Skip to main content
Chat can read your own material — what you’ve published, what performed, how you write — and tell you something specific about it. Start here when you’re not sure what to write next. Asking what’s working usually produces better ideas than asking for ideas directly. Ask in your own words. You don’t need a command:
The message box with the request: analyse my recent posts, let me know what worked, what not and how to improve

​
What comes back

An analysis opens with the answer, not the workings. You get a read on your positioning, then what worked and what didn’t, each quoting the post it’s talking about.
An analysis result: a summary of what the content is performing best at, a High confidence label, then What worked and What needs work sections quoting specific posts
Two things on that screen are worth knowing about. Worked through 8 steps expands. If you want to see what it read and how it got there, it’s in that panel rather than in the answer. High confidence is a claim about the evidence, not the advice. When there isn’t enough published history to be sure, it says so instead of sounding equally certain. Then it names the posts behind the verdict, with the numbers:
Top performers listing four posts, each with impressions, reactions, comments, shares and engagement rate, and a one-line reason it performed
Low performers follows the same shape. That half is the useful one, and it’s the half most people skim. Underneath, Insights is where it stops listing posts and starts naming patterns:
An Insights list, each item a named pattern with an explanation and the specific posts it was drawn from
Each one is a named pattern, an explanation, and the posts it came from. The names are the useful part. Gated Lead Magnet Fatigue is something you can act on and watch for. “posts underperformed” is not. Insights are also where an analysis earns its keep on a second reading. The performer lists tell you what happened. This tells you why, and the reasoning is attached, so you can disagree with it.

​
What you can ask for

That last row matters more than it looks. You aren’t limited to fixed reports — you can ask an open question about your own content and get an answer rather than a template.

​
Some of these need history

Analysis works from your material, so a new workspace has less to work with:
  • Profile, voice patterns, content gaps work as soon as your brand is set up
  • Recent posts, best posts, performance need published history and synced analytics
If an analysis comes back thin, the usual cause is not enough published history rather than a bad question. Check what’s in Analytics first.

​
Getting a sharper answer

“Analyze my recent posts” is vaguer than “analyze my posts from the last 90 days”. Naming a window changes what gets looked at.
“Analyze my posts” gets a general read. “Analyze my posts — I want to know why my reach dropped in June” gets an answer to your actual question.
Analysis of your own content usually doesn’t need Viral Posts or Web Search. Turning them off keeps the answer about you. See context.
Run analyze → ideas → draft in one thread. Each step inherits what the last one found, so the draft starts from your real patterns rather than from nothing.

Find what's trending

Look outward — viral patterns and how you compare.

Get ideas and angles

Turn a read on your content into things to write.

Give chat the right context

Control what the analysis looks at.

Analytics

The numbers behind performance analysis.