Main Comments and Replies Are Different Evidence

A flat count hides the conversation

A page can contain 100 visible records without containing 100 independent opinions. Some records start a thread; others clarify, disagree with, or answer the original comment. Flattening both into one column removes the relationship that gives the reply meaning.

For product research, the distinction affects every later step: issue frequency, feature-request grouping, sentiment interpretation, and the evidence used to support a decision.

What our platform tests showed

In one Google Maps validation case, 77 main comments and 75 replies were available on the page. A final task capped at 100 records produced 77 main comments plus 23 replies. The count was correct only because replies were included in the same requested total while their hierarchy remained explicit.

In a separate Bilibili validation, 20 main comments and 23 replies were collected with parent identifiers. The useful evidence was not merely 43 rows—it was the ability to tell which rows started a discussion and which rows belonged beneath them.

A safer export contract

A defensible comment export should make the following visible instead of implying completeness:

• whether a record is a main comment or a reply;
• the parent relationship when the page exposes it;
• whether replies were included in the requested limit;
• whether the task stopped with a partial result;
• the source URL needed to revisit the evidence.

What this does not prove

A successful test on one page does not establish a universal safe limit or guarantee that every thread can be expanded. Page structure, login state, regional presentation, and platform controls can all change. The honest output is the evidence actually loaded—not a larger number inferred from a counter.