Audience: collecting their language and pains

AI for social media and blogging · Lesson 2 / 20

A profile built from data, not imagination

"Women 25 to 45, interested in self-development" is not an audience, it is a line from an ad dashboard. You cannot write a single hook from it. A useful profile has three parts: the words people use for their problem, what they already tried and abandoned, and what they are afraid to say out loud.

Where the raw material comes from

Your own comments and direct messages. The last 100 are the most honest source. Copy them verbatim, typos and slang included.
Reviews of other products in your field. Mid-range ratings beat glowing ones: that is where people write what was missing.
Questions under other people's popular posts. People ask about what they did not understand — that gap is what you close.
Recordings of client calls. Five transcripts teach you more than a month of guessing.

The analysis prompt

Here are 100 comments from my audience (paste verbatim).
1. Extract 15 recurring PROBLEM statements
   in their own words, no paraphrasing.
2. Mark words that appear more than 5 times.
3. Sort problems into: don't know HOW / don't believe
   IT WILL WORK / no time / no money.
4. Name 5 fears that are only hinted at.

Point four returns the most. People openly write "I don't have time"; between the lines they mean "I am afraid of looking stupid if I try and fail." A post that names the second fear gets read to the end.

A working glossary

Keep a two-column file: how they say it, how I used to say it. "I can't get started" instead of "procrastination", "I don't know where to begin" instead of "lack of onboarding". Run every draft through this glossary before publishing and swap each right-column word for the left one. It is the cheapest edit you can make and the one with the most visible effect on read-through.

Insight. A model fed 100 real comments finds phrases you read yourself and never noticed: humans see meaning and skip repetition, machines count repetition and ignore meaning.

Common mistake. Asking AI to invent an audience profile with no data. It returns an averaged description from its training distribution, you believe it, and you spend six months writing to the wrong people.

Pro tip. Every quarter, rerun a fresh batch of 100 comments and compare the lists. A new phrase entering the top means your audience shifted faster than your content did.

Cheat sheet

A profile is words, attempts and fears. Demographics are useless.
Raw material comes from your DMs, others' reviews and transcripts.
Unspoken fears matter more than stated problems.
The "them vs. me" glossary is applied before every post.

1. What makes up a working audience profile?

2. Why can't you ask AI for a profile without data?

3. What yields the most in comment analysis?

Task — checked by AI

Collect 100 real statements from your audience: comments under your posts, direct messages, questions under other people's posts in your field. Run them through the analysis prompt from the lesson. Write out 15 problem statements verbatim in their words, mark the 5 most frequent words, sort the problems into the four causes, and name 5 unspoken fears. Separately, build a glossary of at least 8 pairs of 'how they say it — how I said it' based on your own recent posts.

← Back

🔒 Answer the question correctly to move on to the next lesson.