Every brand on Instagram has a notifications tab, and every brand on Instagram believes that tab shows them what people are saying. It does not. It shows a specific, narrow slice: posts where your account was tagged, comments where someone typed your @handle, and story mentions. Everything else — a customer typing your brand name in plain text under a creator Reel, a comparison thread under a competitor post, a complaint on your own post from three weeks ago that nobody scrolled far enough to see — arrives nowhere.
That gap is the whole subject of this guide. Not the tools-list version of brand monitoring, but the mechanics: what Instagram surfaces, what it structurally cannot surface, what the official API adds, and how to build a system that catches the rest without a five-figure enterprise contract.
The three kinds of Instagram brand mention
They behave completely differently, and conflating them is why most brand monitoring feels both busy and blind.
Tagged mentions
Someone uses your @handle in a caption, a comment or a story, or tags your account in a photo. These reach your notifications and, if you run a Business account, the API. They are also the smallest of the three categories in almost every brand we have looked at, because tagging requires the person to know your exact handle and to want you to see the post.
Untagged mentions
Someone writes your brand name as plain text. No notification exists, on any account, ever. This is where the honest opinions live, precisely because the writer does not expect you to read them. A customer complaining under a review Reel is not tagging you; they are telling other customers.
Comment mentions on posts you do not own
A creator posts about your category. Two hundred comments arrive. Eleven of them name your brand, four of them ask whether it is worth the money, one of them says your support team never replied. You own none of that surface, and no product Meta ships will push it to you.
A rough working ratio from comment exports across consumer brands: for every tagged mention, expect somewhere between five and twenty untagged ones. The exact number depends on how memorable your handle is versus your name. Brands whose handle differs from their name — an added underscore, a country suffix, an _official — sit at the worst end of that range, because nobody remembers the handle and everybody remembers the name.
What the official API actually covers
Meta does provide mention functionality. It is worth knowing exactly where it stops before you build on it. The Instagram Platform mentions documentation describes an endpoint that lets a Business account read media and comments where it was @-mentioned, and reply to them. Three limits matter:
- It requires the @handle. The mention has to be an actual tag. Plain-text brand names produce no mention object, so there is nothing to fetch.
- It requires a Business or Creator account connected to a Facebook Page, plus app review for the relevant permissions. This is a real integration project, not an afternoon.
- It cannot read arbitrary public posts. You can read comments on media you own and on media where you were mentioned. You cannot point it at a competitor Reel and read the comments, which is exactly where comparison intent lives.
So the API automates the easy half — the mentions you would have seen in notifications anyway — and does nothing for the half that carries the information. We wrote up the same boundary in more depth in the Instagram Comments API guide. The practical conclusion is the same either way: whatever you use for tagged mentions, untagged mentions need a different mechanism.
The system, end to end
What follows is a weekly routine that takes roughly an hour once it is set up. It is deliberately unclever. Clever monitoring systems get abandoned in month two.
Step 1 — Write down every string that means you
Open a sheet and list, in one column: your handle, your brand name, the two or three misspellings people actually type, every product name, your slogan if it is distinctive, and your branded hashtags. Add the shortened form your customers use even if you hate it — if people call it something, that something is a search string.
This list is the entire filter. Brands that skip this step and search only for @handle catch maybe a fifth of what is out there.
Step 2 — Decide which posts to watch
Mentions are not evenly distributed. Four sources cover nearly all of them:
- Your own recent posts and Reels. Obvious, and still under-read: on any post above a few hundred comments, most of the thread is behind Load more and has never been read by anyone at your company.
- Creator posts featuring you. Whether paid or organic. The comments under a sponsored Reel are the closest thing you get to a focus group reacting to your positioning in real time.
- Competitor posts. Especially launch posts and anything comparison-shaped. Your name appears in these threads far more often than most brands expect.
- Category posts. Review accounts, roundup Reels, "best X for Y" content. Nobody is tagged, everybody is named.
Twenty to forty post URLs is a realistic watchlist. Keep them in the same sheet, one per row, with a column for the last date you exported it.
Step 3 — Export the threads
Paste each URL into the Instagram comment exporter, enable replies, and download as CSV or Excel. Each row carries author name, username, avatar URL, comment text, likes, reply count, timestamp, language, pinned status, comment ID and parent comment ID. The parent ID is what keeps replies attached to the comment they answer instead of arriving as a flat unstructured list — which matters here, because complaints usually live in replies rather than top-level comments.
Practical notes: use the canonical instagram.com/p/CODE/ or instagram.com/reel/CODE/ URL; public posts only; and expect the exported count to differ by a few percent from the number Instagram displays, since deleted, hidden and filtered comments account for the gap. If you need many posts in one pass, bulk exporting Instagram comments covers the workflow.
The first 100 comments of a post are free with no signup, three exports a day. Past that it is $1 per 100 comments with a $3 minimum, or a one-time $14 3-Day Pass covering unlimited posts up to 10,000 comments each. Pro Passes are $49 for 3 days, $349 for 7 days and $1,499 for 30 days, lifting the ceiling to 100,000 comments per post and adding AI analysis. Each is a single purchase that expires by itself — there is no subscription. For a monthly monitoring cycle, one pass per cycle is the normal shape.
Step 4 — Filter, do not read
Open the combined export and filter the text column against your watchlist. In Sheets, a formula in the shape of =SUMPRODUCT(--ISNUMBER(SEARCH(brand_list, text_cell)))>0 flags any row containing any of your strings. In Excel, the same idea with COUNTIF over a named range. Nothing more sophisticated is required, and anything more sophisticated tends not to get maintained.
