Different AI tools have meaningfully different strengths. Using the wrong tool for a task produces worse results — not just marginally worse, but categorically wrong outputs. Matching the tool to the job is a foundational habit for effective AI use.
AI is not one thing. "Using AI" is not a single skill — it's a family of skills tied to specific tools built for specific purposes. Treating all LLMs as interchangeable leads to frustration and mediocre outputs.
"If you're looking for answers, talk to Perplexity. If you're looking for creativity, Perplexity is nothing."
— Mark Hope, AI Training Session (Part 2)
Best for: Questions with knowable, verifiable answers. Anything where accuracy and sourcing matter.
How it works: Perplexity restricts itself to recognized sources and always provides citations. It is highly unlikely to hallucinate or fabricate information.
Use when:
- You need a factual answer (e.g., "How did Franz Liszt die?")
- You need to cite a source
- You're doing background research on a client, industry, or topic
- You want to verify a claim
Avoid when: You need creative output, brainstorming, or generative ideation — Perplexity is not built for this.
Best for: Ideation, brainstorming, creative copy, strategy generation, and [1] prompts.
Characteristics: Claude is more willing to explore unconventional, edgy, or low-probability outputs. It responds well to complex, structured prompts and tends to produce more varied creative results than ChatGPT.
Use when:
- Generating brand slogans, ad copy, or positioning concepts
- Running verbalized sampling prompts for tail-distribution ideas
- Exploring market scenarios or strategic possibilities
- You want genuinely surprising or unconventional outputs
Note: Mark Hope uses Claude as his primary tool for creative and strategic work, averaging 4–6 hours/day across AI tools.
Best for: General-purpose tasks where a reasonable, mainstream answer is acceptable.
Characteristics: ChatGPT tends toward more conservative, high-probability outputs. It is less likely to produce edgy or unconventional ideas even when prompted. For factual queries, it may hallucinate without flagging uncertainty.
Use when:
- You need a capable general-purpose assistant
- The task doesn't require deep creativity or verified sourcing
- You're already embedded in the OpenAI ecosystem
Avoid when: You need sourced facts (use Perplexity) or genuinely unconventional creative output (use Claude).
| Need | Tool |
|---|---|
| Verified facts with sources | Perplexity |
| Creative ideation, brainstorming | Claude |
| Tail-distribution / verbalized sampling | Claude |
| General-purpose assistance | ChatGPT |
| Conservative, mainstream outputs | ChatGPT |
| Client research, industry background | Perplexity |
Documented from the Asymmetric internal AI training session "Using AI Part 2: Verbalized Sampling," led by Mark Hope. See also the verbalized sampling prompt document distributed to the team following that session.