AI Chat · OpenAI
✓ Verified Jun 2026
Last audited: 2026-06-29
ChatGPT:
What the pricing
page doesn't tell you
Everything the pricing page doesn't tell you. Based on public docs, community reports, and third-party benchmarks — not legal advice or official statements.
Hidden Behaviors
documented & verified
Transparency
31%
of key policies disclosed upfront
Last Verified
Jun 2026
$20/mo · ChatGPT Plus
Hidden Behaviors
5 documented
Behaviors confirmed through user reports, help docs, and independent testing — none disclosed on the pricing page.
No visible usage counter — limit discovered on hit
▾
Official
No usage dashboard existsOpenAI confirms there is no consumer-facing counter for Plus plan messages.
Observed (community)
Wall appears mid-sessionUsers report messages stop without warning — only a limit popup appears after the fact.
Estimated (not verified)
~150–160 msgs / 3h (community-reported)Inferred from help docs; no official number published on pricing page.
Reported: model downgrade when cap is reached
▾
Official
Not disclosedPricing page and plan comparison make no mention of automatic model downgrade behavior.
Observed (community)
GPT-4o mini substituted silentlyOnce the Plus cap is reached, responses continue but from a smaller model. No modal, no banner.
Estimated (not verified)
majority of users reportedly unawareCommunity surveys suggest most users attribute lower response quality to "the AI getting worse," not a model switch.
Reported: burst usage may extend your wait time
▾
Official
Rolling 3h per-messageEach message carries its own 3-hour timer from send time. Disclosed in help docs, not on pricing page.
Observed (community)
Burst = full 3h waitSending 80 messages in 10 minutes means you wait ~3h for any slot to return — unlike spacing them out across the window.
Estimated (not verified)
Disproportionate for power usersAffects users who batch-send during deep work sessions — coding, research, writing sprints.
Deep Research capped at reportedly ~10 runs/month on Plus
▾
Official
Not featured on pricing pageChatGPT Plus marketing highlights Deep Research as an included feature without specifying monthly limits.
Observed (community)
reportedly ~10 runs/month confirmedUsers and OpenAI support threads confirm the ~10 monthly cap. Heavy researchers exhaust this in 2–3 days.
Estimated (not verified)
Pro plan offers moreChatGPT Pro ($200/mo) includes higher Deep Research allowance — the effective upsell trigger for power users.
Free-tier conversations used for model training by default
▾
Official
Opt-out requiredOpenAI Privacy Policy confirms free-tier data used for training by default. Must disable via Settings → Data Controls.
Observed (community)
Setting not surfaced at signupThe opt-out exists, but new users are not prompted to review it during onboarding. Default is training-on.
Estimated (not verified)
Plus/Team off by defaultPaid plans have training off by default. Free users who upgrade retain their prior opt-in unless manually changed.
Real Limits
Reset Cycle
Rolling 3-hour window (per-message timer, not a fixed clock)
Stated Cap
~150–160 messages / 3h (reported) — listed in help docs, not the pricing page
Usage Meter
None. No dashboard, no counter, no warning before the wall.
Response Speed
GPT-5.5 high-reasoning: 20–30s time-to-first-token (benchmark range) (Artificial Analysis, mid-2026)
Price History
$20/mo since Feb 2023 — most stable price in this category
⚠ Hidden Catch
Each message carries its own 3h timer. Burst 80 messages in 10 minutes and you wait nearly the full 3h for any slots to return — unlike spacing them out.
Transparency Breakdown
How much OpenAI discloses about each dimension — before you pay.
Real Headroom
~150–160 msgs / 3h (reported) rolling. No usage meter — you only find out when a popup blocks you.
disclosed in help docs · not on pricing page
Response Speed
GPT-5.5 (high): 20–30s time-to-first-token (benchmark range). Not disclosed anywhere on the product or pricing pages.
third-party benchmark only
Limit Transparency
Zero usage dashboard. No counter. You discover your limit only after hitting a wall — per community reports.
score: 2/10 — lowest in category
Data & Privacy
Free-tier trains on conversations by default. Opt-out exists but not surfaced at signup. Paid plans off by default.
requires manual action to change default
Reliability
3 incidents / 30 days — ~~99% uptime (est.). Status page is public and up to date.
best-disclosed dimension
Overall Transparency Score ⓘ HOW?
Weighted estimate — see methodology for formula
31%
Policy History
Every confirmed limit change at the same price — documented with source type and dates.
Feb 2023
ChatGPT Plus launched at $20/mo — GPT-4 limited to 40 messages/3h
OFFICIALhelp.openai.com release notes
2024
Cap quietly raised to ~80 msgs/3h, then 150 msgs/3h — no public announcement
OBSERVEDcommunity-reported; no official announcement found
Feb 2026
Ads introduced on Free and Go tiers in the US — Plus and above remain ad-free
OFFICIALopenai.com/blog/ads-in-chatgpt
Feb 2026
GPT-4o, GPT-4.1 and older models fully retired — forced migration with short notice
OFFICIALhelp.openai.com/retiring-gpt-4o
Apr 2026
Message caps silently tightened — a subset of heavy users affected (est. ~5–10%), no official notice given
OBSERVEDcommunity-reported via OpenAI Developer Forum, April 2026
Who It's For
Same data, different verdict depending on how you use it.
✓ Works well for
Daily work & productivity
Stable $20 price since 2023. Broad ecosystem with GPTs, image gen, voice.
One-tool users
Widest feature set of any AI subscription at this price point.
Image & video creation
DALL·E and Sora included — no other $20 plan matches this.
✗ Problematic for
Burst power users
No usage meter — you hit the wall with zero warning, mid-session.
Privacy-conscious users
Free-tier conversations used for training by default. Must manually opt out.
Slow-connection users
GPT-5.5 high-reasoning mode: ~20–30s to first token (estimated range).
Compare Transparency
ChatGPT
31% transparent · viewing now
Claude
view report →
Gemini
view report →
Perplexity
view report →
Frequently Asked
Does ChatGPT show how many messages I have left?
No. As of June 2026, OpenAI provides zero usage dashboard for consumer plans. You only learn your limit when you hit it.
When does ChatGPT Plus's limit actually reset?
It's a rolling 3-hour window per message, not a fixed clock. A message sent at 2:00 PM frees its slot at 5:00 PM — not at a round hour. This means burst usage is disproportionately penalizing.
Has ChatGPT Plus ever raised its price?
No. It has stayed at $20/month since February 2023 — the most stable price among major AI subscriptions in this database.
Is GPT-5.5 slow?
Time-to-first-token is around 20–30 seconds on the high-reasoning setting — among the slowest of major assistants for that mode, per Artificial Analysis benchmarks from mid-2026.
Does ChatGPT train on my conversations?
Free-tier conversations are used for training by default. You can opt out via Settings → Data Controls. ChatGPT Plus and Team plans have training off by default.
Get alerted when policies change
We track silent limit cuts before most users notice them.