How to analyze Instagram data with ChatGPT (no CSVs, no screenshots, no spreadsheets)

Anyone running social knows the routine: open Instagram, export a spreadsheet, paste into ChatGPT, ask. Every week, all over again. Kondado offers a different path: you replicate your Instagram data to your destination and ChatGPT reads it directly through MCP, no CSV in the middle. The question shifts from "how do I export this" to "what do I want to know right now".

Summary

• Kondado connects Instagram to ChatGPT via MCP using 7 canonical tables in your destination, with literal names like instagram_media_insights and instagram_account_daily_insights.
• The setup takes 3 steps: create the Instagram integration, choose Via Kondado as the destination, and point ChatGPT to https://mcp.kondado.io/mcp.
• Analysis becomes a conversation: natural language prompts about views, reach, follows and saves query the same fields that feed your dashboards.
• Meta API quirks become technical advantages when documented (Learnings 96, 109 and 174) and protect your analysis from silent errors.

Why replace the CSV with ChatGPT reading your Instagram directly?

Spreadsheet analysis has three structural problems. The time window freezes at the moment of export. Context is lost in copy and paste. And any cross-source check requires a second spreadsheet, formulas, manual review. When ChatGPT reads the data through MCP, all three disappear: the window moves with the question, the context lives inside the tables themselves, and cross-source checks are just a query.

Kondado delivers this model ready-to-use: the same Instagram connector that feeds dashboards in Power BI or Looker Studio also answers questions in ChatGPT. Not two products—just the same data in two reading paths.

How to connect Instagram to ChatGPT in 3 steps?

The flow has three stages and none requires code.

1. Create the Instagram integration in Kondado. Go to the Kondado data sources page and select Instagram. Authentication happens via Facebook OAuth and requires an Instagram Business or Creator account linked to a Facebook Page where you are Admin, Editor, or Owner. The scopes requested are instagram_basic, instagram_manage_insights, pages_read_engagement, pages_show_list and ads_management.

2. Choose Via Kondado as the destination. When creating the integration, select Via Kondado as the destination: it is Kondado's managed data store that serves both dashboards and MCP. You don't need to configure Postgres, BigQuery or Redshift yourself to get started. If you already use a supported destination, the same integration runs there.

3. Point ChatGPT to the MCP endpoint. In ChatGPT (or any MCP client), configure an MCP server with endpoint https://mcp.kondado.io/mcp and the credential generated in your Kondado account. From there, questions like "how many saves did my last 20 posts get" or "what was last week's reach compared to the previous week" are translated into KSQL queries and answered in natural language.

What you find in Instagram's 7 tables

The Kondado Instagram connector exposes 7 table groups organized by function. Each contains fields that feed both dashboards and ChatGPT answers.

1. instagram_profiles

Profile data: username, business ID, follower count (followers_count). Useful for normalizing business IDs and as a reference when follower_count fails.

2. instagram_media

Each post: media_id, media_type (IMAGE, VIDEO, CAROUSEL_ALBUM, STORY, REELS), caption, timestamp, like_count, comments_count. For third-party media includes Business Discovery fields.

3. instagram_account_daily_insights

Daily metrics: impressions, reach, profile_views, follower_count (delta of new followers, not total). Available only for accounts with 100+ followers.

4. instagram_audience_insights

Demographic data: age, gender, cities, countries. Percentage values (0-1) by dimension.

5. instagram_media_insights

Per-post metrics: reach, impressions, engagement, saved, video_views, story_exits (for Stories). Includes separate Story metrics.

6. instagram_third_party_profiles

Competitor metrics via Business Discovery: target_username, target_followers_count, target_media_count. No additional authentication required.

7. instagram_third_party_posts

Detailed competitor post metrics: target_media_id, engagement metrics, timestamps.

Practical prompts for analyzing your Instagram with ChatGPT

With data replicated to Via Kondado, you can ask natural language questions. Here are practical examples:

The same fields run in Power BI, Looker Studio or Tableau. ChatGPT just changes the reading interface.

Which Meta API quirks become technical context in your answers?

The difference between shallow analysis and reliable analysis lies in knowing the edges of the Meta API. Three quirks documented by Kondado change how you read the numbers.

Same name, different data: follower_count vs followers_count The follower_count field in instagram_account_daily_insights is the daily delta of new followers: accounts that started following that day. The followers_count field in instagram_profiles is the accumulated total. Confusing the two inverts growth reading. Bonus: Meta's API returned zero for two long periods (July 2025 and March 2026) for follower_count, with 200 OK and empty data. Every ETL tool on the market was affected in the same window. Kondado documents instagram_profiles as an alternative source for followers_count when follower_count breaks in the API, with the snapshot intact.

Authentication and New Pages Experience When the linked Facebook Page migrates to New Pages Experience (NPE), page-scoped endpoints stop accepting the user token. Migration happens in silent waves: a Page can be on the classic model on Wednesday and migrated by Thursday. Kondado does the automatic gate in the integration: if the call fails due to NPE, collection tries again with the page token. What does not happen with generic ETLs is the logging of which mode the Page was in at collection time. Kondado includes this metadata in the log tables for troubleshooting.

API limitation: followers are not attributed to campaigns Meta does not expose follower acquisition attributed to ad campaigns via Instagram Graph API. The data exists in Ads Manager but is not available for programmatic extraction via the Instagram API. When you ask "which followers came from that traffic campaign", ChatGPT will respond that there is no table for this — and will suggest crossing instagram_account_daily_insights with campaign flight dates as a proxy. Other ETL tools, if they don't document the limitation, return zero or empty data without explaining why. Kondado documents in each table's __kdd_about which metrics are deltas, which are snapshots, and which depend on conditions (100+ followers for demographic data, for example).

Conclusion

Kondado makes ChatGPT a direct reader of your Instagram data, eliminating the CSV step. The 7 canonical table sets, the MCP endpoint https://mcp.kondado.io/mcp and the proactive documentation of Meta API quirks turn every question into a conversation instead of an export. Start your 14-day free Kondado trial and set up your first Instagram connector today.