You have a 4 pm meeting and a partner who wants the competitive landscape “in broad strokes — but accurate.” Three years ago that meant a panicked afternoon and 30 open tabs. Today, a Perplexity-plus-Deep-Research loop will give you a defensible five-competitor teardown in 30 minutes, with citations you have actually opened. This tutorial walks the loop end-to-end and flags where AI still gets pricing, headcount, and feature parity wrong.
What this covers
A repeatable 30-minute workflow: pick the right comparison axes for your decision, pull five candidate competitors with Perplexity, validate them with one Deep Research pass, dump everything into a comparison sheet, then spot-check the two cells you would stake the meeting on. The output is a single page your partner can read in 60 seconds.
Who this is for
PMs writing competitive slides before a roadmap review, founders sanity-checking a pitch deck the night before, strategy and BD teams scoping a new vertical, and analysts who normally bill the first three hours of a project to “tab juggling.” Not for: long-form market reports — for those, see the AI industry research workflow.
When to reach for it
When you have one short meeting, a partner or investor who will ask sharp questions, and five competitors who are mostly searchable on the web. Skip this workflow when the competitors are stealth-mode startups, when most data lives behind paywalls (legal, medical), or when you actually need a 12-page market brief — that is a different tool.
Before you start
- Write the decision the teardown supports in one line. “Should we ship feature X in Q3?” is a decision; “tell me about competitors” is not. The decision picks the axes.
- Pick 4-6 comparison axes ahead of time: pricing, ICP, distinct feature, distribution, recent funding, team size. Decide which two matter most for the call.
- Open a fresh spreadsheet or table. Rows are competitors, columns are axes, one cell per fact with a link.
- Pick your engines: Perplexity for fast cited answers, ChatGPT or Claude Deep Research for the synthesis pass. Both, not either.
Step by step
- Frame the competitor question in Perplexity. Open perplexity.ai, switch to Pro Search if you have it. Ask “Who are the top 5 competitors to [your product or thesis] as of 2026, ranked by [traction / revenue / mindshare]?” Read the answer; click two citations to confirm the list is not stale.
- Lock the five names. Drop anyone defunct, pivoted away, or acquired into a different segment. If Perplexity offers eight names, pick the five most relevant to your decision — not the five with the loudest marketing.
- Run one Perplexity query per axis, all five competitors at once. “Pricing for Competitor A, B, C, D, E as of 2026 — list each one with currency and tier.” Repeat for ICP, distinct feature, and last funding round. Five queries, four axes, twenty data points. For an example of cited-search workflow ergonomics, Perplexity basics covers Focus modes and Collections that pay off here.
- Drop every answer into the sheet with the source URL. Do not paraphrase. If Perplexity returned “$49/seat/month, billed annually,” that goes in the cell verbatim. The URL goes in the next column. Empty cells stay empty — that signals “unknown,” which is honest.
- Run one Deep Research pass for synthesis. ChatGPT or Claude. Paste the filled sheet as context. Ask “Given these five competitors and four axes, what are the three angles where my product can win, and the two angles where I am structurally behind?” This is the only step where AI writes prose for you.
- Spot-check the two cells your partner will ask about. Usually pricing and recent funding. Click the source, Cmd+F the number, confirm it is not from a 2023 blog quoting a now-dead pricing page.
First-run exercise
Pick a product you already know well — your own, or a tool you use daily. Run the full 30-minute loop. Score every cell: green if right, yellow if approximately right, red if wrong. Most first runs come back with 75 percent green, 20 percent yellow, 5 percent red — and the red cells are almost always stale pricing or feature claims pulled from a 2024 article. That is the failure mode to watch for on every future run.
Quality check
- Every cell has a source URL. Cells without sources are deleted, not guessed.
- The two cells your partner will ask about have been clicked through and verified.
- No competitor has been silently dropped because the AI could not find data. “Unknown” is a valid answer.
- The synthesis paragraph names a specific angle, not a generic “differentiation opportunity in the mid-market.”
- You can read the whole sheet in 60 seconds. If not, you have too many axes.
How to reuse this workflow
- Save the four axes plus the synthesis prompt as a template. New vertical, new five names, same spine.
- Build a personal blocklist of stale sources Perplexity loves to cite (specific Medium accounts, AI-generated SEO sites). Pin them in the prompt as “do not use.”
- Re-run the teardown quarterly on the same sheet. Diffing two quarters tells you who actually shipped versus who only marketed.
Recommended workflow
Decision in one line → axes locked → Perplexity for the list → Perplexity per axis → fill the sheet with URLs → Deep Research synthesis → spot-check two cells → one-page summary. The full loop is 30 minutes the first time, 15 minutes once you have a template.
Common mistakes
- Skipping the decision sentence — every axis looks relevant, the sheet bloats, the meeting drifts.
- Trusting Perplexity’s competitor list without confirming the companies still exist as described.
- Filling cells from memory because “I think their pricing is around X” — that is the cell your partner will catch.
- Asking Deep Research for the competitor list — slower than Perplexity and no better.
- Treating the synthesis paragraph as the deliverable — the sheet is the deliverable; the paragraph is the lede.
- Forgetting the empty cells. Three “unknown” cells across five competitors is a finding, not a gap.
FAQ
- Why Perplexity instead of just ChatGPT?: Inline citations. You can audit the sheet in two minutes; with ChatGPT you cannot tell which numbers were retrieved versus invented.
- Do I need Pro on either tool?: Pro Search on Perplexity helps for the per-axis queries. Deep Research requires a paid tier on ChatGPT or Claude. For a one-off teardown, the free tiers usually suffice.
- What if a competitor has no public pricing?: Leave the cell empty and write “no public pricing” in a notes column. Do not let AI invent a number.
- Can I do more than 5 competitors?: Yes, up to 8 before the sheet becomes unreadable in a 60-second scan. Beyond 8, switch to the AI industry research workflow.
- How fresh is “fresh enough”?: Pricing and funding must be 2025 or 2026. Feature claims tolerate 12 months. Anything older needs a click-through.