Turn comments into leads: the word-in-a-comment funnel one person can run

A 97-second tutorial reel pulled 213 comments against 297 likes. That ratio does not happen by itself. It is a mechanic: ask viewers for one word, deliver the goods in a DM, and a comment becomes a contact. Here is how it works, my own numbers, and a one-evening test that costs nothing.

Published August 19, 2026 · By Dmitry Tolstoi · 9 min read

I counted the comments under someone else's 97-second tutorial reel: 213 comments against 297 likes. In the same batch of public reels, clips with 41,000 and 361,600 likes collected 639 and 957 comments. The smallest reel of the five was beating clips with a thousand times its likes on one metric, and only that one.

The whole reel hung on a single line: comment one word and the full lesson lands in your DMs. Once you notice the pattern you see it everywhere, because it converts a passive scroll into a raised hand you can actually answer.

I run a content operation alone, with AI on most of the production. I have already written about getting a video past a few hundred views; this piece is about the step after reach, where a view turns into a lead you can follow up with. No team, no ad budget, and for the first test, no software either.

Does a comment really beat a like?

In short: a like is applause you cannot follow up on, while a commenter has identified themselves and opened a channel you are allowed to message.

A view is not a lead. One of my reels passed 10,000 views and changed nothing, because nobody raised a hand. Likes are the same problem at a smaller scale: applause from people you cannot name and cannot reach. A comment is different in kind. The person identified themselves, told you what they want, and opened the one door the platform holds open: Instagram lets a professional account send a private reply to someone who commented on its post.

The algorithm angle is real but smaller than the sales pitches claim. Instagram's own ranking explainer lists how likely you are to comment among the signals it predicts for posts; for Reels, watch-through and reshares carry more weight. So treat the ranking boost as a bonus, and run the funnel for the contact. A path from content to product outlasts any single spike of reach.

How does the word-in-a-comment funnel work?

In short: a video asks for one word, a trigger catches the comment, and a DM delivers the link, so the post itself stays clean.

The funnel is three links: a video that asks for one word, a trigger that catches the comment, and a DM that delivers the thing. Nothing in that chain requires a team.

The link travels by DM for a mechanical reason: a link in an Instagram caption is plain text, nobody can tap it. The private message is the only place where the click actually happens, and it keeps the post itself free of link clutter.

Meta publishes the rules of the channel. A private reply, the official name for the DM you send to a commenter, allows one message per comment, sent within seven days; after that the conversation continues only if the person writes back. In practice the funnel answers each comment twice: a short public reply under the post, which is one more visible interaction on your reel, and the private message with the link. The viewer types one word; the system, or you with your thumbs, does the rest.

The numbers that made me take this seriously

In short: likes scale with reach while comments scale with a reason to comment, and without an ask my own reel ran at 7 percent.

Put the reference batch side by side. The reel with 297 likes: 213 comments. The reels with 41,000, 44,500, 140,800 and 361,600 likes: between 639 and 957 comments each. Likes grew a thousandfold across the batch; comments barely tripled. Likes scale with reach. Comments scale with a reason to comment, and the small reel was the only one that gave people a reason.

My own accounts told the same story from the other side. One of my reels collected 13 comments against 186 likes with no call to comment anywhere in it: 7 percent, pure organics. A public reel I benchmarked, built around a direct "comment CODE" ask, ran 4,963 comments against 8,646 likes: 57 percent. The gap between 7 and 57 is the price of not asking.

Do you need the automation at all?

In short: answer the comments by hand for one evening first, because four numbers from a manual run tell you whether software would earn its fee.

Most writeups of this funnel jump straight to the chatbot subscription. I test with my thumbs first. Post the reel with the one-word ask, keep the phone nearby for the next hour or two, and answer every comment by hand: a short public reply, then a DM with the link. It costs nothing, and it is the honest version of the test, because every reply is unique, written by a person, at human speed.

Count four numbers: views, people who wrote the word, DMs that got an answer back, clicks on the link. Then apply the decision rule I use, per thousand views. Fewer than ten people writing the word means the problem is the magnet or the audience, and no subscription fixes that. Around fifty means your thumbs stopped keeping up, and that is the moment automation starts earning its fee, and not a day earlier.

My thresholds, and when I stop

In short: my keep-going bar is 1 percent commenting and 30 percent clicking; three reels under 0.3 percent means the audience is wrong.

