A first-draft competitor scan in 15 minutes

A competitor scan doesn't have to start with a week of desk research. An AI assistant can give you a solid first draft in a quarter of an hour, if you feed it the right material.

The speed is real. In a Harvard Business School study with Boston Consulting Group, consultants using AI completed 12.2% more tasks and finished them 25.1% more quickly, with better quality.¹ But the same study has a warning that matters for competitor work, and this guide is built around it.

Why the AI needs your sources

Asking an AI “what does Competitor X do?” from memory is risky. Its knowledge may be out of date, and it may fill gaps with guesses that sound right. In a 2025 Columbia Journalism Review test, eight AI search tools gave incorrect answers to more than 60 percent of queries, and the researchers found that “Premium chatbots provided more confidently incorrect answers than their free counterparts.”²

OpenAI explains why this happens: “standard training and evaluation procedures reward guessing over acknowledging uncertainty.”³ The fix is simple. Give the AI the material and tell it to stay inside it.

Give it sources

Paste in your competitors' homepage copy, product pages, and two or three recent press releases. Working from material you supply keeps the output grounded.

A good source pack for each competitor, collected in about five minutes:

• Homepage and “about” page. How they describe themselves and who they say they serve.
• One or two product or pricing pages. What they sell and how they package it.
• Two or three recent press releases or blog posts. What they want to be known for right now.
• A handful of public reviews. What customers actually praise and complain about.

Copy the text rather than pasting links, so you know exactly what the AI read. Label each block with the competitor’s name and the date you collected it. Three to five competitors is plenty for a first draft.

Keep it to public material. Don’t log into a competitor’s customer area, sign up under a false name, or paste anything shared with you in confidence. A scan built on public sources is one you can share freely inside the company.

The prompt

“Using only the material below, summarize each competitor's positioning in one sentence, their main claims, their target audience and any gaps you notice. Flag anything you are unsure about.”

Then add three follow-up requests that make the output easier to use:

1. “Put the summary in a table with one row per competitor and columns for positioning, main claims, audience, pricing signals and evidence.”
2. “For every claim, note which document it came from.”
3. “List the three biggest differences between these competitors and the positioning I paste below.” (Then paste your own.)

Asking for the source of every claim makes checking much faster. Asking for uncertainty gives the AI permission not to guess. OpenAI’s own testing shows why that matters: in one benchmark, a newer model that declined to answer when unsure (52% of the time) made errors 26% of the time, while an older model that almost always answered made errors 75% of the time.³

A useful first draft might include a row like this, for an imaginary competitor: “Positioning: the simplest scheduling tool for small clinics. Main claims: set up in a day, no training needed (homepage). Audience: practice managers. Gap: no mention of integrations, which our customers ask about. Unsure: pricing page not supplied.” Short, sourced, and honest about what it doesn’t know.

Notice the “Unsure” line. It is the most useful part of the output, because it tells you exactly where to look next. If the AI never flags anything as unsure, that is a warning sign in itself: either the material was unusually complete, or the AI is filling gaps without telling you.

Coming soon: Prompts that work: a brief template for marketers →

What to double-check

Verify every claim against the source, and treat the gaps as hypotheses, not findings.

This step is where human judgment earns its keep. In the Harvard study, consultants given a task chosen to sit outside AI’s strengths were 19% less likely to reach the right answer when they used AI.¹ The researchers call this the “jagged technological frontier”: AI is excellent at some tasks and misleading on others that look just as hard.¹

A competitor scan has parts on both sides. Summarizing what a competitor says about itself sits comfortably inside the frontier. Judging what that means for your strategy, whether a gap is real or a deliberate choice, often does not. So:

• Check the facts. Each claim should match the document it cites. Delete anything that doesn’t.
• Question the gaps. A gap the AI spots may simply be missing from the material you supplied.
• Add what you know. Sales conversations, lost deals, and industry gossip are not in any press release.

Using the draft

A first-draft scan is a conversation starter. Share the table with the team and ask two questions: what surprises you, and what is missing? The answers usually say more about your own positioning than about the competitors.

Save the source pack and the prompt. Re-running the same scan every quarter, with fresh material, shows how competitors’ messages shift over time, which is often the most useful insight of all.

Where the 15 minutes go

About five minutes to collect sources, two to run the prompts, and the rest to check the output. That balance is deliberate. The time savings from AI are real: in an MIT experiment with 453 professionals, ChatGPT cut the time spent on writing tasks by 40 percent while quality rose by 18 percent.⁴ But the time saved on drafting is best spent on checking.

If you find yourself skipping the check because the draft looks polished, slow down. Polished is not the same as correct, and the CJR test found that paid AI tools could be more confidently wrong than free ones.² When the scan feeds a pricing decision or a board slide, budget an extra half hour to confirm the key facts by hand.

Coming soon: Hand over your first multi-step task →