"Do a competitor analysis" — the prompt that returns a Wikipedia summary
Why an open-ended competitor analysis prompt returns generic profiles, and a version scoped to a decision you actually need to make.
The prompt, as typed
Do a competitor analysis for my company. Be thorough and detailed.
What it leaves the model to guess
- Which competitors, and on which dimensions
- What decision the analysis is meant to inform
- How to handle facts the model cannot verify
- Output format for comparison
Graded before and after
Both versions below were scored by the same rubric the free prompt grader runs.
As typed
5/100 · grade F
| Clarity of context | 2 |
|---|---|
| Structure | 0 |
| Specificity | 0 |
| Reusability | 0 |
| Decision logic | 0 |
Rewritten
90/100 · grade A-
| Clarity of context | 3 +1 |
|---|---|
| Structure | 3 +3 |
| Specificity | 2 +2 |
| Reusability | 6 +6 |
| Decision logic | 5 +5 |
The rewritten prompt
Compare the competitors listed below on the dimensions listed below.
Competitors: {{competitors}}
Our product: {{our_product}}
Decision this informs: {{decision}}
Dimensions: {{dimensions}}
Output a table, one row per competitor, one column per dimension, plus a final
column "confidence" marked high / medium / low.
Rules:
- Mark any cell you cannot support as "unknown" rather than estimating
- Do not use marketing language from their own sites as fact
- After the table, write three lines: where we are genuinely differentiated,
where we are at parity, and which single gap most affects {{decision}}
If you do not have reliable information on a competitor, say so and leave the
row blank rather than filling it in. Why the rewrite works
The "unknown" instruction is the whole point. An unscoped research prompt rewards confident-sounding filler, because nothing in the prompt makes admitting ignorance an acceptable answer. Making it acceptable is what turns the output into something you can act on.
Common questions
- Why does this prompt score badly?
- It scores 5 out of 100 (F) because it leaves the model to decide 4 things you already know: which competitors, and on which dimensions; what decision the analysis is meant to inform; how to handle facts the model cannot verify; output format for comparison.
- Does a longer prompt always score higher?
- No. The rubric rewards named inputs, a fixed output format and an explicit rule for choosing between options. Extra prose without those adds length and no score.
- Will this rewritten prompt work in ChatGPT and Claude?
- Yes. It is plain text with placeholders you fill in before running, so nothing in it is specific to one model or vendor.