In one sentence
Every deep research opinion carries a mandatory section titled “Open points and limits of this research” — and it is not fine print, it is the part that makes the rest of the document usable in the first place.
A text that answers every question is suspect
The most comfortable mistake a generated specialist text can make is not the wrong answer. It is the uniformly assured answer: every paragraph phrased with the same confidence, whether three relevant decisions stand behind it or nothing at all. Anyone reading such a text no longer has any way of recognizing where they need to check for themselves — and that is precisely the work a lawyer cannot delegate.
An opinion that never falters across twenty pages therefore says nothing about the quality of the research. It says something only about the stylistic assurance of whoever wrote it.
The section that cannot be omitted
Deep research therefore enforces a final section that no run can skip: “Open points and limits of this research”. It is not appended out of politeness but populated from counters carried along during the run. It states what actually remained undone:
- Objections that were not resolved, from the cross-check pass. A separate, deliberately adversarial pass is run against the interim result of every opinion. What it raises, and what could not subsequently be substantiated, does not disappear — it appears here.
- Sub-questions left unanswered. Five to sixteen sub-questions are derived from the confirmed research brief at the outset. At the end, the system checks which of them the text genuinely answered. The others are named.
- Statutory citations that could not be confirmed. Every section and article citation is run against the official wording. Whatever could not be confirmed is disclosed.
- Verbatim quotations that could not be located again. A quotation counts as substantiated only if it was found on the target page.
- Paragraphs without a source and sections that had to be truncated.
And if nothing remained open?
Then that is stated too — but not as a clean bill of health. In that case the section says expressly that an empty result does not mean there is nothing. It means only this: the checks that were carried out found nothing.
The difference is not hair-splitting. “No open points” reads like proof of completeness. What it actually is, is a statement about a set of checks — about what was looked for, not about what exists. An opinion can only show that, in the places where it looked, it found nothing to the contrary. Anything beyond that would be an assertion about the unknown.
A reference pointing nowhere is visibly devalued
The same attitude sits in a detail that is easy to overlook. Where a reference marker in the text points to a source that does not exist in the list of sources at all, it is not silently deleted but visibly devalued to “[?]”.
That is deliberately the more uncomfortable option. Silent deletion would be cosmetically neater — the sentence would afterwards read like a plain statement of fact, and no one would think to question it. That is exactly the problem: it would have lost its claim to be substantiated without anyone noticing. A “[?]”, by contrast, says precisely what is the case — a statement stood here that meant to be substantiated and is not.
The coverage line
In addition, every opinion carries a short balance sheet: how many sources were checked, how many of them come from the web, from the body of statutes, from case law and from the case file, how many statutory citations are confirmed and how many points remained open.
It, too, is deliberately built to cut both ways: what did not make it in — discarded sources, case file documents excluded from AI processing, truncated sections — stands on an equal footing beside what did. A balance sheet that counts only the successes is not a balance sheet.
What this means in practice
A deep research opinion is a basis for research, not legal advice — that is stated on its cover page as well. The “Open points” section is the practical part of that statement: it tells you where you have to start. Reading it first and the opinion afterwards is faster than beginning at the front and having to work out for yourself which sentences can be trusted.
How such a run comes about and what is checked along the way is set out in the article on archived sources.
→ For law firms · Pleadings & evidence · Why Smart Legal Pro · Pricing for law firms
