Scope note: This cheatsheet reflects observed behavior in ChatGPT models. It does not generalize to all LLMs.
Stability note: Model names, limits, and UI signals change over time. Treat specific numbers and indicators as indicative, not contractual.
With the release of GPT-5 in August 2025, OpenAI adjusted usage limits and internal reasoning behavior. The guidance below focuses on increasing the likelihood of deeper internal reasoning in general (non-coding) projects without unnecessarily consuming manual Thinking quotas.
ChatGPT (web/mobile) – observed usage tiers
| Tier | GPT-5 Standard | GPT-5 Thinking (deeper reasoning) |
|---|---|---|
| Free | ~10 messages / 5 hours; then fallback to smaller model. ~1 Thinking message/day. | ~1 Thinking message/day. |
| Plus | ~160 messages / 3 hours (temporarily elevated at time of review). | ~200 manual Thinking messages/week. |
| Pro/Team | Effectively unlimited standard usage (subject to abuse guardrails). | Access to extended/pro Thinking variants. |
Key distinction
- Manual Thinking requests → counted toward weekly limits.
- Automatic internal escalation → typically not counted.
When a standard request internally escalates, you gain depth without explicitly spending a manual Thinking slot.
Heuristics that often correlate with deeper internal reasoning
These are correlations, not guarantees:
- Slower responses, even for concise prompts.
- Structured output with explicit comparisons, trade-offs, or dependency analysis.
- Occasional UI labeling indicating deeper reasoning after the fact.
You cannot reliably detect or force escalation; the goal is to increase probability, not exert control.
1) Multi-step comparison
Template
“Compare [Option A] and [Option B] across [Factor 1], [Factor 2], and [Factor 3] over [timeframe]. Identify dependencies between factors and note edge cases where the conclusion changes.”
2) Scenario evaluation
Template
“Given [context/background], evaluate outcomes under Scenario 1, Scenario 2, and Scenario 3. Highlight trade-offs, failure conditions, and decision inflection points.”
3) Cross-domain reasoning
Template
“From the perspectives of [Domain 1] and [Domain 2], analyze [problem/topic]. Identify agreements, disagreements, and areas of synergy or tension.”
4) Contradiction and assumption check
Template
“Examine the following statements:
- [Statement A]
- [Statement B] Determine whether they contradict, align, or are orthogonal. Surface hidden assumptions that would invert the conclusion.”
5) Root cause and mitigation
Template
“Given [problem description], rank likely root causes by impact and probability. Propose mitigations for each and indicate what additional data would change the recommendation.”
6) Decision framework
Template
“For [decision], evaluate options using criteria [C1], [C2], and [C3]. Score each option on a consistent scale and justify the scoring. Identify threshold values where the preferred option changes.”
Phrases that often nudge deeper reasoning
- “Compare and rank alternatives with explicit trade-offs.”
- “Evaluate multiple scenarios and identify failure modes.”
- “Surface hidden assumptions that could change the outcome.”
- “Score options using consistent criteria and justify.”
- “Identify inflection points where the decision flips.”
- “Explain why a reasonable person might choose the opposite.”
When not to trigger deeper reasoning
Avoid explicit depth when:
- The task is purely informational or definitional.
- A known, canonical answer exists.
- You are brainstorming broadly without commitment.
In these cases, depth adds latency without improving signal.
When to spend manual Thinking
Use it deliberately when:
- Decisions involve long time horizons or compounding effects.
- Multiple domains interact (financial, technical, social, legal).
- The first response is shallow, contradictory, or hand-wavy.
- You need explicit reasoning chains to defend a decision.