Keyword Triggers: Choosing Words Your Audience Actually Types

A keyword is the smallest field in the automation builder and the one that quietly decides everything after it. It looks like a setting. It is actually a guess about what somebody types with one thumb while half-watching your Reel, and every reply you send inherits that guess.

Get it right and the automation feels like it is reading minds. Get it wrong and it either sits silent while comments pile up, or it answers people who never asked a question.

Start with the words that are already there

The temptation is to invent a keyword. SHOPNOW, INFO2026, something clean. Then you have to teach an audience to type it, which is a second job on top of the first.

The better move is to open your own comments from the last couple of months and read what people already write when they want something. It is nearly always one of three or four words, and it is nearly always shorter and blunter than the word you would have chosen. price. link. size. dm.

Three properties make a keyword survive contact with a real audience:

Case and punctuation are already handled

Matching ignores both. Price?, PRICE, price. and Price are one keyword, not four, so there is no reason to list variants of the same word in different shapes.

What matching does not do is read minds about spelling or language. Those you list yourself, and this is where most keyword sets are too thin. A Hinglish-heavy audience types daam and kitne alongside price. People type prize when they mean price often enough that leaving it out is a decision, not an oversight.

Any-of versus all-of

The two matching modes are not difficulty levels. They answer different questions.

In practice most automations should be any-of, because most posts have one intent and the cost of missing somebody is higher than the cost of answering somebody loosely. All-of earns its place when a single word is genuinely overloaded, and the honest alternative is usually to pick a less overloaded word instead.

Negative keywords are the safety rail

Negative keywords say do not reply, even if a trigger word is present. They are the field people skip on the first build and add after the first embarrassing send.

Three shapes are worth having from the start. Words that reverse the intent, like not interested or no thanks. Words that mean the person has already bought, so the sales reply reads as tone-deaf. And words that mean the comment is about someone else entirely, which is what happens when a comment tags a friend and mentions your keyword in passing.

The test for a negative keyword is simple: would receiving your reply after this comment make you wince? If yes, it goes in the list.

When two automations could both match

Once you have more than one automation running, overlap becomes the thing that bites. Two triggers on the same post, or two posts in the same campaign sharing a keyword, and one person's comment satisfies both.

They do not both send. The gate that runs before every reply includes a dedupe check and a cooldown check, and a blocked reply is recorded with its reason in Activity. So the failure is visible rather than silent, which is the good news.

The fix is not to work out which automation wins. It is to make the question never arise. Scope each automation to specific posts under Advanced, keep keyword sets disjoint between overlapping automations, and treat a growing dedupe count as a design smell rather than a log entry. The full order the checks run in, and why each one exists, is documented separately.

A keyword is an audience-intent decision

Here is the part that is not about software at all.

The keyword you choose sorts your audience before you ever speak to them. A broad word catches everyone who is mildly curious, which means high volume and thin intent. A specific word — book, quote, size 9 — catches fewer people, all of whom have already told you something concrete about what they want.

That choice determines what your private reply can even say. If the keyword is price, the reply can be the price. If the keyword is info, the reply has to be generic, because you do not know which information they meant. Vague keywords force vague replies, and vague replies are the ones that get ignored.

This is also why keyword choice and funnel design are the same conversation rather than two settings. The word decides the intent, the intent decides the message, and the message decides whether anyone answers. If you are laying out the whole path from comment to conversation, start with the flow and let it tell you what the keyword needs to be.

Before you go live

Run Test on myself. It sends the real thing through the real pipeline to your own account, so a keyword that matches nothing shows up as a test that never fires rather than as a quiet Saturday.

Then turn it on for one post and read the blocked list that evening. Keywords are the one setting you should expect to edit twice in the first week, and the comments you get are better evidence than anything you can reason out in advance. If none of this is set up yet, the first-week setup guide walks the same ground more slowly.