Here is the most common support message we get, in its usual form: "the post says 2,400 comments, my file has 780 rows, your Instagram comment export is broken." It is a completely reasonable thing to think, and almost every time the file is right and the badge is the thing lying to you.
This post is about accuracy — what the number under an Instagram post actually counts, where the rows you cannot download went, how to verify a file yourself in two minutes, and why we show a retrieved count before you pay anything instead of quoting the badge back at you.
The number under the post is not the number of comments you can get
Treat the badge as a moderation-inclusive tally rather than a list length. It is a counter Instagram maintains about the post, and it is not a promise that the page will serve you that many comments.
Two structural reasons, before we get to the interesting one. The badge counts replies as well as top-level comments, so a thread with lively sub-conversations can show a number two or three times the length of a top-level list. And it is a cached number that lags: it moves up promptly when comments arrive and lags behind when they disappear.
The interesting reason is that it counts comments that are not publicly visible at all.
Where the missing rows go
Four routes out of the visible thread, in rough order of how much damage each does to a row count.
Filtered words. Instagram lets an account owner hide comments containing chosen terms automatically, on top of a default offensive-comment filter. Hidden comments are still attached to the post, still counted, and invisible to everyone but their author and the owner's moderation view. Meta documents the behaviour in its comment moderation docs. A brand that turned on an aggressive word list before a big campaign will have a badge that has almost nothing to do with the readable thread.
Manual hiding. Same effect, one comment at a time, usually applied to spam and abuse after the fact.
Vanished authors. When an account goes private, gets deactivated, gets deleted or is removed for breaching the Community Guidelines, its comments stop being served. Giveaway posts are hit hardest here, because giveaway posts attract exactly the throwaway accounts that get cleaned up.
Deleted comments. The commenter changed their mind. Gone from the page, slow to leave the counter.
None of these are reachable by any tool reading the public page — not ours, not an extension, not a scraper written this afternoon. They are not reachable by the official API either, which only exposes hidden comments to the owner of the media.
A real post: 1,189 on the badge, 165 in the file
The most extreme case we have taken apart was a post displaying 1,189 comments. Our export returned 165. The user reasonably assumed a bug, so we checked it four ways: our own pipeline, a second independent scraper, a logged-in cookie session, and manual scrolling in a normal browser. All four agreed on roughly 165. The other thousand were moderation-hidden — the post had a filtered-word list and had attracted a wave of spam.
That investigation changed how we build. The badge is not a target to hit, so any exporter that quotes it as an expectation is selling a disappointment. What a user needs is the real figure, early, before money changes hands.
Why we show a count before you pay anything
The flow is deliberately front-loaded. You paste a link, the thread is fetched, and you are told how many comments were actually retrieved. Only then do you decide whether the file is worth anything to you.
Three things fall out of that ordering. You never pay for a 4,000-comment badge that resolves to 300 real comments. You learn something genuinely useful about your own post — a huge gap is a moderation report you did not ask for, and several brands have discovered an over-aggressive word list this way. And the free tier means you can test the claim on any post for nothing: the first 100 comments of any public post export without a signup, which is enough to compare a file against what you can see.
Tools that quote the badge in their marketing and then serve a shorter file are not necessarily dishonest. They are just measuring the wrong thing, and passing the discrepancy to the user as a surprise at the end instead of a number at the start.
Checking a file for completeness yourself
Two minutes, no tooling, and it distinguishes "moderation gap" from "genuine bug" reliably.
- Open the post in a logged-out browser — a private window is enough. This matters: logged in, you may see comments that are only visible to you.
- Note the newest comment and scroll to the oldest you can reach. Both should be in the file.
- Pick two or three comments you can actually see in the middle of the thread and search for their text in the file.
- Sort the file by
created_atand look for a cliff — a clean cut-off with hundreds of comments after it suggests a walk that stopped early, which is a real defect.
If everything you can read logged-out is in the file, the file is complete in the only sense that matters. If it is not, send us the link; a stalled walk is a bug we can fix, and a moderation gap is not.
Replies change the arithmetic
Before concluding anything from a count, decide whether you are comparing like with like. Replies are off by default in the exporter, because expanding them multiplies rows and burns a free allowance fast. The badge, though, includes them.
So the honest comparison is: badge versus top-level rows plus reply rows. Turn replies on and re-export, and on a conversational post the two numbers usually converge to within the moderation gap. Each reply arrives as its own row with reply_to_id pointing at its parent, so you can still separate the two tiers afterwards with a single filter.
For giveaway work this is more than bookkeeping. If your rules say one entry per person, replies are usually not entries — and a tool that counts them as entries inflates your entrant list with people talking to each other.
What we measure from operating the exporter
These claims come from running the thing daily across Instagram, Facebook, TikTok, YouTube, Reddit and Threads, which gives us a comparison most write-ups do not have.
Instagram has the widest badge-to-file gap of any platform we serve. Facebook is close behind. TikTok and YouTube are usually tight, and Reddit is nearly exact because removed comments remain in the thread as visible placeholders rather than disappearing. The difference is not API quality — it is how aggressively each platform's audience is filtered and how it accounts for removed content.
Gap size tracks spam exposure, not follower count. Small accounts running their first giveaway routinely show worse gaps than large brands, because they turned on a broad filtered-word list and then attracted bot entries.
Re-exports drift in both directions. Export a post today and in a week and you will see new comments and lose old ones. We keep paid exports downloadable for 30 days for exactly this reason: the file is a snapshot with a date on it, and the snapshot is the evidence.
The counts we display are rows served, and we would rather show a small honest number than a large flattering one. Caps, column names and free-tier rules quoted in this post are the live product values. For Instagram's own account of comment visibility, read Meta's comment moderation documentation and the Instagram Terms of Use, not our summary of them.
When a short file really is a bug
Moderation is not an alibi for everything, so here are the signatures of an actual failure, which we do want to hear about.
A hard round number — exactly 100, exactly 500, exactly 1,000 — usually means a cap rather than a thread ending, and 100 specifically is the free tier doing its job. A clean chronological cliff points at a walk that stopped paging. An empty file on a post with visible comments points at the URL: a story link, a private post, or a profile URL rather than a post URL. And a file that shrinks dramatically between two exports minutes apart means the platform throttled the walk partway through, which is worth a retry before it is worth a report.
Everything else — a file that is 5% or 60% shorter than the badge, stable across retries, containing every comment you can see logged out — is moderation, and it is the same list every reader of that post gets.
Giveaway draws and the fairness problem
If you are drawing a winner, the gap stops being trivia and becomes a fairness question, so answer it explicitly rather than hoping nobody asks.
The defensible position is simple: you drew from the publicly visible entries as of a stated deadline, and here is the file. Every entrant could see the same thread you exported, and nobody — including you — can draw from comments that Instagram is hiding. State the cut-off, keep the export, and publish the entry count you used.
The indefensible position is quoting the badge as your entry count while drawing from a shorter list. That is the discrepancy someone screenshots. A spreadsheet with a timestamp column is the cheapest audit trail in marketing; see verifying giveaway entries in a spreadsheet for the mechanics.
Test it on a post where you know the answer
The fastest way to trust any exporter is to point it at a post whose comments you have read yourself. Export it, open the file, and check the three comments you remember. You will know in a minute whether you are looking at an accurate file or an optimistic one — and if the number is far below the badge, you have just learned something about your own moderation settings. The first 100 comments of any post are free with no signup; larger threads are a one-time payment, never a subscription, and pricing has the detail.
Download Instagram comments free →
Related reading: exporting Instagram comments without logging in, who can see hidden Instagram comments, spotting fake Instagram comments in an export.