These are my thresholds, set for my own operation, and they promise you nothing. I keep running the funnel while at least 1 percent of viewers write the word and at least 30 percent of the DMs turn into a click on the link.

The stop rule matters more than the targets. Three reels in a row under 0.3 percent of viewers commenting, and I stop blaming the mechanic. At that point the magnet is fine and the wording of the ask is fine; I put the funnel in front of the wrong audience, or on an account those people have no reason to trust. I change the audience or the account, and I leave the magnet alone. Rebuilding the prize for a room that was never going to raise a hand is the most reliable way I know to lose a month.

Which tools and accounts does the automation need?

In short: you need a professional Instagram account, a tool built on Meta's official API, and a budget between zero and about 15 dollars a month.

When the manual run proves demand, the setup is boring on purpose. You need an Instagram professional account, business or creator, and the automation has to run through Meta's official API. That distinction decides whether you are inside the rules: Instagram has been removing inauthentic activity from apps that bypass its API since 2018, while official partner tools plug into the same private-replies channel you were just using by hand.

ManyChat is the usual pick, with a free tier to test on and paid plans from 15 dollars a month; Chatfuel offers a free plan and a 49-dollar AI tier. The robot then does the two things you did with your thumbs: posts the public reply and sends the private message. Meta caps private replies at 750 per hour per account, and the tools pace themselves well below that; Buffer's guide to Instagram DM automation describes the common convention of 200 DMs an hour and one message per person per day. If your evening test produced thirty comments, none of these ceilings will bother you for a long time.

Where does this funnel get accounts in trouble?

In short: the ask itself is allowed; trouble comes from identical messages at volume, content gated behind engagement, and automation on fresh accounts.

Meta's engagement-bait rule demotes posts that beg for reactions "for purposes other than a specific call to action". Asking for one word in exchange for a named, concrete thing is a specific call to action; "drop a comment if you agree" is bait. The wording of your ask decides which side you are on, so promise an exact deliverable and then deliver it. The spam policy draws a second line: requiring engagement "to gain access to specific, exclusive content" is against the rules. My way of staying on the safe side is to make the reel teach the point on its own, treat the DM as a convenience of delivery rather than a paywall, and never stack "follow me first" on top of the ask.

The failures you should actually expect are quieter. A DM to someone who does not follow you lands in their Requests folder, with no notification: from your side it looks sent, from theirs it does not exist, so say "check your message requests" right in the public reply. Practitioner threads repeat one script: the funnel converts for the first twenty people, then dies, because identical first messages carrying a link at volume read as a spam pattern. Keep the link out of the first message where you can, vary the wording, reply once per comment. And do not hang DM automation on a fresh account. A week-old profile that answers every comment in two seconds looks exactly like what it is; run the funnel where you already look human.

A one-evening plan to run your first word funnel

In short: pick one word and one magnet, ship one reel with a single ask, answer by hand for two hours, and let four numbers decide.

This is the run I would do tonight, in order. It assumes nothing but a phone and an account where your audience already lives.

1. Pick the word. Short, on topic, hard to mistype. GUIDE, PLAN, CODE. The first one I ran was RECEIPTS, which broke my own length rule and still worked; shorter is safer.
2. Make the magnet worth typing for. One page that solves the exact problem the reel shows. A breakdown of the ten posts that worked best in your niche beats a generic checklist, because it is specific and nobody else has it.
3. Ship one reel with a single ask. Say the word out loud and put it on screen, because feeds play muted. One ask only; a "like, follow and comment" stack reads as bait and splits attention.
4. Answer by hand for two hours. Public reply plus a DM with the link, for every single comment. Mention the Requests folder to people who do not follow you.
5. Write down four numbers. Views, words written, answers received, link clicks. The counting is one evening of honesty; skip it and you will be paying a subscription in order to guess.
6. Apply the ten-or-fifty rule. Under ten words per thousand views, fix the magnet or move audiences. Around fifty, connect an official tool and let it repeat what you just did. From that point managing the robot is its own discipline: read its transcripts weekly, the way you would read a new hire's.

Where this comes from

The 213-comment reel and its batch are public posts I counted by hand, and the 57 percent benchmark is a public reel with a "comment CODE" ask, counted the same way. The 7 percent figure, the thresholds and the ten-or-fifty rule are my own operation's numbers and rules, stated as mine and promising nothing. Platform mechanics link to Meta's published documentation. The engaged-lead framing behind all of it I picked up from Alex Hormozi's $100M Leads and rebuilt on my own data.