How do I analyze Facebook posts with ChatGPT, Claude, or Gemini?

PostSnag exports plug directly into ChatGPT, Claude, or Gemini: export a profile to Markdown or CSV, paste or upload the file into whichever AI tool you use, then ask it a specific question about the posts. That turns a folder of raw captures into a content plan, a competitor breakdown, or a swipe file in minutes, instead of an afternoon spent scrolling and taking notes by hand.

This is one of the main reasons people use PostSnag. Capturing and exporting gets you the raw data; handing that data to an AI tool is what turns it into something you can act on the same day.

How to do it, step by step

1. Capture the profile with the extension first, if you haven't already. Open it on Facebook, scroll, and click Export To Dashboard. See How do I send my captured posts to the dashboard? (Export To Dashboard).

2. Open the profile in your PostSnag dashboard and filter it down to whatever you actually want to study: a competitor's best month, one post type, or their full history.

3. Click Export and choose Markdown (the more AI-friendly format for most questions) or CSV.

4. Open ChatGPT, Claude, or Gemini in a browser tab or app.

5. Paste the file's contents directly into the chat, or upload the file if the tool you're using supports attachments.

6. Ask a specific question. The more precise the prompt, the more useful the answer; "what do you think of these posts" gets a vague answer, a targeted question doesn't.

[Screenshot: A Markdown export pasted into an AI chat interface alongside a specific prompt]

Example prompts to try

Once your export is in the chat, adapt one of these to what you're actually deciding:

Find the pattern behind what's working

• "Find the common structure of the top 5 posts by reactions. Look at hook style, caption length, and where the call to action lands."

• "What do the highest-engagement posts in this export have in common that the lowest-engagement posts don't?"

Compare post types

• "Group these posts by type and rank each group by average reactions, comments, and shares. Which post type is actually carrying this profile?"

• "Is video outperforming photo and text here, or does it just look that way because there's more of it?"

Turn history into a plan

• "Based on what worked in this export, draft a two-week content calendar with 10 post ideas, each modeled after a specific post from the data."

• "Write three caption openers in this profile's voice, based on the captions that got the most engagement."

Build a swipe file

• "Pull the first line of every caption from the 10 highest-engagement posts. I want a list of hooks I can adapt for my own content."

• "List every post here that used a question as a hook, along with its reaction count."

Spot timing patterns (CSV works well for this one)

• "Look at the Posted column in this CSV. Is there a day of the week or time pattern where reactions are consistently higher?"

Compare a competitor to yourself

• "Here's a competitor's export. My own recent posts cover [describe your content]. What is this competitor doing that I'm not?"

Adjust the wording to your niche and your actual goal. An AI tool can only work with what you give it, so a prompt that names what you're deciding, a content calendar, a hook list, a format decision, beats a generic "analyze this."

Markdown or CSV for AI work

Either format works, but they suit different questions:

• Markdown tends to read more naturally when you're asking about tone, structure, or "what's the pattern here" questions, since each post is already a labeled, self-contained section.

• CSV is better when you want the AI tool doing structured, row-by-row comparisons or basic math across a large number of posts, like ranking by engagement or checking for a day-of-week trend.

See Markdown or CSV: which export should I use? for the full comparison.

A few things that get better answers

• Export a real sample, not a handful. A few posts won't give an AI tool much to compare. A few dozen does much better.

• Say if you're working from a capped export. If a profile is on the Free plan's 30-post cap, mention that in your prompt so the AI tool doesn't assume it's seeing that profile's complete history.

• Give the AI tool your goal, not just the data. "I run a similar page and want to grow engagement" gets a more useful answer than pasting the file with no context at all.

• Ask follow-up questions. Treat the first answer as a starting point. Ask the AI tool to go deeper on one finding, or to re-rank by a different metric, rather than expecting one prompt to do everything.

What this doesn't do

PostSnag doesn't include a built-in AI analysis feature of its own; it captures and exports the data, and you bring your own AI tool to interpret it. It also has nothing to hand over on a per-reaction-type basis: exports report a total reaction count only, not the like, love, and haha split, so a prompt asking the AI to analyze which specific reaction type a post got won't have anything to work with. See What's included in a PostSnag export? for the full list of what's in the file.

Common questions

Do I need a paid ChatGPT, Claude, or Gemini account for this?
No. A free-tier AI tool account can read a pasted or uploaded export, though very large exports may hit that tool's own file size or message length limits.

Should I use Markdown or CSV for AI analysis?
Either works. Markdown tends to read more naturally for pattern and tone questions; CSV is better if you want the AI tool doing structured, row-by-row comparisons.

Can I upload the export file directly instead of pasting it?
Yes, if the AI tool you're using supports file uploads. Otherwise, open the file and paste its contents into the chat.

Does PostSnag have its own built-in AI analysis?
No. PostSnag captures and exports the data; you bring your own AI tool to analyze it.

Can the AI tell me the reaction breakdown for a post?
Not from an export. Exports only include the total reaction count, so an AI tool working from the file can't break that number down into like, love, or haha. Check the dashboard's panel or Analytics for that.