Research and results
Is the AI making this up?
No. Every claim in the plan links to a stored record, and a fact the research did not find shows as a marked placeholder rather than a guess.
Large language models read the pages, judge the keywords and write the briefs and drafts. They are good at it, and they can also be wrong, so the product is built around one rule: every claim must trace to a stored record.
What that means in practice
- A page review quotes the page. The quotes are checked against the fetched text character for character; a quote that is not in the page is thrown away.
- A gap is only a gap if the page reviews and the results data show it.
- A feasibility call lists the observations it relied on, with the source of each.
- A customer-language phrase is only kept if its quote is verbatim from the source and its volume was measured; otherwise it is marked unverified.
- A draft that needs a fact the research did not contain shows a marked placeholder for you to fill, never an invented figure, name or date.
Everything a decision relied on is one click away in the Evidence panel. See How do I open the evidence?.
What it does not mean
The judgments themselves, whether a gap is real, whether you can win, what to build, are opinions formed from the evidence. Good ones, formed the way a careful strategist would, but opinions. Check a recommendation against what you know about your business, and read a draft before you publish it.
If you find a claim without evidence
Tell us with the site and the page. We treat it as a defect and fix the cause, not just the instance. See the Editorial Policy.
Was this helpful?
Still stuck?
Tell us what you were trying to do and which site or page it concerns. A person replies from help@legiit.com, usually within one working day.