Competitor Intelligenceis a free tool that researches your company and up to three competitors, live, and turns the evidence into a sales-ready competitive report: battle cards, objection responses, attack angles, a feature parity matrix, pricing intelligence, and a playbook for displacing each competitor. Every insight is tied to what was actually found on the companies' websites, in recent market news, and in customer reviews.
This documentation explains how that report is produced: what the research phase scrapes and searches, how the research confidence score is computed, what every report section means, and what happens when parts of a run fail.
Why evidence-first competitive analysis
Generic AI competitor comparisons are written from training data: often stale, sometimes invented. This tool inverts that. It first collects real, current evidence (up to four pages per company, four categories of market signals per company from the past year, and review sentiment if you provide a review URL) and only then asks the model to analyze it. Where the evidence is thin, the report says so: individual claims carry high, medium, or low confidence labels, and the whole report carries a research confidence score that tells you how much was actually verified.
What a run gives you
- The competitive picture: a snapshot grid of every company (tagline, pricing model, ICP, top features), three "where you're losing" findings with evidence, and three blind spots competitors cover that you don't.
- Sales ammunition: a full battle card (win and lose scenarios, objection responses, landmine questions, positioning lines), 3 to 5 fast attack angles, and 3 to 5 win themes phrased as copy-paste talking points tagged for sales calls, landing pages, cold email, or demos.
- Strategic depth: messaging analysis with the angles competitors are not claiming, review sentiment across competitors, a feature parity matrix, pricing intelligence with price objection responses, and a displacement playbook per competitor.
- Sources and a share link: the market signal sources behind the analysis are listed in the report, and every report gets a link you can share with your team, plus PDF, image, and CSV export.
How a run works, at a glance
| Step | What happens |
|---|---|
| 1. Resolve | Each company you entered (as a name or a URL) is resolved to a verified name and domain. |
| 2. Research | Up to four pages per company are scraped (homepage, about, pricing, features), four categories of market signals are searched per company, and your review URL is analyzed for sentiment. |
| 3. Analyze | One large model call turns the full research payload into all eleven report sections, under rules that forbid fabricating data points. |
| 4. Validate | The output is checked against a strict schema. Output that does not validate is retried once with the exact error; a run that still fails is marked failed honestly instead of shipping a broken report. |
The details live in How reports are generated and Reading the report.
Running your first analysis
- Open Competitor Intelligence.
- Enter your company as a website URL if you can. A URL resolves directly to your real pages; a bare name has to be looked up first, which is less reliable.
- Add one to three competitors, again preferring URLs. One primary competitor is enough; each additional competitor adds cross-competitor patterns to the messaging and review analysis.
- Optionally paste a review page URL (a G2, Capterra, or Trustpilot listing for a competitor works well). This unlocks the review sentiment section with real customer complaints and churn signals.
- Optionally add a sentence of context about how you compete, for example "we win on price, they win on brand". It is passed to the analysis verbatim.
A run typically takes one to two minutes: most of it is the live research, the rest is a single large analysis call. Identical inputs within 24 hours return the same stored report instead of burning a fresh run.