chart. I also founded Cryudine Group, an AI agency for U.S. SMBs.

Does a platform AI partnership equal adoption?

No. Partnership and integration headlines are distribution and deal optics. Adoption is whether users can find the feature, understand the next action, and come back after the announcement cools.

Who is writing this?

Caleb Weinberger — founder of Cryudine Group; shipped work includes FTNS.ai, KnowCap, Conquest, and 3D Trading Chart.

Disclaimer

This article is for educational and informational purposes only. It is not investment, trading, or financial advice. Nothing here is an offer, solicitation, or recommendation to buy or sell any digital asset or security. Trading digital assets involves substantial risk of loss. Views are the author’s alone and do not represent Stillman Digital, Verify Block, Cryudine Group, or any other affiliated entity.