Copy the matched rows into a running mentions sheet with the post URL and export date appended. That running sheet, not the individual exports, is the asset. It is what makes month-over-month comparison possible.
Step 5 — Triage into three buckets
Every mention is one of three things, and each has a different owner and a different clock:
- Problems — complaints, support requests, delivery issues, anything angry. Same-day or next-day response. There is good evidence that responding at all changes outcomes: a Harvard Business Review study of review responses found that businesses which started replying subsequently received better ratings, and that the effect came partly from unhappy customers revising or withdrawing complaints.
- Intent — "where do I buy this", "does it ship to X", "is it worth it versus Y". These are leads sitting in public with nobody assigned to them. Route to whoever owns sales, and treat the recurring ones as a content brief.
- Advocacy — praise, user-generated content, people defending you in an argument. Route to the creator and community queue. This is also your cheapest source of social proof, and consumers weight it heavily: BrightLocal consumer research consistently finds that a large majority of people read peer opinion before purchasing.
Step 6 — Compare runs, not totals
A mention count on its own tells you almost nothing. The signal is in the difference between this run and the last one:
- A complaint theme that was absent last month and appears twelve times this month. That is a product or logistics change, and it is usually visible in comments before it is visible in support tickets.
- A competitor name appearing in your comment threads more often. That is a paid campaign of theirs landing on your audience.
- A question repeating across unrelated posts. That is a gap in your product page, and it is a content brief with the demand already proven.
- A spike with no campaign behind it. Find the source post before deciding whether it is good.
Reading the mentions you catch
Once the matched rows are in front of you, three passes cover most of the value.
Sort by likes. A complaint with four hundred likes is not one customer, it is a consensus with an audience. A comment with two likes may be true and is not yet a public problem. Likes are the cheapest priority signal in the file.
Count the nouns. Strip out the filler and count what remains: sizes, shipping windows, prices, features, competitor names. Frequency is the finding. Ten people asking the same question is a landing-page fix; ten people asking ten different questions is a healthy thread.
Read the replies to your own replies. Whether your response actually settled the matter is visible in the thread, and it is the only honest measure of whether your community management works. The parent comment ID column makes those chains reconstructable.
If you want to score tone systematically rather than by eye, Instagram comment sentiment analysis covers doing it without producing a meaningless positivity percentage, and how to analyze Instagram comments covers turning raw text into themes.
When to stop doing this by hand
The routine above is an hour a week for one brand and twenty to forty posts. Two things break it. The first is volume: a brand watching several markets, a dozen creators and three competitors is looking at hundreds of post URLs per cycle, and pasting them one at a time stops being a system and starts being someone's Monday. That is what bulk comment export exists for — send the URL list, profiles or hashtags, get one merged CSV or Excel back with the source URL on every row, priced per job at well under the self-serve rate.
The second is that your watchlist is not generic. A brand monitoring problem usually has specifics — particular markets and languages, a competitor set that changes, an alerting threshold that matters to your team, a delivery destination that is Slack or a warehouse rather than a download. We build that: a monitoring pipeline scoped to your brief, run on our infrastructure on the cadence you need, with the alerting and analysis layer on top. Managed data and monitoring covers what that looks like, including taking over a listening setup you already have. If you have a specific need and want to know whether it is buildable, tell us what you are trying to watch — the answer is usually yes and the scoping conversation is short.
What monitoring will not tell you
Some honesty about the ceiling, because brand monitoring is routinely oversold.
Comments are not a representative sample. People who comment are more engaged and more opinionated than people who do not. Comment data tells you what your loudest customers think, which is genuinely useful and is not the same as what your average customer thinks. Treat it as qualitative research with a large N, not as a survey.
Stories and DMs are out of scope. A meaningful share of brand conversation happens in story replies and private messages, and none of it is public or exportable. Anything claiming to reach it is either wrong or asking for your credentials.
Private accounts are out of scope, permanently. No tool reaches them, and any service claiming otherwise is a scam.
Sentiment scoring is approximate. Sarcasm, in-jokes and multilingual threads all break naive scoring. The language column helps you segment before scoring, which is most of the fix, but a sentiment number should never be reported to three decimal places.
Where the line is
Public comments are public, and that does not make them consequence-free to process. Comment data identifies individuals, so it falls under the GDPR and comparable regimes; the EDPB guidance is the reference point for what lawful processing looks like.
In practice, for a brand monitoring programme: analyse in aggregate, quote sparingly and with attribution when quoting publicly, keep exported files access-controlled rather than in a shared drive everyone can open, set a retention period and actually delete on it. Do not build ad-targeting lists from commenters. Do not cold-DM someone because they mentioned you — replying in public to a public comment is fine and expected; a sales DM triggered by monitoring is not. None of the useful work above requires crossing any of those lines.
Adapting this to other platforms
The mechanics transfer with one change: which surface carries the untagged conversation. On TikTok the same system runs on video comment threads, and we wrote the platform-specific version in how to monitor TikTok brand mentions. If you are vetting creators rather than watching your own name, brand safety checks through creator comments is the adjacent workflow, and the agency guide to exporting Instagram comments covers running this across a client roster rather than one brand.
Related reading
- Instagram competitor analysis — reading a rival audience from their own comment sections
- How to analyze Instagram comments — from raw rows to themes
- Instagram comment sentiment analysis — scoring tone honestly
- Instagram Comments API — what the official API can and cannot reach
- Bulk comment export — hand over the URL list, get one merged file back
- Instagram Comment Exporter — first 100 comments free, replies included, no signup